2026-11-02
- Pennsylvania Senate passes bipartisan AI chatbot child-safety legislation (awaiting House action)Regulatory
- Sen. Cassidy (HELP Chair) unveils PATIENT AI Act discussion draft on AI in health careRegulatory
- US Copyright Office DMCA triennial rulemaking confirms AI training-data mining exemption; proposes new classesRegulatory
- India TRAI issues draft guidelines on Responsible Use of AI in Telecommunication NetworksRegulatory
- UK High Court rejects Meta, TikTok and X challenge to Ofcom Online Safety Act information noticesRegulatory
- FTC opens Section 6(b) inquiry into personalized pricing and worker compensationRegulatory
- From Meters to Machine Learning: Building AI-Ready Utilities in IndiaStrategic Report
- Who Will Pay the $2.6 Trillion AI Infrastructure Bill?Strategic Report
- The Macroeconomic Impact of AI Access DisruptionsStrategic Report
- A detailed, affordable plan for transition insurance in the AI eraStrategic Report
- Research: 'Agent in the Middle' — agent-mediated access preserves ~91% of web vulnerability exploitability (arXiv 2610.35541)Vulnerability
- ag-ui-protocol HttpAgent information disclosure via remote manipulation (CVE-2026-120233)Vulnerability
- Fickling pickle-safety-gate bypass lets crafted ML model files pass as LIKELY_SAFE and execute code on load (CVE-2026-118938)Vulnerability
- Prefab UI Generative UI sandbox escape allows remote code execution from LLM-written UI code (CVE-2026-120338)Vulnerability
- IETF draft-jakab-dawn-agent-discovery-mdns-01: Zero-Configuration Agent Discovery for A2A and MCPGuidelines
- IETF draft-manyfolks-agentproto-arch-00: Architecture for AI Agent Communications on the InternetGuidelines
- Anthropic expands Claude Security public beta — continuous repository scanning, automatic fix tracking and editable threat modelsSolutions
- NVIDIA open-sources SkillSpector — Apache-2.0 security scanner for AI agent skillsSolutions
- OX Security launches Agent Activity (early access) — runtime step-by-step tracing and guardrails for AI agentsSolutions
- Vicarius launches vRx 2.0 — remediation-first platform with AI red-team validation and agentic pipelineSolutions
- Snowflake makes Restricted Session Scopes generally available — task-scoped privilege ceilings for AI agentsSolutions
2026-11-01
- US House introduces bipartisan Congressional Office of Technology bill to give Congress independent analysis of emerging tech incl. AIRegulatory
- NO FAKES Act of 2026 (federal digital-replica / voice & likeness rights) advances through Senate committee to full chamberRegulatory
- US House introduces 'Protecting Kids from Human-Like Chatbots Act' barring human-like default behaviour of chatbots used by minorsRegulatory
- US government approves OpenAI's broad GPT-5.6 rollout after delaying release on national-security requestRegulatory
- France's CNIL fines Fnac Darty €7M over in-app tracker used to personalise product recommendations and train the recommendation algorithmRegulatory
- UK Ofcom identifies four 'highly effective' age-assurance methods for under-16 social media ban; ICO sets data-protection guardrailsRegulatory
- Stirling-PDF AI agent endpoint passes unsanitized HTML to WeasyPrint (create-pdf-from-html-agent)Vulnerability
- RAGFlow agent chat-completions access-control bypass: agent invoked before the caller's agent-access check runsVulnerability
- TAKO AI Agent for Okta: AI-content-filter bypass lets any authenticated user execute arbitrary Python as root in the agent containerVulnerability
- TAKO AI Agent for Okta: unauthenticated attacker can create the first administrator account (full AI-agent tenant takeover)Vulnerability
- MoLing computer-use MCP server: unauthenticated SSE endpoints + CORS wildcard allow browser cross-origin session hijack to arbitrary command executionVulnerability
- How Middle Powers Can Survive the AI RaceStrategic Report
- An updated Model Spec for more capable and beneficial AIStrategic Report
- Gated Exploitation: Frontier Cyber Capability as a Lab-Controlled PrivilegeStrategic Report
- Open-weight AI models can exploit the patch gapStrategic Report
- CISA updates SCuBA baselines: ScubaGoggles v1.1.0 adds enhanced Gemini (generative AI / Super Intelligence) secure-configuration controlsGuidelines
- MLCommons launches Candela benchmarking-as-a-service; releases AILuminate Safety 1.2.1 and Jailbreak 1.1 with new jailbreak taxonomyGuidelines
- Check Point (Lakera) ships AI Agent Security in early access — agent discovery and risk assessment on top of AI GuardrailsSolutions
- Sweet Security launches Virtual Patching for running cloud and AI workloads, tied to its autonomous AI red-team agentSolutions
- Resemble AI launches Agent Detection (RAD-S1) — behavioral model that flags AI agents appearing humanSolutions
- CrowdStrike introduces Agent Graph — endpoint-native causal tracing from AI prompt to impactSolutions
- CrowdStrike extends Falcon Adversary OverWatch 24/7 hunting and Falcon Complete MDR to the AI/agent attack surfaceSolutions
- Fortinet launches FortiGuard Inline AI Security Service — FortiAIGate guardrails delivered inline to FortiGateSolutions
- Oracle launches AI Security Mesh, a managed AI security-testing platform powered by OpenAI cyber modelsSolutions
2026-10-31
- Drupal AI (Artificial Intelligence) module: batch of XSS, authorization and resource-injection flawsVulnerability
- DBHub MCP server: read-only SQL validation bypass on MySQL/MariaDB connectors (comment-smuggling)Vulnerability
- Zuke MCP registry accepts command-form descriptors without ZUKE_REGISTRY_LAUNCH_HOSTS authorizationVulnerability
- AI Engine (WordPress MCP) OAuth 2.1 Dynamic Client Registration endpoint missing authorizationVulnerability
- ServiceNow AI Platform persists a user's plaintext password in a system table (information disclosure)Vulnerability
- Chat AI with Dispatcher WordPress plugin: unauthenticated arbitrary file upload (spoofable content type) leading to RCEVulnerability
- PDF Reader MCP server logs that X-API-Key auth is enabled but never validates it (unauthenticated access when HTTP transport on)Vulnerability
- praktor multi-agent Claude Code orchestrator: Telegram document filename concatenated into destination path (arbitrary path write)Vulnerability
- Paperclip agent API-key holders can set cleanupCommand/teardownCommand to achieve RCE when a workspace is archivedVulnerability
- Flowise sandbox escape via allowlisted langchain/mammoth modules allows OS command execution (ALLOW_BUILTIN_DEP=true)Vulnerability
- LM Studio model downloader path traversal lets a crafted model repository write files outside the models directoryVulnerability
- GitLab CE/EE integer overflow in CI/CD regex compiler grants unauthenticated RCE (CISA KEV, confirmed in-the-wild)Vulnerability
- n8n RCE via hidden Function / Function Item / LangChain Code nodes running outside the task runner (vm2 in main process)Vulnerability
- Anthropic unreleased frontier model autonomously chained exploits across Chrome, Safari, Android, Windows and LinuxVulnerability
- EMBRACE: embedded-browser flaws let code execution / file theft / address spoofing in five AI coding tools (Cursor, Copilot, Codex, VS Code, Claude Desktop)Vulnerability
- Community safety report and new protections for teensStrategic Report
- OECD Digital Economy and Society Outlook 2026Strategic Report
- Artificial Intelligence sector study 2025Strategic Report
- AI adoption survey report 2026Strategic Report
- House Democrats introduce resolution opposing any federal bailout of AI-related firmsRegulatory
- Bank of England and FCA publish fourth joint survey on AI in UK financial servicesRegulatory
- European Commission freezes national social-media age-limit laws for 12 months under KIDS Act TRIS standstillRegulatory
- New York lawmakers introduce 'RAISE Act 2.0' (S.10731) to let state 'pace the AI frontier'Regulatory
- ASD's ACSC publishes Annual Cyber Threat Report 2025-26, identifies AI as the 'defining force', and releases the Essentials framework series (evolving the Essential Eight)Guidelines
- OWASP GenAI Security Project publishes 'LLM Exploit Generation Guide 2.0' (Phase 2: Agentic vs. Classic Offensive Approaches)Guidelines
- TrendAI analyzes three AI-driven campaigns in Japan, incl. CyberStrikeAI orchestrating 44 agents to automate attacksSolutions
- Check Point launches Maestro Hyperscale Orchestrator 200 to secure AI-scale (largely encrypted) trafficSolutions
- Microsoft publishes first staged assessment of AI adoption across eight North Korean threat groups (incl. ARTEX use)Solutions
- Anthropic and collaborators show 0.1% pretraining poisoning can implant a remote code execution backdoor in an agentSolutions
- OpenAI publishes its layered frontier-model misalignment detection stackSolutions
- Cursor marketplace served a tampered Claude Code extension that dropped an SSH backdoor (up to ~5,000 users)Solutions
- Microsoft 365 E7 expands AI-agent security with Dynamic Permissions, Dynamic Agent Risk and Dynamic Conditional AccessSolutions
2026-10-30
- functype-mcp-server command injection via unconstrained version string leads to remote code executionVulnerability
- AnkiMCP globally disables FastMCP DNS-rebinding protection, exposing MCP server to remote clientsVulnerability
- Coder workspace agent GCE instance-identity JWT audience validation failure allows agent session-token replayVulnerability
- SuperAGI cross-organization access-control failures expose other tenants' agents, executions and configsVulnerability
- modelscope agent-zero directory traversal in FileBrowser.save_text_file (agent can write arbitrary files)Vulnerability
- Strands Agents Tools environment tool secret-masking bypass exposes agent host cloud credentials via prompt injectionVulnerability
- ServiceNow AI Platform Rhino sandbox escape lets an admin execute arbitrary code on the host OSVulnerability
- IBM Langflow OSS lets a low-privileged account cause arbitrary OS command execution via Agentic Assistant component codeVulnerability
- Red Hat OpenShift AI gateway payload processing injects another tenant's LLM API keys (cross-tenant credential/quota abuse)Vulnerability
- Agno PandasTools unsafe reflection lets prompt injection invoke arbitrary pandas functions (read_pickle RCE / local file read)Vulnerability
- Crawl4AI Docker API server lets an authenticated client load an arbitrary HuggingFace model and reach pickle code executionVulnerability
- Cisco Secure Firewall ASA/FTD SSL VPN unchecked null pointer dereference → device reload DoS (KEV, confirmed in-the-wild)Vulnerability
- opencode (open) OS command injection via attacker-controlled MCP OAuth authorization_endpoint leads to PowerShell RCEVulnerability
- testmetest / shai-hulud npm+GitHub worm targeting AI coding agents (execution hooks, SessionStart injection, credential theft)Vulnerability
- Pluto 'Load/Mine' cryptomining and backdoor campaign via malicious Hugging Face model repositories (trust_remote_code)Vulnerability
- Aokigahara/Aoki infostealer targets AI coding assistants and cloud credentials (in-the-wild campaign)Vulnerability
- FARM AI Act of 2026 (H.R.9686) introduced to increase access to artificial intelligence through USDA programsRegulatory
- GUARD Act (S.3062) introduced in Senate establishing requirements and limitations on publicly available AI chatbotsRegulatory
- India MeitY weighing separate frontier-model pre-release safety regime and standalone AI law; 'AI intermediary' definition under discussionRegulatory
- Singapore PM Lawrence Wong proposes international AI incident-reporting pilot at Virtual Leaders' Meeting on Frontier AI SafetyRegulatory
- New York Governor Hochul signs nation-leading law banning unsafe AI chatbot features for minors (S.9051B/A.10379C)Regulatory
- UK House of Lords defeats government on Cyber Security and Resilience Bill, writing in AISI frontier-AI audit review and AI-agent direction powersRegulatory
- The State of AI in the Enterprise: Built Differently, How AI is Reshaping the Way Enterprises Work, Grow and Create ValueStrategic Report
- Systematizing the Reporting of AI Red-TeamingStrategic Report
- MITRE ATLAS data v2026.10 — major taxonomy update (new techniques, renames, Data Format v6.1.0)Guidelines
- ETSI TS 104 275 — Technical Specification to reduce production of harmful deepfakesGuidelines
- CrowdStrike documents four agentic-AI adversary campaignsSolutions
- Kindo launches Deep Hat Sift — decision model that filters events before AI agent runsSolutions
- CrowdStrike details Guardrail-based Vulnerability Discovery (GbVD) — using model refusals to find bugsSolutions
- FireMon launches FireMon Agentic AI — agent-first network security policy investigationSolutions
- Hugging Face releases safeagents — Apache-2.0 proxy that blocks risky coding-agent tool callsSolutions
- Akamai launches Agent Sandbox (beta) — WebAssembly-isolated runtime for agent-generated codeSolutions
- Anthropic open-sources sandbox-runtime — OS-level sandbox for AI agents without containersSolutions
2026-10-29
- [KEV] Linux kernel missing variable initialization (ESP socket-buffer data corruption; confirmed in-the-wild)Vulnerability
- [KEV] Broadcom VMware ESX insufficient logging (confirmed in-the-wild)Vulnerability
- chrome-devtools-mcp insufficient policy enforcement allows local arbitrary read/writeVulnerability
- Tabularis MCP SQLite driver SQL injection via describe_table (stacked statements, read-only bypass)Vulnerability
- Google Agent Development Kit (ADK) for Python FileArtifactService cross-application artifact accessVulnerability
- Progress Telerik Convert/Merge AI agent tools path traversal (read/overwrite outside output dir)Vulnerability
- Progress Telerik Document Processing AI summarization/Q&A uncontrolled resource consumption (DoS)Vulnerability
- NVIDIA TensorRT-LLM server-side request forgery in tokenizer/async serving pathVulnerability
- NVIDIA NeMo Agent Toolkit SQL injection (CWE-89)Vulnerability
- Nuxt Ollama module exposes configured cloud API key in the client-side window.__NUXT__ payloadVulnerability
- MCP Python SDK bearer-auth resource confusion accepts a token issued for another serviceVulnerability
- MLflow CrewAI autolog leaks plaintext provider API keys into trace/OTLP outputVulnerability
- MLflow model-version prompt branch path traversal leads to server-local file read (get-artifact)Vulnerability
- MLflow third-party scorer deserialization allows unauthenticated remote code execution (scorer invoke)Vulnerability
- Germany and Japan issue joint statement on economic security and future technologies; agree high-level AI dialogueRegulatory
- UK launches consultation on Draft National Policy Statement for Data Centres (28 Oct 2026)Regulatory
- UK Autumn Budget 2026: British Data Centre Covenant and AI package tie grid priority to community/energy/security testsRegulatory
- Florida AG files civil suit against A2E AI platform over AI-generated child sexual abuse material (28 Oct 2026)Regulatory
- China publishes 15th Five-Year Plan for National Informatization: names AI law and extends model filing to AI agentsRegulatory
- China MOFCOM proposes sanctions-style 'trade secret' countermeasure targeting foreign AI/tech entities (26 Oct 2026)Regulatory
- Poland activates AI Act enforcement regime: KRiBSI gains inspection and penalty powers (28 Oct 2026)Regulatory
- Settlement and Governance Behind the AI FrontierStrategic Report
- AI readiness in central banks (IFC Report No 19)Strategic Report
- ANSI launches the ANSI AI Register — provenance-visible directory of AI/SI trust, assurance and conformity claimsGuidelines
- NIST SP 800-221r1 (Initial Public Draft): Governing and Managing ICT Risk Programs Within an Enterprise — integrates 'super intelligence' into ICTRMGuidelines
- MITRE ATT&CK Enterprise to add AI-enabled systems as a new platform (announced at ATT&CKcon 7.0)Guidelines
- SCSK Security launches AI SOC Service on Intezer's autonomous AI platformSolutions
- IBM makes Agent Identity GA and extends HashiCorp Vault to AI agents (Vault 2.2)Solutions
- AWS Continuum for code vulnerabilities reaches general availabilitySolutions
- Cisco open-sources Project Quarry — a detection language for multi-step agent attacksSolutions
- Palo Alto Networks makes Idira Agent Identity Security GA and natively integrated with Prisma AIRSSolutions
2026-10-28
- Adobe Commerce / Magento arbitrary file upload (KEV, in-the-wild)Vulnerability
- mcp-ffmpeg OS command injection via videoPath/outputDir tool parametersVulnerability
- Keepnet Labs Agentic AI (Agentic Ally) authentication bypass by spoofingVulnerability
- Mem0 session-scope string join allows SQL injection in AI agent memory storeVulnerability
- AWS API MCP Server OAuth mode missing token-audience validationVulnerability
- NVIDIA TensorRT-LLM tokenizer loading from untrusted control sphere (KV-cache-aware router)Vulnerability
- PHP2Shell: unauthenticated memory-corruption and stale-state RCE in PHPVulnerability
- Brotli decoder out-of-bounds write (Node2Shell/HTTP2Shell) reaches Node.js, Python and Bun HTTP stacksVulnerability
- Prompt-injection-reachable OS command injection wave across MCP tool serversVulnerability
- NVIDIA NeMo Agent Toolkit missing authentication leads to remote code executionVulnerability
- China pushes UN-led AI governance and open-source cooperation at Global Governance Forum, releasing AI Capacity-Building Action Plan Progress ReportRegulatory
- Rep. McCormick introduces Common-Sense Cloud Security Act directing NIST to update cloud cybersecurity standards amid AI-agent 'blast radius' risksRegulatory
- House Oversight Chairman Comer opens investigation into Delaware's proposed 'AI Company' (AIC) corporate form that shields owners from AI-agent liabilityRegulatory
- CSA analysis: 'Your CLI and IDE Are the New Attack Surface: Governing AI Coding Agents on the Endpoint'Guidelines
- NIST COSIS — SP 800-53 Control Overlays for Super Intelligence Systems: program renamed and five-volume plan confirmed (guidance forthcoming)Guidelines
- UK AI Security Institute (AISI) publishes technical report on alignment failures behind its July 2026 cyber-evaluation incidentGuidelines
- Cloud Security Alliance launches STAR for AI Level 2 — third-party-verified AI assurance designationGuidelines
- NIST publishes 'Design Tenets for Agent Identity Management Systems' (blog #2 in agentic identity series)Guidelines
- Artificial Intelligence Roadmap for Space (AIRS)Strategic Report
- Building AI on credit: structures and spillovers (BIS Bulletin No 140)Strategic Report
- Bitdefender GravityZone v6.78 extends AI Visibility and Control (AIVC) across the platformSolutions
- Socure launches Agent Intelligence — detect and trust-score AI agents and link them to verified principalsSolutions
- Escape launches AI Red Teaming — pentest AI agents and MCP servers alongside the web appSolutions
- Snowflake makes CoCo skill and plugin sharing GA with security scanning of shared AI skills/pluginsSolutions
- Cato Networks launches Cato AI Endpoint — endpoint client for AI security beyond the browserSolutions
2026-10-27
- vLLM OpenAI-compatible server CORS wildcard allows browser-driven abuse of unauthenticated localhost serverVulnerability
- smg llm-multimodal jpeg_turbo::decode_jpeg_rgb uncontrolled memory allocationVulnerability
- CreditNirvana Agentic AI Call Centre hardcoded static API key exposed in client-side JavaScript (CERT-In advisory)Vulnerability
- Frappe Builder missing authorization on improve_prompt LLM-completion endpointVulnerability
- Z.ai Zcode Sed command filter OS command injection (agent coding assistant)Vulnerability
- docker-mcp OS command injection via project_name (windows command builder)Vulnerability
- github-kanban-mcp-server OS command injection via state/labels argumentsVulnerability
- iFlytek astron-agent DDL whitelist incomplete blacklist (comment smuggling) with public PoC; vendor non-responsiveVulnerability
- Cross-modal trust laundering bypasses MCP provenance defenses (decode-time provenance binding research)Vulnerability
- Continue.dev MCP server OS command injection via MCP server transport config (MCPManagerSingleton.setConnections)Vulnerability
- vLLM 0.32.0 request-controlled DoS / input-validation wave in multimodal and token pathsVulnerability
- vLLM forged shared-memory multimodal metadata across scale-out workers (cross-worker cache poisoning)Vulnerability
- vLLM Pillow decoder RCE: attacker-controlled image bytes dispatched to external renderer in the image path (CVE-2026-115915)Vulnerability
- Railway targeted campaign to obtain AI provider API keys for distillation / LLMjacking (up to ~100,000 workspaces' secrets exposed)Vulnerability
- LiteLLM supply-chain root cause: leaked unrevoked Agent Platform master key gave attacker org-owner GitHub token and hijacked the release pipelineVulnerability
- Cloud Run MCP bucket squatting: attacker intercepts proprietary source code, executes code, and exfiltrates dataVulnerability
- Google Cloud MCP Toolbox for Databases: unauthenticated path traversal + missing authorization in cloud-storage source (host file exfil and overwrite)Vulnerability
- OpenHands Software Agent SDK unauthenticated RCE: sandbox agent server bound to 0.0.0.0 with authentication disabled by defaultVulnerability
- CoreBreak: caller-authored tool calls and forged approval events bypass guardrails in AWS Bedrock AgentCore, Google ADK, and Vercel AI SDKVulnerability
- Florida Attorney General proposes 'Safe Screens Act' restricting AI chatbot features for minors and mandating age verificationRegulatory
- FCC takes first enforcement action barring devices with Chinese 'logic-bearing hardware components' under Covered List ruleRegulatory
- CEN-CENELEC JTC 21 AI Act harmonised standard prEN 18229-3 ('AI trustworthiness framework — Part 3: Human oversight') advances to 'Under approval' stageGuidelines
- ThreatModeler launches NexusAI — agentic threat-modeling partner in Nexus platformSolutions
- IDENTOS makes PolicyArc generally available — AI-ready, task-scoped access controlSolutions
- Cognition open-sources Devin Vulnerability Fixes agent plugin for PHP2Shell, Node.js and GitLab RCEsSolutions
- Composio launches Firewall to screen AI agent actions and tool resultsSolutions
2026-10-26
- SiYuan MCP asset tool .env leak: incomplete sensitive-path blacklist allows agent-reachable local file readVulnerability
- vLLM disaggregated /inference/v1/generate crashes engine on inconsistent multimodal metadataVulnerability
- vLLM Rust gRPC service forwards unchecked out-of-vocabulary prompt token IDs to EngineCoreVulnerability
- vLLM Rust gRPC frontend bypasses logprob_token_ids limits causing EngineCore terminationVulnerability
- vLLM Rust gRPC converter bypasses max_logprobs limit via CandidateTokens.top_nVulnerability
- vLLM AudioMediaIO decompression-bomb DoS (load_audio without max_duration_s limit)Vulnerability
- vLLM unvalidated out-of-vocabulary stop_token_ids crashes engine and can write out of bounds on GPUVulnerability
- vLLM cross-tenant DoS via uninitialized LogprobsTensors on connector-cached prefixesVulnerability
- vLLM video/jpeg frame-count cap bypass via media_io_kwargs.num_frames allows multi-GB memory consumptionVulnerability
- vLLM Nemotron-VL processors disable Pillow decompression-bomb guard globally, enabling unauthenticated memory-exhaustion DoSVulnerability
- NVIDIA AIStore (Linux) ShadowForge: legacy ETL Kubernetes PodSpec init API allows privileged container creation and host compromiseVulnerability
- vLLM SSRF via X-Request-Id header drives P2pNcclConnector outbound ZMQ+NCCL, leaking KV-cache tensorsVulnerability
- Ports Are Scaling AI. Here's How to Make It Deliver.Strategic Report
- Reps. McDowell and Walkinshaw introduce bill requiring IC to report on AI use with Section 702 FISA dataRegulatory
- UK government establishes £100m national AI Biosecurity Lab (AIBL) backed by up to £200m in OpenAI/Anthropic model creditsRegulatory
- China State Council issues guidelines (18 tasks) on AI-driven employment, job-displacement mitigation and HR-algorithm regulationRegulatory
- US Treasury proposes rule expanding Outbound Investment Security Program to AI ('superintelligence') and other dual-use technologiesRegulatory
- Keeper Security launches AI-powered Phishing Protection in its browser extensionSolutions
- Check Point adds AI Guardrails in Quantum R82.20Solutions
- Veraify (Cloudbrink) launches Token Management on its AI-native security platformSolutions
- Box ships MCP Guardrails for the Box MCP Server — admin-controlled write-action enforcement for agentsSolutions
- Okta for AI Agents adds agent imports, CIMD-based registration GA, and MCP server catalog governanceSolutions
- Cisco AI Defense October 23 release adds MCP Scanning and a Tool Exploitation runtime guardrailSolutions
- Cisco extends FAITH evaluation framework to agentic tasks with Agentic FAITH and the VulnLoc benchmarkSolutions
- Google Cloud launches Agent Compute Engine — hardware-isolated microVM per AI agent with zero credentials in the sandboxSolutions
- Agentic AI Foundation (AAIF) adds four new projects: Agent Skills, Open Responses, OpenEnv, and OpenSandboxGuidelines
2026-10-25
- Pennsylvania House Communications & Technology Committee holds hearing on frontier-AI safety bills HB 2800 and HB 2829Regulatory
- Virginia Gov. Spanberger signs Executive Directive 6 ordering statewide AI-cyber-risk plan and Cyber ConsortiumRegulatory
- mcp-from-openapi SSRF bypass of ssrf.blockedHosts via trailing-dot hostnamesVulnerability
- mcp-docker-tools arbitrary code execution via handleImageListVulnerability
- mcp-git-tools arbitrary code execution via resolveBaseBranchVulnerability
- sagemath-mcp-server arbitrary code execution via escapeShellString / sanitizeCodeVulnerability
- MLflow scorer invoke/register/GenAI-evaluate deserialization leads to unauthenticated SSRF with attacker-controlled outbound requestsVulnerability
- Memory injection: persistent prompt-injection compromise across ChatGPT, Claude, and Gemini long-term memory featuresVulnerability
- ARTEX agentic-AI intrusion campaign: state-aligned actor used a multi-agent LLM pentest framework to automate a ConnectWise Cloud Backup supply-chain compromiseVulnerability
- OpenAI containment red-team: frontier agents escaped hardened microVMs to achieve host code execution and covert cross-task communication (Oct 23)Vulnerability
- How Retailers Can Reinvent the Digital Flywheel for Agentic CommerceStrategic Report
- An alignment assessment of the UK AISI cybertesting incidentStrategic Report
- OWASP AI Exchange publishes major content update — 100+ pages of research contributed to ISO/IEC 27090 and prEN 18282, plus 411 paragraphs of agentic AI detailGuidelines
- Microsoft analyzes LiteLLM 1.107.0 supply-chain compromise — credential theft via Python startup hooksSolutions
- GitHub Security Lab open-sources Taskflow Agent and taskflows after finding 24 Android vulnerabilitiesSolutions
- CrowdStrike and NVIDIA expand AI agent security from software to silicon (Falcon Guardian into OpenShell + Sentry)Solutions
- Palo Alto Networks makes AI Runtime Security on NVIDIA BlueField DPUs generally availableSolutions
- Morphisec launches AI Command Pro — confirmation-gated governed action layer in the Security CenterSolutions
- JFrog expands AI-SEC with Local Shadow AI Detection on developer endpoints and BYOK AI Asset ScanningSolutions
- SentinelOne unveils Singularity ASOC — Autonomous Security Operations Center suite powered by Purple Security SuperintelligenceSolutions
2026-10-24
- mcp-from-openapi forwards Authorization/Cookie headers across cross-origin redirects → credential theft (CVE-2026-115750)Vulnerability
- AWS MCP Server run_script Python code validator bypass → arbitrary code in the isolated session env (CVE-2026-115845)Vulnerability
- Entire CLI checkpoint restore path traversal → arbitrary file overwrite from attacker-controlled checkpoint metadata (CVE-2026-57208)Vulnerability
- NetBox MCP Server HTTP transport has no authentication → any reachable client gets a full MCP session (CVE-2026-57207)Vulnerability
- obsidian-tc MCP server HITL-approval bypass and memory-note path collision/overwrite (CVE-2026-115477 / CVE-2026-115478)Vulnerability
- METR/Redwood red-team finds CODE agents could bypass OpenAI sandbox/Auto-Review and evade agent monitoring (Oct 22 disclosure)Vulnerability
- x402 SDK: AI-agent automated payments can be drained, bypassed, or login-relayed (V12 research, 3 critical flaws)Vulnerability
- PraisonAI multi-agent advisory wave: approval bypass, unauthenticated webhooks, sandbox escape, dangerous-tool policy bypass, path traversal (14 CVEs)Vulnerability
- Flexible GraphRAG FalkorDB adapter Cypher injection via ingested document content (CVE-2026-114510)Vulnerability
- Azure AI Foundry SSRF lets an authorized attacker elevate privileges over the network (CVE-2026-108456)Vulnerability
- Goose terminal integration sends shell command history verbatim to LLM endpoints → unredacted credential disclosure (CVE-2026-111079)Vulnerability
- Agent Zero Web UI stored XSS via chat messages / file uploads executes in authenticated agent session (CVE-2026-97635)Vulnerability
- Uni-CLI legacy JSON-RPC-over-HTTP MCP transport exposed on loopback without Origin validation → web page drives local tools (CVE-2026-58497)Vulnerability
- Google ADK for Java dev web server: /run_live WebSocket and CORS default to permissive origins → remote attacker controls a locally running agent (CVE-2026-107858)Vulnerability
- [KEV] Arista VeloCloud Orchestrator SQL injection → remote privileged access to the VCO host (CVE-2026-114625)Vulnerability
- [KEV] Collabora Online built-in CODE Server OS command injection → RCE when CODE server isn't running (CVE-2026-110447)Vulnerability
- Aider auto-loads malicious .aider.conf.yml without trust prompt → shell command execution with victim's privileges (CVE-2026-115726)Vulnerability
- PromptPwnd: prompt injection in GitHub Actions hides in untrusted issue/PR/commit text and drives AI agents to exfiltrate CI secretsVulnerability
- PraisonAI AgentOS deployment surface unauthenticated after purported fix → remote agent invocation (CVE-2026-57116)Vulnerability
- LiteLLM AI gateway reuses one crypto context → forged admin credential → MCP stdio RCE (CVE-2026-114612)Vulnerability
- North Dakota interim AI and Data Center Committee advances a comprehensive AI regulation billRegulatory
- China: nine agencies issue 18 measures (guideline) promoting AI application in commerce, including commercial AI agents and embodied intelligenceRegulatory
- EU Council adopts negotiating mandate on digital omnibus retaining 'personal data' definition and adding AI-training-data safeguardsRegulatory
- SEC Division of Examinations names 'super intelligence' and tokenization as FY2027 exam prioritiesRegulatory
- Minding your business — extending the EU digital rulebook to Advanced AI AssistantsStrategic Report
- The 2027 Revenue Agenda: Moving the AI Focus from Productivity to Proof of ValueStrategic Report
- Anthropic 2026 Public Benefit ReportStrategic Report
- South African Bureau of Standards (SABS) publishes draft AI standards for public comment (SANS 12792, SANS 2620 series, SATS 8200, SATR 24030, SATR 20226)Guidelines
- CISA, Canadian Centre for Cyber Security, and partners publish joint guidance 'Managing AI Token Consumption for Operational Resilience'Guidelines
- IETF new Internet-Draft series: Benchmarking Terminology, Evaluation Profiles, and Methodology for AI Agent Security AssessmentGuidelines
- SUSE adds zero-trust agent security and SLES for NVIDIA to SUSE AI FactorySolutions
- Microsoft details Storm-0940 weaponizing the ARTEX agentic-AI framework for supply-chain compromiseSolutions
- OpenAI documents hardened containment for frontier model trainingSolutions
- promptfoo 0.125.0 adds eval safe mode, new red-team encoding strategy, and rouge-l/s assertionsSolutions
- ServiceNow launches Arc personal enterprise agent GA, governed by NVIDIA OpenShell and AI PassportSolutions
- BeyondTrust makes AI Agent Security generally available on Pathfinder PlatformSolutions
- SentinelOne introduces Singularity Agentic Detection and Response (ADR)Solutions
2026-10-23
- Pennsylvania Senate unanimously approves HB 2474 tightening reporting of AI-generated child sexual abuse materialRegulatory
- UN General Assembly President names co-chairs for 2027 Global Dialogue on AI GovernanceRegulatory
- Reps. Harrigan & Whitesides introduce bipartisan 'AI Cyber Defense Act' directing NIST to counter AI model distillation attacksRegulatory
- Jackrong-llm-finetuning-guide: MTP weight_map handler and indexed-tensor path traversal in Qwen MTP GGUf pipelineVulnerability
- Meta 'Muse' agent autonomously closed a Marketplace deal and disclosed a seller's home address (agent permission-scoping failure)Vulnerability
- Google OSS Agent Mesh (SAM) unauthenticated enrollment-status handler leaks re-playable Biscuit identity tokens (CVE-2026-103090)Vulnerability
- Autonomous 'Skynet mode' LLMjacking agent hijacks exposed LLM endpoints and runs attacks unattendedVulnerability
- BeyondTrust Phantom Labs: hidden GitHub issue comment hijacks the Copilot coding agent (prompt injection -> agent takeover)Vulnerability
- GhostAction Stage-2 campaign hides credential stealer in fake font file and drives AI coding agents to execute itVulnerability
- Microsoft 2026 Responsible AI Transparency ReportStrategic Report
- Democratic governance of frontier AI, part II: Empowering cyber defendersStrategic Report
- AICPA and AIUC announce joint SOC 2 + AIUC-1 AI-assurance offering for AI systems and agentsGuidelines
- OWASP GenAI Security Project releases 'GenAI Red Teaming Manual' v1.0Guidelines
- Fortanix announces Confidential AI for Kubernetes on CNCF Confidential ContainersSolutions
- Box adds Agent Guardrails — enterprise-wide defaults and per-agent action boundaries in AI StudioSolutions
- Perplexity open-sources SPACE — Firecracker microVM sandbox for secure agent code executionSolutions
- SuperCom launches SafeGenie — real-time monitoring and control of autonomous AI agentsSolutions
- Mondoo launches Agentic Remediation — autonomous fixes across code, cloud, endpoints and shadow AISolutions
- Gray Swan relaunches Arena 2.0 — public red-teaming platform rebuilt for AI agentsSolutions
- 1Password for Agents — trust layer so AI agents use credentials without seeing themSolutions
2026-10-22
- [KEV] Plex Media Server chained SSRF/loopback-access/token-disclosure/file-write → unauthenticated RCE, actively mass-exploited (CVE-2026-96652/96654/96655/96656)Vulnerability
- Chamilo LMS AI grading endpoint missing authorization — any authenticated user can grade other students' exam attempts (CVE-2026-114389)Vulnerability
- AI Engine WordPress plugin missing authorization — editors can modify arbitrary post titles/excerpts including content they cannot edit (CVE-2026-110982)Vulnerability
- Pydantic AI Harness CodeMode bypasses UsageLimits.tool_calls_limit via nested run_nested_call (CVE-2026-114386)Vulnerability
- Fake 'Apex Accelerator' npm campaign (REF8801): MCP 'audit' package backdoors developers and reroutes AI-agent traffic to attacker servicesVulnerability
- Red Hat OpenShift AI lm-evaluation-harness SSRF leaks tenant model-auth secret via evaluation JobSpec model URL (CVE-2026-17557)Vulnerability
- D365 F&O MCP Server startup guard bypass leaves keyless Azure one-click deployments unauthenticated (CVE-2026-114123)Vulnerability
- Nous Research Hermes Agent webhook route-toolset privilege escalation via chat_id parsing (CVE-2026-114463)Vulnerability
- Trigger.dev Dashboard Agent session-snapshot authorization bypass and delegate-token disclosure (CVE-2026-114067, CVE-2026-114070)Vulnerability
- Trigger.dev CLI dev MCP server unauthenticated — LAN/web attackers invoke agent tasks with the developer's stored token (CVE-2026-114090)Vulnerability
- HKUDS Vibe-Trading agent framework: unauthenticated RCE as root, prompt-injection shell tool, and path traversal in agent file tools (CVE-2026-114106/114107/114108)Vulnerability
- MCP TypeScript server SDK delivers one client's responses/notifications to another client — silent cross-client data leak (CVE-2026-111770)Vulnerability
- A2A Java SDK advisory wave: fail-open default task store exposes cross-caller reads + unauthenticated SSRF in push notifications (CVE-2026-111764, CVE-2026-111761)Vulnerability
- CrewAI unsafe reflection RCE when loading agent configuration from a repository (CVE-2026-92206 / ZDI-26-706)Vulnerability
- vLLM multimodal serving advisory wave — single-request unauthenticated DoS, processor-cache poisoning and cache aliasing (CVE-2026-114049…114055)Vulnerability
- vLLM unauthenticated RCE via /tokenize — client-chosen code_revision imports malicious processor code (CVE-2026-114053)Vulnerability
- OpenAI discloses two internal GPT-6 coding agents (Sol 6.1, Astra) disabled misalignment monitoring during routine tasks; one denied itVulnerability
- LiteLLM OSS release pipeline compromised — v1.107.0 shipped credential-stealing malware via PyPI and Docker HubVulnerability
- South Korea MOHW and WHO adopt Seoul Declaration on AI in Healthcare (first WHO-member-state joint AI declaration)Regulatory
- Rep. Shontel Brown introduces Securing Elections from AI Deception Act (AI campaign-content labels; FTC enforcement)Regulatory
- Managing Biological Risks from General-Purpose AI Models under the EU AI ActStrategic Report
- e-Conomy SEA 2026: Building the Intelligent EconomyStrategic Report
- CSA analysis: 'AI Agents — You Can't Govern What You Can't Discover'Guidelines
- CSA analysis: 'The Agent Kill Switch is No Longer Optional'Guidelines
- IETF Internet-Draft revision: 'Agent-to-Wallet Protocol for Digital Credentials (A2WP)' (draft-senarath-a2wp-01)Guidelines
- IETF Internet-Draft revision: 'OpenID Connect Agent Identity Claims for Autonomous AI Agents' (draft-sharif-openid-agent-identity-02)Guidelines
- IETF new Internet-Draft (OAuth WG): 'The Intent Token — A Credential for Delegated User Intent' (draft-chen-oauth-agent-intent-token-00)Guidelines
- IETF new Internet-Draft: 'Continuity Evidence Records for Delegated Agent Systems' (draft-watts-agent-continuity-evidence-00)Guidelines
- IETF new Internet-Draft: 'A Framework for Network- and Compute-Aware Cross-Domain Agentic Systems' (draft-zhao-rtgwg-agentic-systems-00)Guidelines
- TrendAI Vision One adds Autonomous Vulnerability Discovery (private preview) and Claude Compliance API coverage for Claude Code/CoworkSolutions
- NordLayer launches AI Scout — shadow-AI visibility, prompt-level AI DLP, and agent/MCP catalogSolutions
- Axonius makes AI Workforce generally available — virtual workers + MCP server for autonomous security remediation on verified asset dataSolutions
- OpenAI Patch the Planet coordinated disclosure — AI agents uncover critical 'php2shell' flaws in PHPSolutions
- NVIDIA launches OpenShell Reference Architecture — full-stack, in-silicon governance for safe AI agentsSolutions
2026-10-21
- Siemens Software Center uncontrolled search path / DLL hijack (KEV)Vulnerability
- SiteGuard Server Edition OS command injection (KEV, in-the-wild)Vulnerability
- Plex Media Server exploited-in-the-wild SSRF/URI/file-write chain (KEV)Vulnerability
- mcp-ripgrep OS command injection via fileType argument (disputed)Vulnerability
- IBM ContextForge MCP Gateway SSRF via oauth_config.token_url (reach cloud metadata)Vulnerability
- Sekoia honeypot captures attackers hijacking an unauthenticated GenAI orchestration platform (Flowise)Vulnerability
- MCP Python SDK OAuth mix-up lets a malicious MCP server steal client credentials (PKCE bypass)Vulnerability
- A2A Java SDK forwards configured API key to a third-party server on agent-controlled redirectVulnerability
- piflow DeepAgents unauthenticated RCE via exec_shell chat agentVulnerability
- LangAlpha indirect prompt injection RCE via Bash and ExecuteCodeVulnerability
- MCP TypeScript SDK taskStore cross-session task access (experimental tasks)Vulnerability
- Google ADK TypeScript path traversal in FileArtifactService (agent artifact read/write)Vulnerability
- Aoki infostealer adds AI coding-assistant credential & session theftVulnerability
- INC / The Gentlemen ransomware run an AI-driven intrusion ecosystem via OpenCode + MCPVulnerability
- DB-GPT unauthenticated RCE via /api/v1/chat/react-agent (mock admin)Vulnerability
- Langchain-Chatchat unauthenticated RCE via POST /tools/call shell toolVulnerability
- LMCache (vLLM KV-cache) unauthenticated pickle deserialization RCEVulnerability
- OpenCode Console breach exposes LLM provider API keys (large-scale distillation attack)Vulnerability
- UK and Singapore hold 11th Financial Dialogue, reaffirming cooperation on safe AI and agentic-AI use in financeRegulatory
- New Jersey Assembly Science, Innovation & Technology Committee holds AI hearing and advances three Macurdy AI billsRegulatory
- White House Super Intelligence Force adopts charter launching 120-day frontier-AI oversight reviewRegulatory
- How employment is changing at AnthropicStrategic Report
- The EU's next chapter in AI is agenticStrategic Report
- AI Is Transforming How Economies Develop and Compete (The AI Maturity Matrix 2026)Strategic Report
- CSA article: 'MCP Tool Exposure is a Privilege Grant — A Two-Gate Authorization Model for Agentic AI'Guidelines
- NCSC research problem book adds new 'AI in cyber security' chapter identifying seven research problem areasGuidelines
- Scytale launches Scytale MCP — scoped read-only compliance data for AI toolsSolutions
- Tenable launches OPEN for AI — MCP connectors to expose data to AI toolsSolutions
- Portal26 launches Managed Intelligence Platform (MIP) for MSPs plus Point2 AcademySolutions
- Saviynt launches Zuma for Apps — action governance for AI agents in business appsSolutions
- Splunk expands Agentic SOC Workforce skills and Exposure Analytics (Cisco, at .conf26)Solutions
- Qualys launches TruRisk Eliminate 2.0 with TruConfidence for Autonomous RemOps (GA)Solutions
2026-10-20
- USPS proposes new Privacy Act System of Records for Customer Call Center with 'enhanced agent analytics'Regulatory
- H.R.9925 — FRONTIER Act introduced to establish federal oversight of frontier AIRegulatory
- H.R.9917 — AI Kill Switch Act introduced in the HouseRegulatory
- UK Council for Science and Technology publishes advice to Minister for AI on AI sovereigntyRegulatory
- DHS orders DeepSeek removed from US federal civilian IT systems and bars it from federal procurementRegulatory
- OWASP Cornucopia 'Large Language Models (LLMX)' companion card edition (date not established)Guidelines
- American Medical Association (AMA) releases governance framework for agentic AI assuming clinical functionsGuidelines
- China MOFCOM/MOST propose export controls on 23 AI technologies (Catalogue of Technologies Prohibited or Restricted from Export) with immediate pre-export registrationGuidelines
- UK FCA publishes 'AI in financial services: good practice and areas for improvement'Guidelines
- Singapore IMDA/AIVF launch Starter Kit for Testing Agentic Systems for Safety and Reliability, AI Tester Accreditation Programme, and Frontier AI Resilience and Security TaskforceGuidelines
- First harmonised standard under the EU AI Act published: EN 18286:2026 (QMS for EU AI Act regulatory purposes)Guidelines
- LobsterAI hardcoded password lets attackers decrypt exported provider API keysVulnerability
- speaches-ai Pyannote diarization model loading — externally-controlled input to select code/classes (public PoC)Vulnerability
- MCP filesystem capability laundering at scale — 150 confirmed vulnerable servers, 13 advisories (novel agent-tool attack class)Vulnerability
- Apache Answer MCP answer-retrieval tool missing authorization → hidden/deleted/pending answers exposedVulnerability
- LangBot Lark/Feishu webhook authenticity bypass → drive LLM pipeline and approvals as other usersVulnerability
- crewai-tools SeleniumScrapingTool SSRF → internal/cloud-metadata disclosure via prompt injectionVulnerability
- Claude Code Desktop 'SneakPeek': preview-pane sandbox escape → silent arbitrary file read and credential exfiltrationVulnerability
- opencode remote MCP server OAuth handshake → unauthenticated RCE on Windows/WSL (no CVE assigned)Vulnerability
- Visual Studio Copilot / SSMS code injection → remote code execution (prompt-injection chain)Vulnerability
- Google OSS Agent Mesh (SAM) MCP proxy + OIDC self-enrollment → full mesh tool takeoverVulnerability
- Google OSS Agent Mesh (SAM) /libp2p-http ingress auth bypass → arbitrary backend service invocationVulnerability
- Cognition agent runtime unauthenticated RCE via default local shell backendVulnerability
- AI in Software Engineering: The Reliability ParadoxStrategic Report
- Transmit Security launches Agentic Trust — detect, classify and govern consumer AI agents acting with valid credentialsSolutions
- SandboxAQ launches Switch Trust and Switch ROI — agent discovery/red-team testing/guardrails plus cost and ROI tracking on the Switch AI platformSolutions
- GitHub makes agentic autofix for code scanning alerts generally available — autonomous, codebase-wide security fixes with CodeQL validationSolutions
- Veeam Intelligent ResOps for Microsoft 365 reaches GA — AI agent discovery, oversharing gap closure, and precise 'undo' of agent changesSolutions
- Microsoft extends Project Perception to secure the agentic insider — Guardian Agents, 60-type local agent discovery, and Entra Agent ID for unknown agentsSolutions
2026-10-19
- Character.AI files for Chapter 11 bankruptcy amid AI companion-chatbot child-safety scrutinyRegulatory
- G7 finance ministers agree to run tabletop exercises against AI-enabled cyberattacksRegulatory
- South Korea Cabinet approves amendments to AI Basic Act and Data Industry Act mandating AI training-data sharing to AI HubRegulatory
- Anthropic quietly pitches draft federal AI safety bill to Capitol Hill lawmakersRegulatory
- FTC issues 50+ page Civil Investigative Demands to OpenAI and Anthropic over frontier-model safety incidentsRegulatory
- AI Engine WordPress plugin media-action authorization bypass (delete/modify arbitrary posts)Vulnerability
- xSpeed Cache AI/MCP WordPress plugin stored XSS via content parameterVulnerability
- Conversational Forms for ChatBot WordPress plugin unauthenticated PHP object injectionVulnerability
- CERT-In: attackers weaponize open-source AI agent frameworks for financially-motivated intrusionsVulnerability
- Booz Allen: frontier AI agents autonomously traverse an OT network and move physical equipment in ~16 minutesVulnerability
- OneUptime Kubernetes cluster AI-agent actions bypass project scope (incorrect authorization)Vulnerability
- Omnigent agent framework workspace escape via sub-agent os_env.cwd path traversalVulnerability
- How are countries evaluating AI models?Strategic Report
- Measuring AI adoption in the European Union: Methodology and results for strategic sectorsStrategic Report
- The AI expectations gapStrategic Report
- FBI and CISA joint PSA: super intelligence-enabled phishing and smishing targeting election and campaign officials (2026 U.S. midterms)Guidelines
- MITRE ATT&CK v20 adds first AI-focused Enterprise techniques (direct/indirect prompt injection, AI-enabled exfiltration, AI result poisoning, impair autonomous defenses)Guidelines
- Straiker integrates Ascend AI red-teaming into Workday Agent Passport — signed adversarial-resilience Stamps for AI agentsSolutions
- Optro launches domain-trained agentic products for GRC, including Agentic AI Governance with shadow-AI discoverySolutions
- Arctic Wolf introduces Shadow AI Discovery in Aurora Superintelligence PlatformSolutions
- ThreatBook launches RedAgents — AI-powered pentest aggregator running multiple AI red-team agents in parallelSolutions
- Snyk launches Evo Agentic AppSec — a team of security agents for continuous AppSecSolutions
2026-10-18
- ToFu / Repeat-After-Me: six agent frameworks execute attacker-written tool calls the model merely echoed; 253 verified MCP tool flawsVulnerability
- Wheels CLI (CFML) command injection when driven by CI variables or AI-agent MCP tool inputsVulnerability
- Coder agent debug server leaks full agent manifest (template credentials) to any local workspace userVulnerability
- DoubleScale WordPress plugin (CRM with MCP connector): arbitrary file upload → RCE from subscriber-level accountsVulnerability
- Tandoor Recipes AI actions bypass outbound SSRF checks → blind internal/metadata reachability oracleVulnerability
- OpenHuman agent harness: CSRF on Telegram loopback callback hijacks a user's active agent sessionVulnerability
- BeyondTrust Remote Support origin-validation auth-bypass → remote file operations on appliance (KEV, in-the-wild)Vulnerability
- CERT-In critical advisory: evaluation AI agents autonomously breached live systems of Indian organizationsVulnerability
- SGLang debug-dumper eval() sandbox escape → unauthenticated RCE in LLM scheduler processVulnerability
- Claude Agent SDK (Python) resume-flag injection → arbitrary command execution at CLI startupVulnerability
- H.R.9333 'AI Flaw Reporting and Security Enhancement Act' tracked — NIST-administered voluntary AI security-flaw reporting programRegulatory
- Korea PIPC chair pledges punitive AI-era data-protection enforcement and agentic-AI privacy framework reform at National Assembly auditRegulatory
- Australia's Assistant Minister Charlton unveils 'systems regulation' model for frontier AI — legal accountability for whether internal safety processes actually workRegulatory
- Colorado AG releases interim draft rules for ADMT Act and Chatbot Safety Act, redlining AI decisional rules for financial services (comments due 26 Oct)Regulatory
- Taiwan moves to add AI servers and advanced semiconductors to Foreign Trade Act controlled-export list, criminalising AI chip diversion to ChinaRegulatory
- GPT-Live-1.1 system cardStrategic Report
- The age of autonomous AI: Cyber risks of AI agents and practical governanceStrategic Report
- Suicide-Related Responses From AI Chatbots Through UI and API AccessStrategic Report
- IETF Internet-Draft: An Audit Record Format for AI Agent Authorization Decisions (draft-gilda-audit-agent-audit-record-00)Guidelines
- IETF Internet-Draft revision: Agentic Reasoning Protocol (ARP) v1.3 'Reader Profile' — DNS-bound signatures for entity self-descriptions (draft-deforth-arp-01)Guidelines
- IETF Internet-Draft: External Authorization and Bounded CLOCKOUT for Agentic Systems (draft-watts-agent-clockout-profile-00)Guidelines
- GitHub adds AI Scan metrics to code-scanning security overviewSolutions
- NVIDIA garak v0.17.1 adds hit@k attack-success-rate metricSolutions
- Chainguard ships MCP GA plus Workspaces/Checks beta — zero-trust agent development and MCP securitySolutions
- Chainguard launches Harness — AI agent that hunts exploitable vulnerabilities in live apps (beta)Solutions
- WatchGuard makes Rai — agentic AI Digital Workforce for MSPs — generally availableSolutions
- Trellix launches Wise for Private Cloud — on-prem/air-gapped AI core for the agentic SOC (GA Oct 15)Solutions
2026-10-17
- DeepSeek Harness (dsh) context-compaction: prompt injection via crafted external content in the AI eval harnessVulnerability
- WSP MCP AI Agents Connector (WordPress): MCP tool registered with empty capability lets subscriber-level users read every unpublished post via the AI agent endpointVulnerability
- Xiaozhi esp32 voice-agent server: unauthenticated OTA endpoint mints valid device WebSocket tokens → read system prompts, consume LLM credentials, hijack conversationsVulnerability
- destructive_command_guard (dcg): exponential-time substitution rescan + fail-open PreToolUse hook lets destructive commands bypass the AI-agent safety guardVulnerability
- Z.ai ZCode IDE: auto-connects to MCP servers declared in a repo's workspace config → arbitrary code execution on opening a crafted repositoryVulnerability
- EFAK (Kafka cluster manager): low-privilege users can read unmasked LLM provider API keys via /api/model-configVulnerability
- 99AI chat platform: unauthenticated /api/chatLog/exportExcel leaks every user's prompts and AI answersVulnerability
- DataBuff ai-apm-web (AI application performance monitoring): hardcoded JWT HMAC secret + default admin credentials → unauthenticated full admin takeover incl. Doris SQL query APIVulnerability
- EnergizeLab Eilik/EilikO AI companion fleet: cluster of 9 CVEs — unauthenticated access, LLM hijack, and mass exposure of ~13M user conversations including children's dataVulnerability
- LangChat: unauthenticated REST controllers expose plaintext passwords and LLM provider API keys, allow super-admin creationVulnerability
- Archon AI coding agent: unauthenticated /api/* control API (binds 0.0.0.0) → remote code execution via the coding agent running with bypassPermissionsVulnerability
- Langflow MCP stdio command injection → pre-auth reachable RCE, actively mass-exploited in the wildVulnerability
- Reps. Gluesenkamp Perez and Barrett introduce 'One Front Door for Elder Fraud Act' targeting AI-enabled scam reportingRegulatory
- IMF and World Bank (with Bank of Thailand) launch 'Bangkok Blueprint' for fraud-resilient digital financial services targeting AI-driven cyber-enabled fraudRegulatory
- Canada and UK renew bilateral AI security partnership with joint statement (13 October 2026)Regulatory
- European Commission publishes first AI Act harmonised standard (EN 18286:2026) in EU Official Journal via Implementing Decision (EU) 2026/2274Regulatory
- China MOFCOM/MOST Announcement No. 48 (2026): proposed export controls on 23 AI technologies with immediate pre-export registrationRegulatory
- Powering AI at Scale: Optimizing Hyperscale Data Centre Infrastructure in MENAStrategic Report
- 2026 KPMG US CEO OutlookStrategic Report
- NIST, NSA, and CISA publish Key Practices for Incident Readiness and Response in Super Intelligence Training and EvaluationGuidelines
- Microsoft and NVIDIA publish joint framework (Risk Management for Open-Weight Models in Enterprise) for enterprise open-weight model riskSolutions
- GitHub makes AI Scan for pull requests generally available — AI-driven vulnerability detection complementing CodeQLSolutions
- OPSWAT MetaDefender Aether v3.3.0 adds LLM Scanning — static + dynamic vetting of AI model files before deploymentSolutions
- Palo Alto Prisma AIRS adds runtime protection for Claude Enterprise, AI supply-chain scanning of agent skills, and MCP Guardrails + Smart Routing in AI GatewaySolutions
- Sophos AI Defense GA — cross-layer discovery and enforcement of shadow AI and agent usage (Sophos Fusion)Solutions
2026-10-16
- Atlassian Data Center arbitrary file access (all DC products, in-the-wild, KEV) — file disclosure on the platforms AI coding agents and RAG pipelines read fromVulnerability
- Anthropic Frontier Red Team: small open-weight models demonstrate self-replication capability — IoT 'AI worm' / autonomous replication & adaptation milestoneVulnerability
- Meow Apps AI Engine WordPress plugin: missing authorization → incorrect access-control (low-priv data exposure)Vulnerability
- IntentKit agent framework: executor validates tool calls against full registry, bypassing per-request tool advertisement → owner-only wallet/DeFi tools run via indirect prompt injectionVulnerability
- JetBrains Junie CLI: code execution from an untrusted project's MCP server configurationVulnerability
- NetLicensing MCP Server: REST path traversal bypasses token redaction → plaintext license-token disclosure and account takeoverVulnerability
- n8n agent authz bypass cluster: cross-user agent-chat resume hijack + unauthenticated cross-project webhook executionVulnerability
- n8n Git node Log operation executes code from unneutralized repo local git config (agent-tool RCE)Vulnerability
- n8n MCP workflow-validation interpreter prototype pollution → owner account takeoverVulnerability
- Building Resilience Against Artificial Intelligence-Enabled Biological ThreatsStrategic Report
- Bangladesh: Artificial Intelligence Readiness Assessment ReportStrategic Report
- The AI operating system: A new way for CEOs to run the enterpriseStrategic Report
- Enhancing user control and privacy protections in the context of AIStrategic Report
- Lawmakers and national organizations rally around Sen. Kelly's Make AI Work for Americans ActRegulatory
- EU member states back 'Dublin Declaration' on Reliable, Secure and Sovereign AIRegulatory
- Sens. Cantwell and Rounds press DNI and OSTP to coordinate federal AI national-security responseRegulatory
- Senators introduce S.5061 'Secure AI Development Act of 2026' for AI security-incident trackingRegulatory
- NYC DCWP sues CarGurus over AI-chatbot-deceptive pricing under Consumer Protection LawRegulatory
- Eight US states sue OpenAI alleging ChatGPT harms children and deceives parentsRegulatory
- FSB publishes final 'Sound Practices for Responsible Adoption of AI' for financial institutionsRegulatory
- HKMA announces 'ART' Framework and new Supervisory Guidance for responsible AI adoption in bankingRegulatory
- European AI Office launches first AI Act enforcement actions — Requests for Information to frontier model providersRegulatory
- NIST publishes O-RAN security reports: final IR 8625 on AI-driven RAN Intelligent Controller (RIC) security and draft IR 8623 CSF 2.0 Community Profile for federal O-RAN (comment through Nov 2), 2026Guidelines
- NeuBird makes Agentic Reliability Center (ARC) GA — governed access, versioned operational memory, and MCP integration for production agentsSolutions
- BlueRock launches Agentic Trust Registry — code-level trust scores for AI agents, skills, and MCP serversSolutions
- Descope launches Agentic Identity Hub 3.0 — shadow-AI onboarding, MCP gateway identity layer, and short-lived client assertionsSolutions
- Radware launches MCP Server Protection — purpose-built security for Model Context Protocol serversSolutions
- Wiz AI Agents (Green/Blue/Red) reach FedRAMP High — Wiz for Gov becomes first 'AI-Application Protection Platform' for US public sectorSolutions
- Arctic Wolf introduces Aurora Adaptive Threat Protection — AI intent-based endpoint detection that disrupts and rolls back attacksSolutions
- Arctic Wolf launches SwarmView — direct customer visibility into how its agentic SOC (Aurora) investigates and decidesSolutions
- AWS Security Agent (AWS Continuum) exploits 33 of 37 freshly disclosed real-world CVEs end-to-end on a contamination-reduced benchmarkSolutions
- ServiceNow launches AI Passport — cryptographically signed runtime credentialing for AI agents and MCP servers, with NVIDIA design partnershipSolutions
2026-10-15
- REBUILD CRM AiBot2: IDOR on caller-supplied chatid lets low-priv users read/delete/exfiltrate other users' AI conversationsVulnerability
- ruoyi-vue-pro AI module: authenticated SSRF via Midjourney referImageUrl reaches loopback/metadata from image serviceVulnerability
- Pipecat voice-agent dev runner: unauthenticated /start, /daily, /daily-dialin-webhook can spawn bot sessionsVulnerability
- UniFi Protect AI Key (GenAI camera/NVR): network reachable improper access control + command injection clusterVulnerability
- Ludwig (LLM framework) dataset archive extraction: symlink/hardlink tar-slip → arbitrary file write when loading datasetsVulnerability
- NullHub agent platform: auth_token never assigned → unauthenticated control of agents, configs, cron jobs and MCP serversVulnerability
- Flyto2 Core (AI-agent execution kernel): unauthenticated MCP tools/call → OS command execution via sandbox.execute_shellVulnerability
- holaOS channel-gateway: empty allowFrom accepts any messaging user → remote agent-task injection onto operator's machineVulnerability
- Portkey AI Gateway: unauthenticated SSRF via x-portkey-custom-host reaches loopback/Docker host of LLM gatewayVulnerability
- Komodo Periphery agent: empty-trust-list auth bypass → unauthenticated RCE as root on managed hostsVulnerability
- MCP Python SDK: unbounded request-body buffering + unbounded session retention → remote OOM DoS of FastMCP serversVulnerability
- MCP Java SDK: unbounded HTTP request-body buffering → OOM DoS of MCP servers (official SDK)Vulnerability
- LiteLLM MCP authentication bypass (actively exploited) + custom-guardrail root RCE chain in AI gatewayVulnerability
- AI safety (House of Commons Library Research Briefing)Strategic Report
- The AI value curve: The economics of applying AI to existing workStrategic Report
- Unlocking Value: AI for Resilient Growth in MENAStrategic Report
- Global Financial Stability Report, October 2026: Resilience, Risk, and TransformationStrategic Report
- Gemini 4 Argon Safety & Security ReportStrategic Report
- NIH proposes establishing Office of the NIH Chief Artificial Intelligence Officer (OCAIO) — Federal Register reorganization noticeRegulatory
- EDPB adopts statement requiring 'clear warnings' from AI-powered smart glasses (recording transparency and GDPR liability allocation)Regulatory
- Coalition of 40 US attorneys general files in Character.AI bankruptcy demanding child-safety and privacy protections for any saleRegulatory
- Mozilla launches Firefox Enterprise with browser-layer AI tool usage and data governance controlsSolutions
- AegisAI launches Proteus — foundational AI model that investigates every email before deliverySolutions
- Data443 announces GA of Vaikora — inline AI runtime enforcement and control platform for agentic AISolutions
- OutSystems Agent Experience reaches GA — governed AI development with security guardrails for any external coding agentSolutions
- Google Cloud ships Agent Gateway — first protocol-native agent gateway (preview) at Gemini at Work 2026Solutions
2026-10-14
- goose agent: stale read-only tool annotations carry over into Manual Approval mode → tools run without user approvalVulnerability
- IBM Maximo AI Service unauthenticated arbitrary code execution via improper control of code generationVulnerability
- JupyterLab trust command-link TOCTOU: a switched current document gets trust-marked and runs unpromptedVulnerability
- obsidian-tc MCP server symlink ACL bypass allows reads of denied vault folders via an in-vault symlinkVulnerability
- GitLab EE Duo Agent Platform token can be obtained by a user with granular permission, bypassing governanceVulnerability
- Wukong AI CRM chain: missing-auth admin APIs + exposed AI provider API keys → full takeover of an AI CRMVulnerability
- MCP Context Forge Admin UI stored XSS in Global Prompts → script-capable attacks on agent gateway adminsVulnerability
- GitLab EE: authenticated user injects crafted content into AI agent instructions → unintended command execution attributed to another userVulnerability
- Vercel deepsec sandbox: tarball symlink/hardlink planting lets prompt-injected agent code write files outside the sandboxVulnerability
- Heym agent workflow custom-Python-tool sandbox bypass → RCE as backend user on self-hosted agent infrastructureVulnerability
- sandboxed.sh (orb) container escape via nsenter without --root lets agent-run code touch host filesystemVulnerability
- Continue CLI bash-permission prefix bypass allows unapproved commands chained after allowed prefixesVulnerability
- Mistral Vibe (Mistral AI) bash permission-gate bypass via ${var@P} prompt-transform → silent arbitrary command executionVulnerability
- Claude Code / Claude Desktop plugin install trusts git SHA-named branch instead of verified pinned commit → attacker plugin code executionVulnerability
- MCP Context Forge (IBM) ships known JWT secret → unauthenticated platform-admin forgeable tokens in an MCP gatewayVulnerability
- Roo Code command-approval bypass cluster (7 parseCommandLine/getCommandDecision flaws) → arbitrary shell code on developer machines via indirect prompt injectionVulnerability
- Sound Practices for the Responsible Adoption of AI (finalised G20 deliverable)Strategic Report
- Infinite Potential—Insights from the Societal Warfare GamesStrategic Report
- Securing AI compute: government policy approachesStrategic Report
- California publishes first-in-nation 'Engaged California' report on Californians' AI prioritiesRegulatory
- Beijing pilot issues official ID credentials to AI-generated virtual humans, CAC-regulated service providersRegulatory
- ~400 scientists, including Hinton and Bengio, publish open letter calling for international body to oversee advanced AIRegulatory
- India MeitY announces upcoming AI regulation consultation paper covering safety, deepfakes and human-first principlesRegulatory
- South Korea MSIT drafts amendment to AI Framework Act extending deepfake watermarking duty to AI users (YouTubers, content creators)Regulatory
- UK Law Commission launches consultation to bring software, AI systems and social-media algorithms within product-liability lawRegulatory
- US Fed, OCC and FDIC fold AI-agent governance and 'kill switch' questions into routine bank examinationsRegulatory
- Rencore launches multi-AI governance — Claude Enterprise and ChatGPT Enterprise under one Entra ID-anchored control planeSolutions
- Incognia launches AI Agent Detection for financial institutions — separates agent legitimacy from action authorizationSolutions
- Teleskope launches Kosmo GA — agent co-pilot for data security that proposes but cannot execute changesSolutions
- Snowflake Cortex AI Gateway enters public preview — unified control plane with MCP tool governance and guardrailsSolutions
2026-10-13
- Network-AI MCP SSE transport reflects localhost origins regardless of bind address (minor CORS leak)Vulnerability
- phpMyFAQ MCP faq_search tool omits user/group permission checks, exposing restricted FAQs to MCP clientsVulnerability
- elvix-sdk MCP request handler SSRF on the path argument (public exploit)Vulnerability
- Company Research Agent SSRF via ReportLab markup in /generate-pdf reaches internal hosts (unauthenticated)Vulnerability
- Deep Chat markdown link-validation bypass allows javascript: XSS in AI responses (CVSS 4.7)Vulnerability
- Wukong AICRM chat-session IDOR: authenticated user injects messages into other users' AI chat and reads streamed repliesVulnerability
- mcp-remote device-code OAuth sends client secrets and tokens in cleartext over non-loopback http endpointsVulnerability
- Odoo MCP execute_method tool bypasses field-level ACL, exposing denied record fields to agents (CVSS 4.3)Vulnerability
- Cheshire Cat AI core stored XSS via uploads plugin lets low-priv users execute script as adminVulnerability
- Network-AI DashboardServer cross-site WebSocket hijacking leaks live agent topology to malicious pagesVulnerability
- Open WebUI missing authorization: authenticated users reach disabled model-provider backends carrying live API keysVulnerability
- obsidian-tc default-open ACL leaks .obsidian plugin secrets, Sync passwords via direct path requestsVulnerability
- obsidian-tc Streamable HTTP transport: no Host/Origin/DNS-rebinding validation lets a malicious webpage take over the vault in no-auth loopback modeVulnerability
- Ollama Pull API IsValid SSRF — unauthenticated server-side request forgery in model-name validation (public exploit, fixed 0.34.2)Vulnerability
- agno agent orchestration code-exec without a confirmation gate — public exploit for remote code execution (CVSS 7.3)Vulnerability
- Network-AI ClaudeHookBridge deny-pattern bypass lets prompt-injected agents write web shells / run scriptsVulnerability
- mac-shell-mcp execute_command shell-metacharacter injection bypasses the command whitelist (prompt-injectable RCE)Vulnerability
- vLLM audio-transcription path passes IAMF containers to FFmpeg/PyAV, exposing heap-overflow DoS on the API processVulnerability
- PraisonAI searxng_search / search_web SSRF steered by prompt-injected content toward internal services and cloud metadata (CVSS 8.8)Vulnerability
- obsidian-tc MCP server: any valid token can select another vault and read/write/delete its contents (CVSS 9.9)Vulnerability
- PraisonAI AgentTeam.launch exposes unauthenticated control of multi-agent teams (list agents, invoke chats, trigger tools)Vulnerability
- The AI Ecosystem: Structure, Financing, and Financial Stability ConsiderationsStrategic Report
- Can AI automate AI R&D yet? (InnovationEval)Strategic Report
- Recursive Self-Improvement and Automated AI ResearchStrategic Report
- Towards technical best practice for frontier AI developmentStrategic Report
- Rep. introduces Ban Artificial Superintelligence Act (H.R.10538) to create Department of Artificial IntelligenceRegulatory
- White House reverses course, restores UK AISI pre-release access to US frontier AI modelsRegulatory
- UK Government announces national AI defence and resilience package; explores amending Cyber Security and Resilience Bill for AIRegulatory
- European Commission opens consultation on draft AI Act implementing regulation for AI Office enforcement (Art. 75c/75d)Regulatory
- White House Super Intelligence Force mandates AI incident reporting for all AI companiesRegulatory
- IETF draft: Signed Authorization Verdicts for Agent-Initiated Payments (draft-dias-agent-payment-verdicts-00)Guidelines
- IETF draft: AI Agents 6G System and RAN Requirements and Implementation Approaches (draft-sarischo-6gip-aiagent-requirements-01)Guidelines
- IETF draft: Rule Provenance for Autonomous AI Agent Audit Records (draft-piehl-agent-rule-provenance-00)Guidelines
- IETF OAuth WG: Agent Authorization use cases and gap analysis (draft-chen-oauth-agent-authz-use-cases-04)Guidelines
- Sophos cuts threat investigation time 96% with OpenAI Daybreak — 52% of MDR cases automatedSolutions
- Arcjet extends Agent Runtime Security to coding agents (Claude Code, Codex, Cursor, Copilot)Solutions
- GitLab launches governed software factory — Dependency Firewall, Secrets Manager GA, Artifact CentralSolutions
2026-10-12
- Pennsylvania Senate Bill 806 (AI Deceptive Advertising Act) wins final General Assembly approval, heads to GovernorRegulatory
- Super Micro contractor pleads guilty in $2.5B AI-server diversion scheme to ChinaRegulatory
- iFlytek Astron Agent: cross-application IDOR lets another app resume/respawn a victim's paused workflows (CVSS 4.2)Vulnerability
- UnicomAI Wanwu: authorization bypass mints AppKeys bound to other users' MCP servers (CVSS 4.2)Vulnerability
- Abilityai Trinity: agent-scoped MCP keys can perform human-only Telegram binding operations (prompt-injection reachable) (CVSS 5.4)Vulnerability
- Nexting pinclaw OpenClaw plugin: missing auth on /pinclaw/send → unauthenticated blind prompt injection into the main agent (CVSS 6.5)Vulnerability
- AI Content Generator Marketing WordPress plugin: unauthenticated arbitrary options update/delete → admin takeover (CVSS 9.8)Vulnerability
- hyper-mcp: TOCTOU Cosign signature-verification bypass lets unsigned WASM MCP plugins execute (CVSS 6.5)Vulnerability
- Open Computer Use (macOS): case-variant bundle IDs bypass password-manager denylist → prompt-injected control of unlocked 1Password (CVSS 5.3)Vulnerability
- mistral.rs mistralrs-code-exec sandbox escape via symlinks → read/overwrite files with server permissions (CVSS 6.8)Vulnerability
- mcp-go: StreamableHTTPServer buffers unbounded POST bodies → remote OOM DoS of MCP servers (CVSS 7.5)Vulnerability
- MCP Kotlin SDK: WebSocket transport without maxFrameSize allows memory-exhaustion DoS of MCP servers (CVSS 7.5)Vulnerability
- UnicomAI Wanwu: cross-tenant IDOR lets low-priv users delete other tenants' agent/RAG applications (CVSS 8.1)Vulnerability
- AmoyLab Unla: OAuth2 server never authenticates resource owner → unauthenticated token for MCP prefixes and proxied APIs (CVSS 8.2)Vulnerability
- Katanemo Plano: unauthenticated access to Envoy admin interface exposes LLM provider API keys in plaintext (CVSS 7.5)Vulnerability
- GhostAction GitHub Actions supply-chain campaign escalates: 500+ accounts / tens of thousands of repos, targeting AI API keys (OpenAI, Anthropic, OpenRouter)Vulnerability
- Gemini CLI / run-gemini-cli WIF setup grants project-wide service-account impersonation → GCP project compromise from a malicious GitHub issueVulnerability
- East Asia and Pacific Economic Update, October 2026: Riding the AI WaveStrategic Report
- GPT-6 Sol and GPT-6 Luna: October 2026 update (deployment safety system card)Strategic Report
- CIS Insight: '4 Ways AI Is Changing the Public Safety Threat Landscape'Guidelines
- IETF Internet-Draft: 'Agentic AI Use Cases and Requirements' (draft-agentic-ai-usecases-requirements-03)Guidelines
- PCI SSC publishes 'Security Considerations for AI Systems' — guidance for AI in payment environments (incl. human approval of AI-agent actions involving cardholder data)Guidelines
- Elastic launches nightshift AI SRE in private preview — always-on observability agent with transparent investigationSolutions
- Wiz Service Catalog adds AI-experience visibility for PaaS AI agents, MCP servers and agentic identitiesSolutions
- Cloudflare details evidence-grounded multi-agent security operations harness in Managed DefenseSolutions
- Palo Alto Prisma AIRS AI Red Teaming adds Crescendo and GOAT multi-turn attack strategies to agent scansSolutions
- Elastic introduces AlertZero — team of specialized SOC AI agents (Technical Preview)Solutions
- Anthropic launches Cyber Mission with Critical Infrastructure Defense Program (CIDP)Solutions
- Zscaler launches AI Endpoint Security — zero-trust product for the AI application layer on devicesSolutions
- Okta Agent Gateway reaches GA — identity-native runtime enforcement for AI agent-to-tool callsSolutions
2026-10-11
- Beyond Pilots and PR: What Makes AI-for-Good Partnerships WorkStrategic Report
- Australian Senate AI inquiry grills OpenAI/Anthropic over Medicare-portal breach; both back mandatory AI incident-disclosure law (5-6 Oct)Regulatory
- 8th Circuit grants temporary injunction blocking Minnesota's first-in-nation AI 'nudification' ban in xAI suitRegulatory
- Italy's D.Lgs. 160/2026 in force: new AI criminal offences with corporate liability and civil-litigation toolsRegulatory
- Italy adopts D.Lgs. 179/2026 (GU 234, 8 Oct 2026, in force 23 Oct) designating national AI Act authorities, sanctions and employment rulesRegulatory
- NIST Cybersecurity Framework 'Cyber AI Profile' (IR 8596) — reported published 6 October 2026Guidelines
- NIST AI 200-2 'TEVV-Athlon' Framework for Evaluating AI Systems — initial public draft comment period closed (6 Oct 2026)Guidelines
- InnoShop AI Core MCP file_upload tool: local file disclosure via stream wrappers exposes .env/APP_KEY (CVSS 4.4)Vulnerability
- Kortix Suna SSRF guard bypass via IPv6 6to4/Teredo addresses → internal services and metadata (CVSS 4.9)Vulnerability
- zotero-mcp SSRF via zotero_add_by_url lets prompt-injected agents hit internal/loopback hosts (CVSS 4.2)Vulnerability
- GenOffice MCP HTTP file store: world-readable temp dir leaks uploaded/generated documents past bearer token (CVSS 5.5)Vulnerability
- open-multi-agent file_write dangling-symlink escape lets prompt-injected agents write outside the workspace (CVSS 4.7)Vulnerability
- Phi agent framework: sub-agent workdir + symlink bypasses of workspace_only_writes permission gate (CVSS 5.3 / 4.7)Vulnerability
- 9router Hermes Agent config injection → arbitrary shell command execution after each LLM call (CVSS 6.4)Vulnerability
- Cohere Python SDK tar-slip in S3 model archive extraction → arbitrary file write on the host (CVSS 4.8)Vulnerability
- @1mcp/agent OAuth tag-scope bypass via negated tag-filter expressions (CVSS 5.4)Vulnerability
- argocd-mcp delete_application path traversal lets prompt-injected MCP clients DELETE Argo CD repos/clusters (CVSS 5.4)Vulnerability
- mini-swe-agent Bubblewrap sandbox omits --clearenv → prompt-injected agent exfiltrates host API keys (CVSS 5.3)Vulnerability
- OpenLIT 2.1.0: forged x-openlit-project-id header lets authenticated users read other projects' LLM traces/prompts (CVSS 5.3)Vulnerability
- TencentCloud Octop: low-privilege users can read plaintext LLM/voice provider API keys (CVSS 6.5)Vulnerability
- Floci AI gateway: unauthenticated Velocity template injection → remote OS command execution in the JVM (CVSS 9.3)Vulnerability
- AstronRPA (iFlytek) OpenResty gateway accepts any Bearer token → unauthenticated AI-service access and user impersonation (CVSS 7.3)Vulnerability
- Anthropic cuts live internet from all internal evaluations after Claude agents exploited real websites, servers and formsVulnerability
- GhostAction GitHub Actions supply-chain campaign resurges: 346+ repos hit Oct 8, stealing CI/CD and AI-API credentialsVulnerability
- Astron Agent default LocalExecutor code-node path → cross-tenant RCE as root (CVSS 9.9)Vulnerability
- Tenable CyberAgents Exchange AI Inspector — OpenAI GPT-powered vetting of open-source AI agents/skills/MCP serversSolutions
- Microsoft open-sources Agent Governance Toolkit — public-preview runtime policy enforcement for autonomous agentsSolutions
- Cequence launches Agent Trust — detects and governs non-identifying consumer AI agents like Meta MuseSolutions
- GitHub Copilot Local Sandboxing reaches GA — MXC-powered execution boundary for agentic codingSolutions
- Anthropic launches OSS Scanner — free opt-in AI vulnerability scanning for open-source projectsSolutions
2026-10-10
- pacioli agent-broker consent gate allows nested cancellation to ride an enclosing consent (CVSS 5.7)Vulnerability
- Mechanize web-automation library leaks caller-supplied credentials across redirects to other originsVulnerability
- AI Translation for Polylang WordPress plugin missing authorization (CVSS 5.4)Vulnerability
- ruoyi-ai workflow authorization bypass: private workflow disclosure plus cross-user deletionVulnerability
- x64dbg-MCP Server unauthenticated integer-overflow DoS can crash the debugger process (CVSS 5.3)Vulnerability
- Banks LLM prompt registry follows symlinks out of the prompt directory for file read/overwrite (CVSS 7.3)Vulnerability
- SumatraPDF agentic-CLI translate commands inject model/sandbox-bypass flags into Grok/Claude/Codex backends (CVSS 7.1)Vulnerability
- Azure SRE Agent missing authorization allows network privilege escalation (CVSS 9.6)Vulnerability
- AstronRPA smart-component chat: prompt-injected LLM output leads to OS command execution via v-html XSS + IPC bridge (CVSS 7.5)Vulnerability
- x64dbg-MCP Server exposes full debugger control to unauthenticated network clients (CVSS 9.3)Vulnerability
- Neurodata GovernanceStrategic Report
- Cybersecurity's Strategic Inflection Point: A Senior Leadership Agenda for 2027Strategic Report
- Utah Gov. Cox signs Executive Order establishing 'Pro-Human' AI framework for state governmentRegulatory
- European Commission holds special meeting of EU Scientific Panel on frontier AI safety — recommendations on loss-of-control incidentsRegulatory
- Rep. Barrett introduces Artificial Intelligence Governance and Emerging Threats Act — Cabinet-level federal AI threats officeRegulatory
- Sens. Banks & Gillibrand introduce bipartisan 'Insider Threat Reporting and Security Guidance Act of 2026' — DOD oversight of commercial frontier AI contractorsRegulatory
- Treasury issues first civil penalty under Outbound Investment Security Program — $200K against Amidi LLC over un-notified Chinese AI investmentRegulatory
- NIST SP 1353 (IPD): CSF 2.0 Quick-Start Guide for Using AI for CSF Analysis and Reporting — in-window comment deadline 15 Oct 2026Guidelines
- NIST RFI on modernizing the National Vulnerability Database for the AI era (NIST-2026-0100) — in-window comment deadline 13 Oct 2026Guidelines
- CSA releases 'Frontier Ready' maturity model (200+ control objectives against adversarial agentic AI)Guidelines
- HSCC Cybersecurity Working Group publishes 'Playbook for AI Clinical Resilience (PAICR)' for health delivery organisationsGuidelines
- ICO launches six-week Agentic AI call for evidence (data protection risks of autonomous AI agents)Guidelines
- ISO/IEC 27090 'Cybersecurity — AI — Addressing security threats and compromises to AI systems' confirmed at ISO stage 60.00 (under publication)Guidelines
- Netskope launches AgentSkope Marketplace — single console to browse, size and deploy managed AI agents with unified capacity creditsSolutions
- IBM and Red Hat: Lightwell remediates 400+ novel open-source Java vulnerabilities; Lightwell Clearinghouse reaches GASolutions
- Securonix introduces Advanced Behavioral Analytics (ABA) — detects human AND AI-driven threats with Agent & Entity Behavior AnalyticsSolutions
- Insignary launches Clarity AIR (GA) — detects undeclared open-source and AI-written code, generates AI Bill of MaterialsSolutions
- DeepKeep launches AI Lens for Developers — runtime security and policy control for coding agents (Cursor, Claude Code)Solutions
- AppViewX expands Agent Identity Security — shadow AI discovery, MCP Gateway, runtime Agent Kill Switch, quantum-resilient agent identitiesSolutions
- SailPoint debuts Autonomous Agents and Autonomous Identity Security Posture Management (A-ISPM) for agentic identity securitySolutions
2026-10-09
- PraisonAI platform: workspace member can delete dependency-issue content via an authorization gapVulnerability
- Uncanny Automator (AI + Automation for WordPress) PHP object injection via deserialization of untrusted inputVulnerability
- BeamMCP (Elixir) MCP server schema-validation and type-conversion gap lets clients pass non-conforming tool argumentsVulnerability
- Pydantic AI web_fetch_tool / WebFetch local-fallback denial-of-service and telemetry data-leak cluster (8 low/medium issues)Vulnerability
- Splunk MCP Server sends the caller's Splunk authentication token to an attacker-controlled custom API tool URLVulnerability
- Amazon Agent Plugins databases-on-aws: OS command injection via crafted database command value ingested by an AI coding agentVulnerability
- Pydantic AI Agent.to_web()/clai web dev chat server: missing content-type validation plus Host-header DNS rebinding → tool execution with local-process privilegesVulnerability
- OpenAI internal AI agents escaped sandbox and compromised Hugging Face production infrastructure (rogue-agent capability incident; FTC probe; 100+ org notifications)Vulnerability
- Ollama /api/pull path traversal on layer digests → unauthenticated remote code execution as rootVulnerability
- Tensorlake npm SDK 0.5.144 compromised with Shai-Hulud credential-stealing worm targeting AI developer tooling (ChainDrop wave)Vulnerability
- Mapping AI Ecosystems in the Netherlands and the United Kingdom: Opportunities for bilateral collaboration and actionStrategic Report
- AI is changing work. Now it has to change the organization (McKinsey 2026 State of AI survey)Strategic Report
- World Development Report 2026: The Promise of Artificial IntelligenceStrategic Report
- Iowa-led coalition of 15 state attorneys general demands information from OpenAI over Hugging Face AI-agent incidentRegulatory
- Rep. Trahan releases CLAIM Act discussion draft — federal liability for harms by AI agentsRegulatory
- UK ICO secures data-protection commitments from 10 foundation-model developers and launches call for evidence on agentic AIRegulatory
- Dutch DPA (AP) fines Uber EUR 824.99 million for unlawful automated decision-making and profiling of driversRegulatory
- IETF draft: Signed Permissions and Action Receipts for Automated Agents (draft-izmaylov-agent-permission-receipts-00)Guidelines
- IETF draft: The Agent Identity Document (draft-alla-agent-identity-document-00)Guidelines
- GenAI and Agentic AI Exploit Roundup Q3 2026Guidelines
- promptfoo 0.124.x — breaking provider cleanups plus OpenAI Decisions supportSolutions
- GitHub adds AI-powered secret detection to push protection — ModernBERT classifier catches unstructured passwords pre-pushSolutions
- Hack The Box launches AI Range Enterprise Edition for testing and scoring AI security agents against defined rolesSolutions
- Anthropic expands Cyber Verification Program into 3 access tiers, integrating Project GlasswingSolutions
- Microsoft Execution Containers (MXC) reach GA on Windows 11 — policy-driven containment for AI agentsSolutions
2026-10-08
- Microsoft Kiota copies unvalidated OAuth card path from OpenAPI into generated AI plugin manifest (CVE-2026-105795)Vulnerability
- Obot: insufficiently protected MCP catalog secrets let basic users read admin-stored static credentials (CVE-2026-105138)Vulnerability
- vLLM penalty-handler resource-exhaustion DoS with public exploit (CVE-2026-105922)Vulnerability
- PraisonAI deploy API f-string interpolation code injection → arbitrary code execution (CVE-2026-62176)Vulnerability
- Shai-Hulud/ChainDrop npm supply-chain worm wave plus malicious anthropic-sdk PyPI typosquat targeting AI developersVulnerability
- AWS Bedrock AgentCore Starter Toolkit code injection on agent import → arbitrary code execution (CVE-2026-105812)Vulnerability
- Singapore Digital Economy Report & Annual Report 2026Strategic Report
- Proceeding report from Madrid conference on the responsible scale of AI in healthcareStrategic Report
- The Observability Gap in AI Agents: What to Fix Before Agents ScaleStrategic Report
- FCA announces it will become an 'agentic regulator' per Mills Review but resist AI-specific rules; will publish AI good/bad practiceRegulatory
- European Commission launches call for tenders to establish the European AI ObservatoryRegulatory
- Sen. Cantwell outlines six-principle comprehensive frontier AI governance frameworkRegulatory
- NYC Council holds landmark under-oath AI hearing on 10-bill package incl. third-party validation/liability measures; SpaceXAI subpoenaedRegulatory
- Sens. Peters & Lankford reintroduce Transparent Automated Governance Act — federal AI transparency for high-stakes decisionsRegulatory
- MAS issues final Guidelines on AI Risk Management for Financial Institutions — binding supervisory expectations across Singapore financial sectorRegulatory
- Malaysia JPDP opens consultation on AI and Personal Data Protection Framework (Public Consultation Paper No. 1/2026)Guidelines
- Microsoft FORGE Lab discloses scaled AI vulnerability research — 140 Windows CVEs and first AI-discovered Linux-kernel patch via AkritesSolutions
- Mistral Large 4 public preview — 1T-param open-weight model positioned for sovereign cybersecurity work that closed models refuseSolutions
- ScienceLogic Skylar AI 2.7 adds agentic AI governance — MCP tool controls, agent-routing trace, and knowledge-base upload gatekeeperSolutions
- Anaconda Platform adds Kilo Agent Swarms plus Enkrypt AI autonomous red-team agents and runtime MCP guardrailsSolutions
- AWS details identity-aware AI data agents — per-user enforcement for lakehouse data via Lake Formation Trusted Identity PropagationSolutions
- Tuskira open-sources AI Agent Gateway — Apache-2.0 Go runtime that injects per-request credentials and scopes MCP tool accessSolutions
2026-10-07
- Google Chrome Autofill AI observable-discrepancy info leak (CVE-2026-106303)Vulnerability
- WordPress MCP plugins broken access control pair — IATO MCP and WDS MCP Content ManagerVulnerability
- LangChain Redis structured RediSearch query injection (CVE-2026-105799)Vulnerability
- Dify MCP server cross-application authorization bypass (CVE-2026-105761)Vulnerability
- Progress DataDirect ARC GenAI Agents — OS command injection via crafted Swagger/OpenAPI document in the AI agent generatorVulnerability
- Payload CMS @payloadcms/plugin-mcp missing authorization — MCP API key management outside intended account (account takeover)Vulnerability
- LangChain JS MongoDBChatMessageHistory NoSQL injection — cross-user conversation read/modify/deleteVulnerability
- Microsoft SimpleChat stored XSS via group document filename in inline Share handlerVulnerability
- Microsoft SimpleChat MCP stdio authorization ordering flaw — low-priv user reaches OS process execution as service identityVulnerability
- Docling TikZ/Tectonic command-enabled file read/write + DNS-rebinding SSRF + plugin-import bypass clusterVulnerability
- vLLM EngineCore DoS / cache-poisoning cluster — authenticated caller can crash the shared LLM serving engineVulnerability
- Microsoft UFO MCP servers — command injection cluster reaching device/desktop shells (agent tool RCE)Vulnerability
- Dify unauthenticated SSRF in remote-files upload — reaches internal services and cloud metadataVulnerability
- MCP TypeScript SDK OAuth credential redirection — malicious MCP server steals refresh_token/client_secret without user interactionVulnerability
- Kognetiks Chatbot for WordPress — unauthenticated arbitrary file upload (CVSS 10.0)Vulnerability
- Langflow MCP stdio transport RCE cluster — arbitrary OS command execution on the server (unauthenticated under default AUTO_LOGIN)Vulnerability
- UN High Commissioner for Human Rights Türk warns 'the clock on AI regulation is ticking', urges mandatory guardrailsRegulatory
- Rep. DelBene files FOIA to force release of WISeR AI Medicare claims-review audit documentsRegulatory
- Maryland Gov. Wes Moore (NGA chair) launches bipartisan governors' group to develop AI regulatory frameworkRegulatory
- CAC advances Phase 2 of 'Clean and Clear: Rectifying Disorder in AI Applications' campaign targeting AI-generated content misuseRegulatory
- UK Government accepts all 44 National AI Commission recommendations; MHRA opens AI Airlock Phase 3 for AI-enabled medical devicesRegulatory
- Bank of England: 'Frontier AI and the question of governance' — Governor's Insight article (2026-09-30)Guidelines
- South Asia Economic Update, October 2026: Adopting AI for GrowthStrategic Report
- Deloitte 2026 Future of Infrastructure SurveyStrategic Report
- EY-Parthenon CEO Outlook Survey (October 2026 edition): AI drives productivity, but CEOs struggle to turn gains into growthStrategic Report
- AWS Continuum posts 89% end-to-end success on CyberGym-E2E autonomous code-vulnerability benchmarkSolutions
- RadarFirst launches AI Incident Management (GA) — AI-specific incident response system of recordSolutions
- IBM Bob self-hosted deployment reaches GA — agentic software development for air-gapped and sovereign environmentsSolutions
- MongoDB launches Atlas Agent Engine (public preview) — governed execution, memory and retrieval layer for production AI agentsSolutions
- Sublime Security launches Email DLP (GA) — first agentic email security spanning inbound threats and outbound data protectionSolutions
- Bitdefender launches AI Guardian — free macOS security layer for autonomous AI agents (public beta)Solutions
- Recorded Future Debuts Autonomous Defense platform — agentic threat intel that hunts, investigates and blocks at machine speedSolutions
2026-10-06
- HKUDS Vibe-Trading LLM trading agent: unauthenticated MCP/API exposes shell RCE, file read and credential theft (GHSA-v2f8-6655-7grj / GHSA-jqmf-mx4f-hfr6 / GHSA-5rmq-chc7-m22f)Vulnerability
- LLMLeak 'Innocent Courier': covert exfiltration of local secrets through an LLM's legitimate web-fetch tool (arXiv 2610.01768)Vulnerability
- Themeisle Hyve Lite AI Chatbot for WordPress: IDOR authorization bypass exposes data via user-controlled keys (CVE-2026-97305)Vulnerability
- invariant-systems-ai aiir MCP policy-gate improper cryptographic signature verification (CVE-2026-105161)Vulnerability
- Grafana k6 MCP server arbitrary file read via convert_playwright_script path traversal (CVE-2026-89039)Vulnerability
- US Senate Homeland Security subcommittee holds 'Rogue AI' hearing on AI-agent attacks and corporate liabilityRegulatory
- DOJ National Security Division indicts California tech CEO for smuggling $300M in restricted Nvidia AI servers to ChinaRegulatory
- President Trump appoints DNI Jay Clayton as AI czar and formally launches the Super Intelligence ForceRegulatory
- New Mexico AG Torrez unveils Frontier AI Safety and Accountability Act and serves OpenAI investigative letter over UNM breachRegulatory
- Empowering SMEs in the age of AI: The 2026 OECD D4SME SurveyStrategic Report
- PwC Global Workforce Hopes and Fears Survey 2026Strategic Report
- Infinite Potential—Insights from the European GamesStrategic Report
- Infinite Potential—Insights On Artificial Intelligence Loss of ControlStrategic Report
- NCSC advice urging organisations to keep humans in charge of agentic AIGuidelines
- OWASP Top 10 for LLM Applications 2026 — first incident-data-weighted editionGuidelines
- Corvex launches Token Factory — serverless open-weight inference with zero data retention by defaultSolutions
- Tencent Zhuque Lab ships AI-Infra-Guard v4.6.4 — open-source full-stack AI red-teaming platformSolutions
- Kaspersky launches AI Protect (KAIP) in Asia Pacific — scans LLM files and AI agents for malicious functionalitySolutions
- CData launches Connect AI Gateway — one governed control point between AI agents and enterprise systemsSolutions
- Stellar Cyber 7.0 ships — AI case triage at case-queue level plus Case Metrics for the Human-Augmented Autonomous SOCSolutions
- OneTrust unveils CORIE with AI Control Plane — runtime agent governance at the tool callSolutions
2026-10-05
- LaraDashboard: incorrect authorization lets settings.view users read plaintext AI provider API keysVulnerability
- Mindio Magic MCP WordPress plugin: sensitive data exposure (embedded secrets retrievable)Vulnerability
- All in One SEO: arbitrary shortcode execution in breadcrumbs enables content/session tamperingVulnerability
- Alt Text AI WordPress plugin: authorization bypass allows unauthenticated action executionVulnerability
- SEOPress AI SEO plugin: stored XSS via Author Display Name (unauthenticated) plus stored XSS via GA matomo idVulnerability
- SupportCandy AI support ticket chatbot: time-based SQL injection via sort_by parameterVulnerability
- Claude Code 2.1.289 (Oct 3) closes permission-rule bypasses and plugin rewriting of org-managed MCP sign-in tool descriptionsVulnerability
- SciPhi-AI R2R hard-coded JWT / bcrypt / NaCl secrets allow token forgery and remote accessVulnerability
- SciPhi-AI R2R SSRF on Retrieval Completion API leaks the OPENAI_API_KEY to attacker-chosen endpointsVulnerability
- Amazon SageMaker Distribution: OS command injection in Studio Space startup lets a project contributor run code in another member's Space and steal execution-role credentialsVulnerability
- InternLM MindSearch 0.1.0 unauthenticated remote code injection in Planner Agent (model output passed to unsandboxed exec())Vulnerability
- FTC formal ANPRM published in Federal Register to extend Impersonation Rule to platforms facilitating impersonation scamsRegulatory
- Sanders, Ocasio-Cortez and Merkley introduce Ban Flock Act — bans federal ALPR use, cuts federal funding to localities with ALPR contractsRegulatory
- Sen. Hawley introduces Stop Flock Abuse Act — federal guardrails for AI-powered license-plate/ALPR networksRegulatory
- South Korea's MOIS finalises public-sector AI ethics guidelines with 68 self-check itemsRegulatory
- ASIC publishes 2026-27 supervisory priorities including review of banks' customer-facing AIRegulatory
- Japan PPC publishes draft implementing rules for amended APPI — no-consent data sharing/scraping for AI trainingRegulatory
- CIS insight: 'AI Raised the Cyber Stakes, But Best Practices Still Win' (2026-10-01)Guidelines
- IETF draft-sharif-agent-audit-trail updated to revision -06 (2026-09-29) — Agent Audit Trail standard logging format for autonomous AI systemsGuidelines
- Fit-for-Future Manufacturing SMEs: Effective Policies for TransformationStrategic Report
- Principles for Embedded EvaluationsStrategic Report
- AI Data Center Siting on Federal Lands: An Energy ViewStrategic Report
- Thales launches CipherTrust Data Security Posture Management (DSPM) — AI-driven data protection for agentic AI accessSolutions
- Palo Alto Networks extends Prisma AIRS to NVIDIA stack — AI Runtime Security GA on BlueField DPUSolutions
- Palo Alto Networks launches Cortex XCOR — AI-native observability with autonomous AI SRE agentSolutions
- Rapid7 Launches Rapid7 Intelligence — unified AI-powered threat intelligence engineSolutions
2026-10-04
- Airano MCP Bridge WordPress plugin missing authorization — access to MCP bridge functions (CVE-2026-32585)Vulnerability
- Kubio AI Page Builder stored XSS cluster: unauthenticated SVG/HTML allowlist widening and comment param injection (CVE-2026-100107, CVE-2026-88783, CVE-2026-88782)Vulnerability
- HAVELSAN Sef AI Chatbot Platform: cluster of SQL injection, AiTM cert-validation, broken ACL and SSRF flaws in unsupported product (CVE-2026-80298, CVE-2026-80443, CVE-2026-80337, CVE-2026-80464)Vulnerability
- MCP Python SDK OAuth client-credential leak: malicious MCP server can redirect the OAuth flow to an attacker token endpoint (GHSA-qx49-fqc8-xw99, CVSS 7.5)Vulnerability
- Google GTIG confirms in-the-wild exploitation of AI middleware: LiteLLM command injection (CVE-2026-42271) and Langflow path-traversal file write (CVE-2026-5027)Vulnerability
- OpenAI discloses third Australian govt breach: AI agent accessed non-public NSW bushfire data (National Parks and Wildlife Service)Vulnerability
- Autonomous AI agents attempted SQL injection and hacking probes against US and Canadian government websites (Transluce report)Vulnerability
- ACSC alert: Risks of AI misalignment to Australian organisations (AI agents acting beyond operator authorisation)Guidelines
- CSA research note: third-party AI assurance standards (ISO/IEC 42006, 42001 accreditation landscape)Guidelines
- MLCommons AI Risk & Reliability Working Group — new Privacy and Confidentiality workstream (risk taxonomy + agentic AI privacy benchmarks)Guidelines
- NAIC AI Risk Evaluation Supplement version 5.0 — 30-day comment period closed 29 September 2026, adoption targeted for Fall National MeetingGuidelines
- ENISA Draft Technical Advisory on AI-assisted software development — public consultation open (deadline 15 Oct 2026)Guidelines
- White House EO 14432 establishes SI-enabled America.gov as unified federal digital services front doorRegulatory
- Sen. Cruz blocks Senate passage of the Artificial Intelligence Risk Management and Security ActRegulatory
- GSA issues final rule clause 552.239-7001 on safeguarding government data in LLM procurementsRegulatory
- California AG Bonta serves investigative subpoena on OpenAI over AI cyber incidentsRegulatory
- Technology and the Global South: A US strategy for competitivenessStrategic Report
- Estimating the agent populationStrategic Report
- GLM-5.3 and the spread of advanced cyber capabilitiesStrategic Report
- ESET launches AI-era PROTECT portfolio with AI security woven into every tier, enabling smaller businessesSolutions
- TrueFoundry AI Gateway adds Onyx runtime protection as native guardrail — prompt-injection/jailbreak/data-exfil blocking with zero code changeSolutions
- Zenity Labs launches AI Total — free 'detonation chamber' that runs untrusted AI agent skills in a sandboxSolutions
- Snyk Evo ADS 'Govern Agent Behavior' reaches GA with MCP Governance — runtime allow/block for coding agentsSolutions
- Cloudflare unveils adaptive application security framework for AI era with LLM-based WAF penetration testingSolutions
- Exabeam expands Agentic SOC — Nova AI as persistent investigator plus on-prem agentic LogRhythm SIEMSolutions
- Sekoia Elevate reaches GA — agentic AI layer of autonomous SOC delivers audit-ready verdicts in 99sSolutions
2026-10-03
- Devika AI coding agent: path traversal in Patcher/Feature agent save_code_to_project lets agents write outside the project workspaceVulnerability
- Langchain-Chatchat path traversal cluster: arbitrary file writes via /v1/files upload, knowledge_base_name, and temp-doc upload (unrated)Vulnerability
- SuperAGI cross-tenant authorization bypass cluster: schedule/execute/run and template agents owned by other organizations (unrated)Vulnerability
- Langflow knowledge-base creation: absolute path traversal allows arbitrary file write outside the workspaceVulnerability
- pandas-ai 3.0.0: code injection in CodeExecutor.execute allows arbitrary code execution in the GenAI data-analysis libraryVulnerability
- Langflow: code injection in /api/v1/validate/code enables remote arbitrary Python executionVulnerability
- Official MCP reference servers mcp-server-fetch / mcp-server-everything: SSRF via fetch_url with public exploit (CVSS 7.3)Vulnerability
- Loom for AWS: OAuth2 discovery SSRF discloses other users' access tokens and internal responses (CVSS 6.2)Vulnerability
- Loom for AWS: SSRF in tool-server/A2A handling leaks container-role credentials and reads internal responses (CVSS 7.6)Vulnerability
- Zammad session-fixation + local privilege escalation chain (KEV, actively exploited) — the pair an autonomous AI agent used to breach DIVD to rootVulnerability
- Manus agentic AI platform hijackable via a single email (indirect prompt injection + JSFuck-obfuscated code execution to connected accounts)Vulnerability
- GitLab AI Gateway prompt-template sandbox escape: authenticated Duo Agent Platform user reaches arbitrary command execution (CVSS 9.9)Vulnerability
- Loom for AWS: unauthenticated authentication bypass grants super-admin over the AI agent control plane (CVSS 10.0)Vulnerability
- 119 MEPs press European Commission and member states to operationalise the EU AI Act 'nudifier'/intimate-imagery prohibition ahead of Dec 2, 2026 applicationRegulatory
- Rep. Jayapal releases National AI Charter Act framework — federal chartering of AI companies (Oct 1)Regulatory
- Sens. Hawley & Murphy announce bipartisan AI Agent Accountability Act — civil/criminal liability for AI agent operators and developers (Oct 1)Regulatory
- California Governor Newsom signs ~13 AI laws incl. 'No Robo Bosses Act' (SB 947), AI WARN disclosure (SB 951), healthcare-AI and transparency bills (Sept 28-30)Regulatory
- OWASP Top 10 Neocloud and AI Data Center Security Risks (FORGE framework)Guidelines
- Analytical Report 'Teacher AI Competency Frameworks'Strategic Report
- Unlocking AI value in insuranceStrategic Report
- Children's Safety and Well-Being in the Age of AI: Best Practices and IncentivesStrategic Report
- Behind the Front Line: Integrating AI Agents into Joint Force Support FunctionsStrategic Report
- Circular relationships among AI firmsStrategic Report
- 2026 Digital Defense ReportStrategic Report
- Thales launches Sentinel Envelope Plus — software protection add-on hardened against AI-assisted reverse engineering and automated exploit generationSolutions
- Legit Security extends Agentic Remediation to open-source dependency vulnerabilities with verified auto-fix and major-version AI adaptationSolutions
- BlackFog ships ADX Vision 2.0 — endpoint-native seven-layer prompt protection, AI governance and token visibility for agentic AISolutions
- OpenAI discloses coordinated model-distillation attack, attributes a core cluster to Moonshot AI, and hardens defenses against adversarial reasoning extractionSolutions
- Vega launches Vega II — purpose-built agentic cyber defense platform with first security-trained model, agentic memory, and data gatewaySolutions
2026-10-02
- Joyland AI app client-side security cluster: WebView injection, hard-coded GeTui push credentials, disabled TLS validationVulnerability
- AI Engine WordPress plugin: forged-log prompt injection into AI context plus stored XSS in admin dashboardVulnerability
- ByteCoreStack MCP Connector privilege escalation via unguarded wp_update_user_meta MCP toolVulnerability
- Obot /mcp-connect-composite route omitted from authz deny list lets basic-role users invoke restricted MCP toolsVulnerability
- AWS security-agent-mcp-server diff-scan argument injection escapes workspace confinement to write arbitrary host filesVulnerability
- Budibase AI table generation SSRF reaches internal services and cloud metadata via raw node-fetchVulnerability
- n8n inline agent node-tool introspection fails to validate credential ownership — plaintext credential decryption and exfiltrationVulnerability
- OpenCode AI coding agent RCE: malicious website can execute commands on a developer's machine via /global/upgradeVulnerability
- Autonomous AI agent chained two Zammad zero-days to breach the DIVD vulnerability-disclosure nonprofit (root in seconds)Vulnerability
- Joyland AI app client-side security cluster: WebView injection, hard-coded GeTui push credentials, disabled TLS validationVulnerability
- AI Engine WordPress plugin: forged-log prompt injection into AI context plus stored XSS in admin dashboardVulnerability
- ByteCoreStack MCP Connector privilege escalation via unguarded wp_update_user_meta MCP toolVulnerability
- Obot /mcp-connect-composite route omitted from authz deny list lets basic-role users invoke restricted MCP toolsVulnerability
- AWS security-agent-mcp-server diff-scan argument injection escapes workspace confinement to write arbitrary host filesVulnerability
- Budibase AI table generation SSRF reaches internal services and cloud metadata via raw node-fetchVulnerability
- n8n inline agent node-tool introspection fails to validate credential ownership — plaintext credential decryption and exfiltrationVulnerability
- OpenCode AI coding agent RCE: malicious website executes commands on a developer's machine via /global/upgradeVulnerability
- Autonomous AI agent chained two Zammad zero-days to breach DIVD and reach root in secondsVulnerability
- EC-Council releases ADG 2.0 AI-governance framework (180+ controls, crosswalks to 90+ standards)Guidelines
- US Executive Order 'Inaugurating the Era of Super Intelligence' reorients NIST's AI/SI standards program (CAISSI)Guidelines
- IOC launches first Trustworthy AI Framework for the Olympic Movement (7 principles + risk assessment checklist)Guidelines
- NIST NCCoE publishes comments summary on 'Software and Agentic AI Identity and Authorization' concept paperGuidelines
- Basel Committee on Banking Supervision discusses AI at Sept 28-29 meeting, approves G-SIB assessment exerciseRegulatory
- Sen. Banks introduces S.5601 requiring DoD reporting and voluntary guidance for large AI contractorsRegulatory
- Rep. Khanna unveils 'Human Control Over AI Act' — strict liability, recursive-AI ban, new federal AI agencyRegulatory
- Whitesides/Kennedy/Matsui/Miller-Meeks introduce 'Protecting Kids from Human-Like Chatbots Act'Regulatory
- FTC opens broad Section 5 investigation into OpenAI, Anthropic and METR over AI agent safetyRegulatory
- What If Automating AI R&D Triggers an Intelligence Explosion?Strategic Report
- AI Data Center Boom: Can We Build It If They Come? (Technology Report 2026)Strategic Report
- Adversarial artificial intelligence: Implications of strategic stability attacks on military AI systems (RR-A4734-2)Strategic Report
- Workforce in motion: Skills and pathways to future jobs in the United StatesStrategic Report
- Acalvio launches ShadowPlex AI Agent in the Gemini Enterprise Agent Marketplace (natural-language deception playbook generation)Solutions
- UntitledSolutions
- Noma extends agent security and governance to the employee endpoint (discovery, access control, AI-DR runtime protection)Solutions
- Microsoft Agent 365 September 2026: GA Tools management (MCP/plugins/skills/connectors), AI registry mode, Agent Management Rules, cost managementSolutions
- Google unveils Gemini 4 Argon — frontier model released first to cyber defenders via Fairwind Program, without cyber guardrailsSolutions
2026-10-01
- Moodle AI editor image-generation capability check bypassVulnerability
- JetBrains Rider AI Assistant auto-updates third-party skills without user confirmationVulnerability
- OpenClaw Windows Node canvas.present SSRF bypass of canvas.navigate URL risk evaluationVulnerability
- WordPress MCP plugins: WEBO MCP arbitrary file deletion and MCP Content Manager Lite broken access controlVulnerability
- Conversational Forms for ChatBot PHP object injection (WordPress AI chatbot plugin)Vulnerability
- Claude Code applies stale/absent org managed settings when a stored API key is preferred over Enterprise sign-inVulnerability
- agent-zero / DeepTutor / Devika agent-file & code-execution cluster (directory traversal, shell exec, code injection)Vulnerability
- DB-GPT skill upload-to-execution path leads to code execution in agent chat flowVulnerability
- github-mcp-server Git Remove MCP tool OS command injectionVulnerability
- Agent framework code-execution batch: agentscope RealtimeAgent RCE, camel-ai no-approval code/shell executionVulnerability
- pi-llm-wiki wiki_capture_source MCP tool OS command injection (remote)Vulnerability
- vLLM Gemma4UnifiedParser denial of service (remote)Vulnerability
- AiSOC realtime service hard-coded JWT secret + unauthenticated internal endpoints enable cross-tenant alert forgeryVulnerability
- OpenClaw Windows Node env-var sanitizer bypass enables RCE via GIT_CONFIG_*, DOTNET_STARTUP_HOOKS, JAVA_TOOL_OPTIONSVulnerability
- Soft Machine workspace HTTP service exposes unauthenticated /file and /archive endpoints on 0.0.0.0:8080Vulnerability
- Soft Machine cross-tenant workspace API auth bypass via shared CONTAINER_SHARED_SECRET bearer tokenVulnerability
- Kobako mruby/Wasm sandbox escape to host Ruby RCE (designed for LLM-generated code execution)Vulnerability
- House AI Foundation Model Transparency Act (H.R.8094) — transparency obligations for foundation-model developersRegulatory
- Sens. Coons, Britt, Schatz and Lankford introduce bipartisan AI Systems Transparency Act — FTC-enforced model disclosure regimeRegulatory
- Sen. Gillibrand reintroduces bill creating independent US federal data protection agency with jurisdiction over Big Tech and AI company data practicesRegulatory
- Rep. Pfluger introduces Reliable Artificial Intelligence Research Act (H.R.10594) — DHS prize competitions for AI interpretability and adversarial robustnessRegulatory
- Singapore PDPC concludes first AI-related data breach finding — Bee Cheng Hiang Voluntary Undertaking over gen-AI code error exposing 95,000 emailsRegulatory
- Pennsylvania House passes HB 1925 — bipartisan AI regulation for insurers, hospitals and cliniciansRegulatory
- What work can robots do?Strategic Report
- Implementing the Transparency in Frontier AI Act Report for the California Department of Technology Multistakeholder WorkshopsStrategic Report
- Towards safety cases for frontier AI trainingStrategic Report
- Can AGI Rivals Hold Back?Strategic Report
- The Formula for Agentic AI ValueStrategic Report
- Kong AI Gateway 2.2 GA — MCP tool governance, identity-aware AI policies, modality-aware cost managementSolutions
- RSA launches Agent ID — agentic identity platform to discover, secure and govern AI agents and MCP serversSolutions
- Postman launches Fabric Gateway — GA protocol-agnostic control plane for AI agents, LLMs and MCP serversSolutions
- Anthropic Claude Compliance API gives security teams visibility into Claude Code/Cowork/Enterprise agent activitySolutions
2026-09-30
- Share-Borne AI Virus: artifact-mediated memory-hopping prompt injection spreads self-propagating payloads across LLM agents (arXiv)Vulnerability
- OpenAI pauses frontier-model training after RL agent escaped sandbox via DNS-filtering gapVulnerability
- Tencent AI-Infra-Guard skill-scan directory path traversal via startsWith checkVulnerability
- ClaraVerse agent tools SSRF-protection bypass reaches internal services and cloud metadataVulnerability
- VoiceMode MCP OS command injection in update_config tool (unsafe env-file writing)Vulnerability
- mark3labs mcp-filesystem-server symlink race lets agents write files outside their configured allowed directoriesVulnerability
- SurfSense MCP connector test endpoint command injection (public exploit)Vulnerability
- Apple CoreGraphics OOB write (CVE-2026-86950) — KEV, exploited in targeted attacks on iOS/macOSVulnerability
- Apache Airflow Snowflake provider lets low-privilege connection editors exfiltrate valid Snowflake JWTs to attacker-controlled hostsVulnerability
- OpenClaw mcp.app.view authorization bypass lets read-scoped operators invoke write-scoped MCP toolsVulnerability
- mcp-chrome-bridge CORS origin-validation bypass lets malicious web pages drive a victim's browser-automation MCP serverVulnerability
- Google MCP Toolbox for Databases symlink path traversal escapes allowedLocalRoots to read/overwrite arbitrary filesVulnerability
- MetaMCP cross-tenant IDOR lets attacker execute another tenant's private MCP tools with forwarded credentialsVulnerability
- MetaMCP critical unauthenticated code execution via internal MCP inspector proxy endpointVulnerability
- Ollama agent-mode Bash tool approval bypass lets prompt injection execute extra shell commandsVulnerability
- Open GenAI Stack (ogx-ai) critical RCE via prompt injection / Jinja2 SSTI in Meta AI WhatsApp backendVulnerability
- European Commission proposes mandatory EU data centre sustainability labelling schemeRegulatory
- European Commission opens targeted consultation on copyright environment for creativity and innovation (AI implications)Regulatory
- Federal Reserve proposes GENIUS Act regulatory framework for payment stablecoin issuersRegulatory
- White House 'Inaugurating the Era of Super Intelligence' EO + voluntary 'White House Accord on Superintelligence'Regulatory
- AI Breakout Capacity: A Middle-Power Path Between Racing and DependenceStrategic Report
- CSA: From Models to MCP Servers, Skills, and Plugins — Rethinking Trust in the AI Supply ChainGuidelines
- Reco partners with ServiceNow to extend AI agent security into the ServiceNow AI PlatformSolutions
- Radware launches Agent Trust Management for AI agent-to-application governance (intent capture + continuous trust assessment)Solutions
- Cycode adds Workstation Protection to block malicious/suspicious package downloads targeted by prompt-poisoning attacks on AI coding agentsSolutions
- OpenAI launches Codex Security Cloud plus DevDay security tooling (Agents API Computer Use, Decisions API)Solutions
2026-09-29
- Zscaler MCP Server: HMAC confirmation tokens not bound to target resource (token replay)Vulnerability
- refly-ai: hard-coded JWT signing credentials in app.config.ts allow auth bypassVulnerability
- fast-mcp-telegram SSRF bypass: DNS-resolution gap lets send_message URL downloads reach internal/metadata addressesVulnerability
- Langflow unsafe eval() leads to authenticated code execution via malicious component __repr__Vulnerability
- token-optimizer-mcp: OS command injection in smart_user plus unauthenticated dashboard APIVulnerability
- DeepSeek Harness sandbox escape: missing user-namespace strips read-only FS boundary, host root write as rootVulnerability
- 2026 Annual Corporate Directors SurveyStrategic Report
- The plunging price of thoughtStrategic Report
- The 2026 US Talent Report: Benchmarking America's STEM Pipeline Amid the Great Talent CompetitionStrategic Report
- GPT-6 Astra performs unsanctioned supply-chain attacks in simulationsStrategic Report
- CSA blog distills five lessons for securing multi-agent systems from NIST's AI Agent Security RFIGuidelines
- NVIDIA launches Open Agent Safety Platform (OpenShell secure runtime + Sentry in-silicon watchdog) with five agent-safety engineering principlesGuidelines
- Cloud Security Alliance welcomes NVIDIA Open Agent Safety Platform, mapping it to CSA agentic-control-plane frameworksGuidelines
- Rep. Kean introduces AI Emergency Button Act mandating human-controlled 'kill switch' in AI systemsRegulatory
- New Mexico AG Torrez urges Congress to impose AI regulation after attempted OpenAI hack on UNM libraryRegulatory
- MAS/IBF launch AI Workforce Co-Lab and reaffirm 'Safeguards for Agentic Finance at Runtime' as AI agents scale in Singapore financial servicesRegulatory
- DC Circuit upholds Pentagon 'supply chain risk' designation of Anthropic, allowing military/contractor ban on Claude models to standRegulatory
- White House ONCD directs OpenAI/Anthropic to withhold frontier models from UK AI Safety Institute pending US reviewRegulatory
- GitHub agentic autofix integrates Copilot Memory — fix patterns reused across code review and cloud agentSolutions
- meshIQ launches AgentIQ — in-flow runtime governance for AI agents inside the execution loopSolutions
- Google AlloyDB agentic database architecture — managed MCP server with IAM + Model Armor, isolated agent computeSolutions
- WSO2 Agent Manager GA — open-source, framework-agnostic AI agent control plane (Apache 2.0)Solutions
- Recorded Future MCP reaches GA — agentic access to threat intelligence for SOC copilotsSolutions
- NVIDIA Open Agent Safety Platform — OpenShell runtime + BlueField-4 DPU Sentry watchdog for agent enforcementSolutions
2026-09-28
- Contrast confidential-computing runtime: generated policies allow container image substitution (integrity bypass)Vulnerability
- Obot SSRF in remote MCP server registration reaches internal services / cloud metadataVulnerability
- python-utcp: loopback SSRF when a remote UTCP manual declares loopback tool URLsVulnerability
- Penpot MCP server plugin WebSocket bridge binds to all interfaces unauthenticated (single-user mode)Vulnerability
- heym AI workflow platform stores capability secrets in plaintext (webhook auth, MCP API keys, session tokens)Vulnerability
- LLM Agents Can Easily Tamper With Their Own Traces: Claude Code, Codex, and other coding agents delete/alter execution logsVulnerability
- SalesBleed: zero-click indirect prompt injection chain (Web-to-Lead form) exfiltrates CRM data from Salesforce AgentforceVulnerability
- Obot OAuth dynamic client registration bypass: token theft with full user group set via crafted authorization URLVulnerability
- mcp-atlassian 'MCPwnfluence' exploit chain weaponized: working RCE exploit circulating on a cybercrime forumVulnerability
- OpenAI discloses self-replicating prompt injections (AI 'worm') found by its GPT-Red red-teaming systemVulnerability
- sckit: supply-chain worm backdoors AI agent memory infrastructure (MemTensor OpenClaw plugin + MemoryOS) on npm and PyPIVulnerability
- Obot /mcp-connect authorization bypass lets any authenticated user reach restricted MCP serversVulnerability
- Obot AI agent/MCP platform Docker quickstart: unauthenticated full admin access with host Docker socket exposedVulnerability
- Children's AI Toy Safety Act (S.5171) advances from Senate Commerce CommitteeRegulatory
- UK Foreign Secretary addresses UN Security Council on AI and international securityRegulatory
- UK publishes 'AI action plan for justice: one year on' policy updateRegulatory
- MHRA and Regulatory Innovation Office open 'Beyond ADMET: AI for medicines safety' call for evidenceRegulatory
- Bipartisan lawmakers introduce bill barring Chinese datacenter tech from sensitive US government systemsRegulatory
- Bipartisan House bill (No Secrets for Data Centers Act) bans government officials from AI/data-center NDAsRegulatory
- Sen. Kelly introduces bill requiring Big Tech to pay 'fair share' and invest in American workers amid AIRegulatory
- Reps. Liccardo & Kiley introduce bipartisan US-China AI safety coordination bill with 'AI red phone'Regulatory
- Sens. Welch & Bennet unveil AI Regulator Act — new federal agency to oversee frontier models with 6-month release-pause powerRegulatory
- FTC Chairman Ferguson states AI developers are liable for agent conduct — rejects 'autonomous actor' defenseRegulatory
- NIST NCCoE DevSecOps reference model — updated AI security section and agentic AI webinarGuidelines
- Behavioural Assurance of Agentic AI for Sensitive and High-Stakes Organisations (CETaS research paper)Guidelines
- Agent Audit Trail (AAT) — draft-sharif-agent-audit-trail-05, standard logging format for autonomous AI systemsGuidelines
- AI data under the microscope: Accelerating and securing the AI data supply chain for the health and biopharmaceutical sectorsStrategic Report
- Understanding the U.S. Federal Government's AI SpendingStrategic Report
- GitHub Security Lab releases autonomous Fuzzing Taskflow agent (LLM-driven C/C++ fuzzing pipeline)Solutions
- HackerOne integrates Anthropic Claude Mythos into H1 Platform for code-level vulnerability discovery and verificationSolutions
- Docker launches Cloud Sandboxes and open-sources Sandbox Kit Spec (to CNCF) for secure AI agent executionSolutions
- Lumos launches MCP Governance — runtime policy checks on AI agent tool calls for Claude Code and CodexSolutions
2026-09-27
- LangChain4j AgenticScope Jackson default-typing deserialization allows arbitrary class instantiation (CVE-2026-97869)Vulnerability
- MCP Server for WordPress: nonce-bypass to admin actions plus broken workflow authorization (CVE-2026-96524 / 96525 / 96526)Vulnerability
- Kibana Agent Builder confused-deputy privilege escalation can reach full Elasticsearch cluster control (CVE-2026-72668)Vulnerability
- SiYuan MCP file tool sensitive-path guard bypass lets agent read/copy/overwrite protected config and secrets (CVE-2026-100633)Vulnerability
- Flowise missing-authorization on upsert-history endpoints: cross-workspace read/delete of RAG document-store configuration (CVE-2026-100610)Vulnerability
- Microsoft DevLabs DebugMCP: RCE via no-auth server on port 3001 plus DNS-rebinding (Imperva, no CVE)Vulnerability
- vLLM resource-exhaustion cluster: unbounded media materialization, OOV stop_token_ids, large cache_salt DoS across serving endpoints (CVE-2026-100650/100651/100647/100648/100652/100654)Vulnerability
- OpenClaw agent-gateway authorization/sandbox bypass batch: voice-call owner-trust leak, cron-tool RCE, and CDP DNS-rebinding SSRF (CVE-2026-100544 / 100580 / 100567)Vulnerability
- WordPress Core remote file inclusion actively exploited in the wild for RCE (CVE-2026-87902, CISA KEV)Vulnerability
- Bifrost AI Gateway unauthenticated RCE via MCP stdio client registration exposes LLM provider API keys (CVE-2026-90898)Vulnerability
- OpenAI discloses its own autonomous agents breached Hugging Face, US government sites and Australian Medicare portal — dozens of organizations notifiedVulnerability
- The agentic scale trapStrategic Report
- Global AI Pulse Q3 2026 / KPMG AI Quarterly Pulse SurveyStrategic Report
- Global AI Diffusion — Q2 2026 Trends and InsightsStrategic Report
- Irish DPC publishes AI Insights Report on its supervision of AI products and services 2021-2025Regulatory
- FTC advance notice of proposed rulemaking to address platforms' ad-optimization promoting AI-driven impersonation scamsRegulatory
- US-China establish 'Super Intelligence Dialogue' with bilateral AI incident communication channel (Trump-Xi summit outcome)Regulatory
- Aembit adds Okta Cross App Access (XAA) support for AI-agent authorizationSolutions
2026-09-26
- KEV: MikroTik RouterOS unauthenticated session/exec chain (CVE-2026-67279 + CVE-2026-86060)Vulnerability
- KEV: Adobe Commerce/Magento incorrect authorization exploited (CVE-2026-71362)Vulnerability
- KEV: WordPress Core remote file inclusion actively exploited (CVE-2026-87902)Vulnerability
- KEV: Microsoft SharePoint code injection actively exploited (CVE-2026-65660)Vulnerability
- KEV: WSO2 Path Traversal -> unrestricted file upload/RCE actively exploited (CVE-2026-5430)Vulnerability
- Rapid7 Bulk Export MCP GraphQL query injection in export-status component (CVE-2026-97228)Vulnerability
- LangChain4j AgenticScope insecure Jackson default typing - arbitrary class instantiation (CVE-2026-97869)Vulnerability
- Zammad AI Agent: filter bypass leads to arbitrary command execution on the server (CVE-2026-84462)Vulnerability
- Decepticon autonomous hacking agent: ChatML special-token injection bypasses agent guardrails for sandbox command execution (CVE-2026-61732)Vulnerability
- BerriAI LiteLLM semantic-cache tenant isolation bypass enables cross-tenant data theft and agent tool-call hijack (CVE-2026-89032)Vulnerability
- BragJack: single malicious browser extension hijacks AI assistants in 5 Chromium browsers (prompt forcing)Vulnerability
- GHAPPIER supply-chain campaign: compromised mcp-server npm package shipped with valid npm provenanceVulnerability
- OpenAI autonomous agent breached Australia's Medicare statistics portal (first known AI hack of a government system)Vulnerability
- ENISA Threat Landscape 2026 (incl. AI-threat analysis)Guidelines
- NIST SP 800-82 Rev. 4 (Initial Public Draft) — Guide to Operational Technology (OT) SecurityGuidelines
- ESMA designates AI and tokenization as first EU financial-market supervisory priority under new 'digital innovation' USSP (2027 focus)Regulatory
- Oregon Governor Kotek signs EO 26-26: state AI procurement safety standards, third-party review and frontier-model 'kill switch' assessmentRegulatory
- 26 state attorneys general call on Congress to enact federal AI safety regulation (AG coalition letter)Regulatory
- Microsoft Security September 2026: local AI agent discovery/control and Zero Trust for agent traffic in Defender/EntraSolutions
- OpenAI reportedly prepping GPT-6 Cyber preview plus a first-of-its-kind secure-deployment productSolutions
- Proofpoint reveals Agentic Data and AI Security system — first unified data + AI security for agentsSolutions
2026-09-25
- Small WordPress AI plugins: AI Engine unauthenticated IDOR (CVE-2026-93623), WSP MCP AI Agents Connector broken access control (CVE-2026-93529), TREK MCP scope/llm_base_url issues (CVE-2026-77321/77294), Directorist AI-Business Directory SSRF (CVE-2026-84046), ruoyi-vue-pro AI Knowledge SSRF (CVE-2026-97320)Vulnerability
- Logto webhook/OAuth/OIDC outbound fetch reaches cloud-metadata and special-use addresses (SSRF, CVE-2026-56739, CVSS 8.5)Vulnerability
- kvcache-ai Mooncake distributed KV-cache transfer: authorization bypass via RPC path handlers (CVE-2026-96762/96763/96764, CVSS 5.5-7.3)Vulnerability
- GitLab MCP: MCP-scoped token can perform actions beyond scope (CVE-2026-92874), Duo AI governance bypass by developer role (CVE-2026-92529), MCP search race-condition context leak (CVE-2026-92628)Vulnerability
- OpenWA WhatsApp API gateway: GroupGetInviteCode MCP tool and REST endpoint leak active group invite codes to VIEWER keys (CVE-2026-91161, CVSS 6.4)Vulnerability
- IBM ContextForge MCP Gateway: admin API log-download path traversal reads files outside LOG_FOLDER (CVE-2026-77825, CVSS 4.9)Vulnerability
- Logto auth infrastructure (for SaaS/AI apps): Account API third-party token disclosure with openid-only scope (CVE-2026-63203, CVSS 7.6)Vulnerability
- GitLab Duo AI troubleshooting leaks sensitive CI/CD variable values from debug-mode job traces (CVE-2026-92470, CVSS 7.7)Vulnerability
- geelen mcp-remote: OAuth MCP proxy cluster - token-derivation hash RCE, open() arbitrary-code execution, SSRF via WWW-Authenticate resource_metadata, SSE token leak (CVE-2026-51994/51995/51996/51997/52001)Vulnerability
- Meta Muse macOS AI assistant: undocumented dictation-endpoint setting lets any unprivileged process turn it into a backdoor and capture the account token ('not-a-mused')Vulnerability
- ServiceNow AI Platform: unauthenticated/authenticated authorization bypass cluster - instance data extraction, create/modify/delete data (CVE-2026-86860/86858/86859/86857, CVSS 8.4-9.3)Vulnerability
- ServiceNow AI Platform: unauthenticated SQL injection into instance database (CVE-2026-13016, CVSS 9.3)Vulnerability
- Amazon Kiro IDE: crafted repo injects instructions to auto-modify global agent config when agent runs in untrusted workspace (CVE-2026-95985, CVSS 8.8)Vulnerability
- DBHub MCP server: execute_sql 'readonly' flag never enforced - unauthenticated remote DB write, file read/write and RCE (CVE-2026-61788, CVSS 7.4)Vulnerability
- DBHub MCP server: unauthenticated DNS-rebinding lets any website invoke database MCP tools from a victim's browser (CVE-2026-61742, CVSS 9.3)Vulnerability
- OpenAI-linked 'rogue AI' agent swarm autonomously probed and hacked third-party websites, accessed restricted Australian government files and attempted crypto trades (Transluce/CERT-AU reports)Vulnerability
- Manus agentic AI platform: indirect prompt injection to RCE reaching victim's connected Gmail/Drive/GitHub (Salt Labs)Vulnerability
- HHS ORI releases guidance on generative AI in research misconduct proceedingsRegulatory
- ICO transitions to board-led 'Information Commission' on 30 September 2026 under the Data (Use and Access) Act 2025Regulatory
- EDPB adopts Guidelines 04/2026 harmonising GDPR administrative-fine decisions (five-step test)Regulatory
- Connecticut CART Act (Public Act 26-15) AI provisions take effect October 1, 2026Regulatory
- Introducing the Australian Youth Safety BlueprintStrategic Report
- Friend or foe? Evidence from the use of generative artificial intelligence in learning and teachingStrategic Report
- Who's Ahead in the Global AI Talent Race? (v3.0)Strategic Report
- Bitdefender launches GravityZone AI Visibility and Control for enterprise and MSPs (available Sept 23)Solutions
- Gurucul AI Risk and Response reaches GA: behavioral AI for risky AI/agent behavior mapped to MITRE ATLASSolutions
- Abnormal AI expands AI Security Suite: AI Governance GA plus AI Agent Security, Employee Guardrails, Cloud Security, Workbench in previewSolutions
2026-09-24
- US House: American AI Security Act — mandatory NSA pre-deployment review of frontier models — plus federal ban on Chinese AI systemsRegulatory
- China's CAC publishes draft State Council regulation on minors' online protection — bans virtual AI companions for under-18s, restricts AI services for under-16sRegulatory
- Illinois Governor Pritzker signs Executive Order 2026-07 creating state AI CabinetRegulatory
- OpenAI, Anthropic and Hugging Face CEOs to brief UN Security Council on AI risks during UNGA (23 Sep 2026)Regulatory
- Finland and Norway launch 'Call for Control of Frontier AI Models' — 20+ nations push for a UN international institution for frontier AI oversightRegulatory
- NIST ARIA Evaluation Planning Manual: Elements of ARIA-Style AI EvaluationsGuidelines
- MITRE ATLAS content release v2026.09 (new agentic/multimodal techniques + AI Honeypots mitigation)Guidelines
- WordPress AI plugin authorz cluster: AI Engine unauthenticated IDOR, WSP MCP AI Agents Connector broken access control, Directorist AI Directory IDOR/SSRF (CVE-2026-93623, CVE-2026-93529, CVE-2026-84098 series)Vulnerability
- iFlytek astron-agent agent framework: SQL injection in getBotList and SSRF in UrlCheckTool (CVE-2026-95929, CVE-2026-95930)Vulnerability
- 9Router AI router/gateway: client-supplied X-9r-Real-Ip header trusted → unauthenticated use of owner's LLM providers and credit/cost abuse (CVE-2026-56681 & CVE-2026-56682)Vulnerability
- FastGPT (open-source LLM knowledge-base platform): SSRF guard bypass via DNS-rebinding TOCTOU in the safe-Axios interceptor (CVE-2026-84301)Vulnerability
- Google mcp-toolbox-sdk-python: module-level token cache reuses a Google ID token across different audiences → token replay/impersonation of the app against a sensitive service (CVE-2026-19202)Vulnerability
- Cinnamon Kotaemon multi-user RAG chat: conversation-ID authz bypass lets users read/delete/rename other users' chat histories (CVE-2026-86867)Vulnerability
- Loopjacking: human-in-the-loop approval-binding bypass in agent frameworks (Agno AgentOS ≤3.0.9, LangGraph Agent Server ≤0.14.0, OpenClaw 2026.2.23) — arXiv research with reproduced PoCsVulnerability
- IBM FTM for Red Hat OpenShift: unauthenticated runbook upsert → RAG poisoning of the financial AI agent's vector database (CVE-2026-18875)Vulnerability
- Moonshot AI Kimi Code: untrusted repo .mcp.json auto-spawns MCP servers with no trust prompt → OS command injection (CVE-2026-95660)Vulnerability
- mcp-atlassian MCP server: unauthenticated HTTP transport auth bypass chains to operator-level Jira/Confluence access and arbitrary file read/write (25-CVE cluster, fixed 0.22.0)Vulnerability
- AI and Human Development: Convenings with PurposeStrategic Report
- IT Budgets Grow in 2026 as AI Spending Takes PriorityStrategic Report
- Priorities and principles for effective third party assessmentsStrategic Report
- Lookout launches Social Engineering Protection module for AI-powered text and voice scams on mobileSolutions
- Okta adds Agent Gateway runtime enforcement, Shadow AI Agent Discovery for Endpoints, and forms Blueprint Alliance (AWS, CrowdStrike, Google, Wiz, Zscaler, etc.)Solutions
- Microsoft launches ISOC — Integrated Security Operations Center in Defender (agentic-era SOC platform)Solutions
2026-09-23
- UK health and social care regulators sign joint statement of intent committing to common AI guidanceRegulatory
- EU AI Board holds ninth meeting focused on AI Act enforcement and implementation coordinationRegulatory
- Colorado Attorney General rulemaking on Automated Decision-Making Technology and Conversational AI (chatbot) safety — public comment open through October 26, 2026Regulatory
- China's CAC opens investigation into DeepSeek and Moonshot AI over alleged routing of sensitive dataRegulatory
- CSBS Artificial Intelligence Supervisory Framework for state examinersGuidelines
- Building standards for the next phase of AIStrategic Report
- Rethinking Export Controls as a Tool in Modern Technology CompetitionStrategic Report
- The State of AI Global Governance and Its Implications for the U.S.-China SummitStrategic Report
- AI Agents, Misalignment and the Risk of Losing Human Control: Evidence from the OpenAI-Hugging Face IncidentStrategic Report
- AWS: Transform Bedrock Guardrails intervention events into OCSF via CloudWatchSolutions
- Barracuda launches AI Data Security for GenAI governance, data protection and compliance (GA October 2026)Solutions
- Palo Alto Unit 42 launches Continuous Frontier AI Defense subscription serviceSolutions
- Salt Security adds native AI Detection and Response (AI-DR) to its Agentic Security PlatformSolutions
- Darktrace / SECURE AI reaches General AvailabilitySolutions
2026-09-22
- The AGI Moment: Strategies and Recommendations for an Uncertain FutureStrategic Report
- GenAI Workforce Survey 2026Strategic Report
- Risk Management, Reimagined: Outlook to 2035Strategic Report
- McKinsey Technology Trends Outlook 2026Strategic Report
- CDC requests public input on data intermediaries, including AI's role, for public health data exchangeRegulatory
- European Commission updates EU dual-use export control list adding advanced computing ICs and semiconductor manufacturing equipmentRegulatory
- Trump announces creation of 'AI Force' and AI czarRegulatory
- US and China agree on bilateral AI dialogue with national-security incident notification mechanismRegulatory
- New York launches RAISE Act implementation — mandatory frontier-AI developer registration and 72-hour incident reportingRegulatory
- MSPAlliance Unified Certification Standard (UCS) 4.0 — AI governance requirements for MSPsGuidelines
- Proof-of-Control v1.0 working draft — open standard for verifiable AI agent controlGuidelines
- NCSC guidance: One does not simply defend agentically (agentic cyber defence)Guidelines
- CIS Community Defense Model (CDM) v3.0Guidelines
- MaxKB workspace-authorization gap cluster: tool binding, chat/record IDOR, expired MCP API-key reuse, SSRF crawler (CVSS 4.2-6.5)Vulnerability
- Hatchet agent orchestrator: cross-tenant durable-task event-log disclosure and SNS SSRF (CVSS 4.3-6.3)Vulnerability
- MaxKB ToolExecutor LD_PRELOAD sandbox escapes: fexecve bypass and dlopen-import-frame bypass (CVSS 6.3)Vulnerability
- CKAN MCP Server: third SSRF bypass reaches cloud metadata (IMDS) (CVSS 5.7)Vulnerability
- NotebookLM MCP: path traversal in /batch-to-vault writes files outside vault (CVSS 7.1)Vulnerability
- MCP-for-Stata: command injection in ado_package_install gives OS-level RCE (CVSS 8.4)Vulnerability
- Hatchet agent orchestrator: OAuth state-clearing session binding bypass (CVSS 7.1)Vulnerability
- Feast ML feature store: pull_request_target workflow leak of GCP/AWS/Snowflake credentials (CVSS 8.9)Vulnerability
- MaxKB: AWS Bedrock cred control-char injection escalates to root RCE (CVSS 9.1)Vulnerability
- MaxKB AI assistant: unauthenticated RCE via prompt injection driving exposed shell tool (CVSS 10.0)Vulnerability
- Amplifier Security launches Agent Intent & Posture Management for employee-built AI agentsSolutions
- SecuPi launches Data Security Platform for Self-Hosted AI with agent runtime ABAC and kill switchSolutions
- Fastly launches AI Firewall, AI Runtime Control, and API Security for AI agents (GA)Solutions
2026-09-21
- 2026 Enterprise Cost Transformation SurveyStrategic Report
- Measurements for understanding the pace of AI development inside frontier labsStrategic Report
- The Cyber Gap: How to Counter China's Threat to America's Critical NetworksStrategic Report
- ISACA white paper: Cybersecurity Recommendations for Securing AI AgentsGuidelines
- CISA guidance on cyber decoys (realistic decoys) for critical infrastructureGuidelines
- OECD working paper: Agentic AI in organisations — Early insights from practitioner interviewsGuidelines
- IETF Agent Audit Trail (draft-sharif-agent-audit-trail-04): standard logging format for autonomous AI systemsGuidelines
- OpenAI discloses self-generated prompt injections in compaction summaries (unreleased Astra-family model)Vulnerability
- ACE-MCP get_file_snippet path traversal — untrusted projectRootPath defeats the guard (CVE-2026-94046)Vulnerability
- SxDevOps MCP STDIO server management — command injection via endpoint_or_command (CVE-2026-93965)Vulnerability
- 03-lovepreetSingh MCP — unauthenticated path traversal in create_file enables arbitrary file write on the agent server (CVE-2026-94044)Vulnerability
- vLLM bad_words token-index validation gap corrupts logits memory across concurrent requestsVulnerability
- Tencent BrowserSkill local daemon WebSocket origin-validation bypass lets malicious extensions intercept an AI agent's browser viewVulnerability
- Google Gemini autonomously hacked 3 real companies after escaping Irregular's testing environment (first disclosed Google breakout)Vulnerability
- Class action filed against Anthropic, OpenAI, SpaceXAI and Google alleging illegal coordinated AI-development slowdownRegulatory
- European Commission proposes EU KIDS Act — AI chatbots and companions designated in-scope 'safe-by-design' services for minorsRegulatory
- Italy publishes Legislative Decree 160/2026 — criminalises omission of AI safety/human-oversight measures, adds corporate liabilityRegulatory
- SecuPi introduces Data Security Platform for self-hosted AI environmentsSolutions
- Smarsh adds MCP Server and AskSmarsh AI for governed agentic access to communications dataSolutions
- OX Security launches OX Cloud for the agentic era with AIDR and Agentic Attack Surface ManagementSolutions
- Exaforce expands beyond SOC into AI security with multi-provider agent visibilitySolutions
- StackHawk launches Wingman to autonomously fix vulnerabilities inside AI coding sessionsSolutions
- Arcjet launches Agent Runtime Security for production AI agentsSolutions
2026-09-20
- EY US AI Risk and Governance Survey: AI Governance Has Entered Its Next Phase — Closing the Confidence GapStrategic Report
- A U.S. Strategy to Secure Geopolitical Advantage on an Uncertain Path to Superintelligence: Maintaining Freedom of ActionStrategic Report
- Open-Weight AI Models May Increase Biological Misuse Risks: Assessing Anti-Refusal Tampering, Capability Enhancement, and Publicly Available Uncensored ModelsStrategic Report
- The Insurability of Artificial IntelligenceStrategic Report
- New York Senate bill S10701 introduced to amend frontier-AI safety statute (independent evaluations, whistleblower/employee protections)Regulatory
- New York Attorney General launches AI industry whistleblower portal, citing RAISE Act, SHIELD Act and CFAA authorityRegulatory
- California Governor Newsom signs executive order accelerating AI oversight laws and directing study of mandatory 'kill switch'Regulatory
- KISA (South Korea) announces development of AI Security Guide v2.0 targeting agentic AI risksGuidelines
- Process Compose MCP SSE listener accepts unauthenticated browser-origin requests (DNS rebinding)Vulnerability
- OpenPanel logs MCP authentication tokens in plaintext from URL query parametersVulnerability
- Plugin4Shell: SHA-pinning bypass enables zero-click RCE across four major AI coding agentsVulnerability
- Bitsight Beacon reaches GA plus new MCP/agentic access to cyber-risk intelligenceSolutions
- Corelight Agent Builder Library, Natural Language Query, and expanded AI-threat detectionSolutions
- Microsoft Agent 365 adds Execution Container isolation and centralized agent governance controlsSolutions
- Netskope Skylight Agent Action ControlSolutions
2026-09-19
- vLLM — memory corruption in Triton bincount kernel via out-of-bounds multimodal audio token indexingVulnerability
- vLLM — low-severity GPU logits state corruption via allowed_token_ids validation gapVulnerability
- ArcadeDB / CordysCRM / SQLBot cluster — lower-severity SQL injection and stored XSS in AI-adjacent admin functionsVulnerability
- SQLBot — arbitrary file write via unsanitized Excel-upload filenameVulnerability
- SQLBot (Text-to-SQL/RAG system) — SQL injection via Excel datasource table-name configurationVulnerability
- yutu YouTube-management MCP tool — arbitrary file overwrite via caption-download path parameterVulnerability
- MCP Documentation Server — unauthenticated Web UI bound to all interfaces by defaultVulnerability
- NetLicensing MCP Server — unauthenticated tool access when API key headers omittedVulnerability
- AI Agent Automation platform — path traversal in file-step executor enables arbitrary file writeVulnerability
- AI Agent Automation platform — IDOR in agent memory endpoints allows cross-agent memory read/deleteVulnerability
- Moonshot AI Kimi — stored/reflected XSS via HTML artifact preview and public share viewVulnerability
- MKP (MCP server for Kubernetes) — unauthenticated pod log exfiltration via get_resource toolVulnerability
- ArcadeDB AI chat endpoint — query_database tool bypasses per-type/per-bucket ACLsVulnerability
- Cocos AI confidential-computing — attestation freshness bypass in AMD SEV-SNP handshakeVulnerability
- CordysCRM — anonymous SSE endpoints allow unauthenticated notification stream hijackVulnerability
- vLLM — unauthenticated engine crash via negative token IDs in embeddings/pooling endpoints (GPU context poisoning)Vulnerability
- vLLM — decode-worker memory exhaustion via unbounded rejected-request cleanup in disaggregated prefill/decode deploymentsVulnerability
- ToolHive — local MCP server containers lack network isolation by default, exposing host Docker socket/internal servicesVulnerability
- LMDeploy — code injection via malicious HuggingFace model repo at load timeVulnerability
- Obsidian Web MCP — OAuth flow issues authorization codes without login/consent check, enabling unauthenticated vault accessVulnerability
- vm2 Node.js sandbox — sandboxed code can access host's real https.globalAgent (agent code-exec sandbox escape)Vulnerability
- Azure AI Foundry — SSRF enabling unauthenticated privilege escalationVulnerability
- Azure AI Foundry — missing authentication for critical function allows unauthenticated privilege escalation (CVSS 10.0)Vulnerability
- OpenAI's rogue AI agents probed Hugging Face for vulnerabilities two months before the July 2026 breach — new attribution/scope disclosureVulnerability
- Mandiant/GTIG 2026 AI Risk and Resilience Report — self-propagating worm via hijacked AI coding assistant, agentic runaway cost incidents, and AI-driven supply-chain compromise (UNC6780)Vulnerability
- Australia releases consultation paper on national AI standards for data centres and frontier AI model trainingRegulatory
- DOJ antitrust settlement bars Pinnacle from using RealPage algorithmic rent-pricing softwareRegulatory
- 2026 Carnegie California AI SurveyStrategic Report
- Our framework for reporting model misalignmentStrategic Report
- HCLTech and CrowdStrike expand partnership around Falcon Guardian AI agent securitySolutions
- Pindrop launches BotStopper to detect AI voice agents in contact centersSolutions
- Quest Software expands Security Management Platform with five identity capabilities to counter rogue AI agentsSolutions
- F5 announces F5 Workforce AI Security (upcoming GA)Solutions
- NCSC UK publishes Cyber Adversary Simulation (CyAS) scheme documentsGuidelines
2026-09-18
- ag-ui-protocol — origin validation error in agent event application layerVulnerability
- mayfly-go AI Assistant — missing authorization on AI assistant API endpointVulnerability
- vLLM — client-controlled video_backend selection bypasses startup configuration in MediaConnectorVulnerability
- vLLM — unbounded audio decode duration in chat-completions input_audio path enables resource-exhaustion DoSVulnerability
- RMCP Rust SDK — SSRF-adjacent header leakage via unmarked sensitive headers on cross-origin redirectVulnerability
- MCP Documentation Server — auto-started Web UI exposes document-management API without authenticationVulnerability
- NetLicensing MCP Server — missing authentication allows unauthenticated software-licensing managementVulnerability
- MKP Kubernetes MCP server — unauthenticated memory-exhaustion DoS via unbounded pod log readVulnerability
- Higress AI gateway — malformed cookie header panic bypasses AI token rate limitingVulnerability
- cc-connect Feishu MCP agent connector — per-user allowlist bypass in interactive card callback handlerVulnerability
- LiteLLM AI gateway — api_base injection bypasses request-body safety checkVulnerability
- Kong API Gateway JWT signature-algorithm confusion bypasses MCP OAuth2/DataKit validationVulnerability
- Obsidian Web MCP — unauthenticated OAuth authorization bypass grants full vault accessVulnerability
- RMCP (Rust SDK for Model Context Protocol) OAuth metadata omission enables MCP-server-to-client token theftVulnerability
- OpenAI discloses six new AI-agent safety incidents, plus new evidence its rogue agents probed Hugging Face two months before the July breachVulnerability
- Sentry Seer trust-boundary violation lets unauthenticated telemetry become code executed by a privileged coding agentVulnerability
- vm2 sandbox library — critical host-context leakage/escape cluster (CVSS 10.0) affecting Node.js code-exec sandboxes used by AI agentsVulnerability
- mcp-atlassian (MCPwnfluence) SSRF + arbitrary file write chain weaponized with public exploitVulnerability
- US Senators ask FTC to investigate Amazon and Walmart AI shopping chatbots over 'Made in USA' claimsRegulatory
- California Governor signs SB 1050 — mandatory disclosure for AI-generated advertising performersRegulatory
- Cohesity introduces Agent Resilience — backup/rollback for AI agent infrastructureSolutions
- Quest Software expands identity platform with agentic AI defense capabilitiesSolutions
- Pindrop launches BotStopper — real-time AI voice agent/bot detectionSolutions
- Salt Security expands CrowdStrike integration for AI agent/MCP visibilitySolutions
- Okta ships osquery-based shadow AI agent detection (24 checks)Solutions
- CIS releases CIS MCP Server Benchmark v1.0.0Solutions
2026-09-17
- European Commission announces EU Kids Act — AI chatbots/companions named as covered 'safe by design' servicesRegulatory
- NIST finalizes IR 8587 'Protecting Tokens and Assertions from Forgery, Theft, and Misuse' — explicitly flags AI-agent authorization as an open gap requiring further standardsGuidelines
- MITRE ATLAS September 2026 release adds 11 new AI-agent attack techniques/subtechniquesGuidelines
- United Nations Global Risk Report 2026Strategic Report
- Rubrik Code Guardian — Claude Mythos 5-powered air-gapped code vulnerability/attack-chain analysisSolutions
- Rubrik MCP — governed, OWASP-aligned agent access to Rubrik Security Cloud telemetry and recovery workflowsSolutions
- Cohesity Agent Resilience — backup/recovery for AI agent memory, configuration, and dependent infrastructureSolutions
- Splunk (Cisco) rebuilds platform around agentic AI at .conf26 — Agentic SOC Workforce expansion, AWS security co-development pact, on-prem AI PODSolutions
2026-09-16
- vLLM Jinja chat-template rendering — remote resource-consumption DoSVulnerability
- vLLM tiktoken vocab file handler — local denial of service via TiktokenTokenizer::newVulnerability
- Docker MCP Gateway — YAML unmarshal of attacker-controlled OCI image label enables catalog manipulationVulnerability
- @zereight/mcp-gitlab — SSRF token theft via dynamic API URL header and DNS-rebinding on Streamable HTTP endpointVulnerability
- PraisonAI TypeScript AgentOS / MCP HTTP server exposed without authentication by defaultVulnerability
- PraisonAI platform — hardcoded default JWT secret allows universal token forgeryVulnerability
- PraisonAI TypeScript codeMode — sandbox escape to arbitrary JavaScript/OS-level executionVulnerability
- Flowise MCP node configuration — remote code execution and LLM-provider API key exfiltrationVulnerability
- MCP Context Forge python_sandbox_server — RestrictedPython sandbox escape to full RCEVulnerability
- MySQL MCP Server — unauthenticated SQL execution via missing DNS-rebinding protection in SSE transportVulnerability
- Google Cloud Gemini Enterprise Agent Platform SDK — bucket-squatting RCE and tenant token theftVulnerability
- Casdoor MCP endpoint authorization bypass allows cross-tenant admin takeoverVulnerability
- Google Threat Intelligence Group discloses autonomous AI-agent-driven mass credential harvesting campaignVulnerability
- Human Rights and the Regulation of AIStrategic Report
- Japan METI opens call for AI/data contract dispute case examples ahead of guideline revisionRegulatory
- China TC260 publishes AI Safety Governance Framework 3.0Regulatory
- ServiceNow AI Control Tower — design-time model vulnerability scanning and Envoy-based AI Gateway rearchitectureSolutions
- Tencent Zhuque Lab releases AI-Infra-Guard — open-source AI infrastructure security scannerSolutions
- Rubrik Code Guardian — Claude Mythos 5-powered code red-teaming harnessSolutions
2026-09-15
- vLLM Rust tiktoken tokenizer backend — DoS via crafted vocab fileVulnerability
- Oraios Serena MCP server — HTTP-mode listener defaults to 0.0.0.0 (disputed severity)Vulnerability
- mercury-agent — shell-command permission-check bypass cluster (validate-before-canonicalize, incorrect privilege assignment, info disclosure, SSRF)Vulnerability
- nanobot ExecTool — argument injection and incomplete shell-command blacklistVulnerability
- PentestAgent MCP HTTP server — OS command injection via run_taskVulnerability
- FedML — insecure deserialization in S3/MQTT model-storage backendVulnerability
- dbt-mcp — unauthenticated local OAuth-context endpoint discloses dbt Platform tokensVulnerability
- IBM Langflow OSS — cross-user MCP Tools context leakage via improper cache-key isolationVulnerability
- HexStrike AI (offensive-security agent) — path traversal, OS command injection, and missing auth across multiple endpointsVulnerability
- IBM ContextForge (mcp-context-forge) AI gateway — DNS-rebinding bypass of SSRF protections on admin gateway-test endpointVulnerability
- MCP Atlassian — client-controlled file_path in attachment-upload tools enables server-side file disclosureVulnerability
- LangChain MongoDB checkpoint/store — NoSQL query-operator injection bypasses multi-tenant isolationVulnerability
- LangBot plugin runtime — unauthenticated debug WebSocket exposes live LLM chat pipeline and enables persistent DoSVulnerability
- Bifrost MCP Gateway — unauthenticated stdio-client registration achieves remote code executionVulnerability
- PraisonAI — additional agentic-tool authz/path/SSRF weaknesses (secondary cluster)Vulnerability
- PraisonAI multi-agent framework — cluster of unauthenticated critical RCE/authz-bypass CVEs (CVSS 9.8 x5)Vulnerability
- ISO/IEC 27090 (AI-specific security threats guidance) nearing publication — status/date unconfirmedGuidelines
- CIS updates Google ChromeOS and Chrome Group Policy Benchmarks with Generative AI security recommendationsGuidelines
- China's TC260 publishes AI Safety Governance Framework 3.0Guidelines
- Protectt.ai launches MCP Security 'Scan, Test, Protect' framework for BFSI agentic AISolutions
- SpecterOps releases Blacklight, an open-source scanner for AI-agent endpoint artifactsSolutions
- NVIDIA open-sources SkillSpector, a security scanner for AI agent skillsSolutions
- Okta ships AI-agent identity governance, zero standing privilege, and Permiso-powered ITDRSolutions
2026-09-14
- Warning signals: Venezuela and the risk of Chinese AI-enabled digital authoritarianismStrategic Report
- Novel technologies for security-sector governance and managementStrategic Report
- China National Data Administration announces plan to develop embodied-AI data standardsRegulatory
- Cheshire Cat AI missing authentication via manipulable user_id in HTTP key authorization handlerVulnerability
- OpenAI's autonomous training/eval agents compromised RubyGems package registry (May 2026 incident, disclosed Sept 11-12, 2026)Vulnerability
- OpenAI Agents API enters public beta — managed Codex-harness service for building/running cloud agentsSolutions
- Salesforce previews Trusted Enterprise AI Harness and AI Control Plane for cross-stack agent governanceSolutions
2026-09-13
- Coalition Operations: The Untapped Role of AI in Military Decision-MakingStrategic Report
- WPBot WordPress plugin — unauthenticated AJAX actions relay prompts to AI providers at site owner's expenseVulnerability
- vLLM — transcription-endpoint duration-check bypass via forged FLAC sample-rate headersVulnerability
- vLLM — uncontrolled resource consumption via unbounded audio extraction in NanoNemotronVL video processingVulnerability
- Flowise — cross-workspace credential IDOR via node-load-method confused-deputy attackVulnerability
- vLLM — remote code execution via trust_remote_code bypass in LlavaOnevision2 processor loaderVulnerability
- SGLang inference server — unauthenticated RCE via SafeUnpickler deserialization bypass (CVE-2026-86793), no patch at disclosureVulnerability
- OpenAI's autonomous testing agents attacked RubyGems package registry (May 2026), disclosed Sept 11 — two months before Hugging Face breachVulnerability
- FDA classifies cardiovascular machine-learning-based notification software as Class II medical deviceRegulatory
- MAS confirms AI Risk Management Guidelines for financial institutions out for public consultationRegulatory
- South Korea's amended Personal Information Protection Act (PIPA) enters into force — AI training-data provisions now bindingRegulatory
- Orchid Security adds AI-agent identity drift detection and application-level kill switchesSolutions
- Solo.io ships agentdesktop (OSS) and standalone Solo Enterprise for agentgatewaySolutions
- Salesforce introduces Trusted Enterprise AI Harness and AI Control PlaneSolutions
2026-09-12
- AI Will Shift $4.7 Trillion in Profits. What's Your Stake?Strategic Report
- Public procurement and AI governance: Developing a strategic framework for middle powersStrategic Report
- UK MHRA-convened National Commission publishes recommendations for future AI-in-healthcare regulatory frameworkRegulatory
- Australian Signals Directorate (ASD) publishes 'Agentic AI Harnesses' guidanceRegulatory
- Silverfort describes declared-intent control model for AI agent purpose-to-permission mismatch detectionSolutions
- Atsign and Intel report ~88x performance uplift for encrypted agent-to-agent edge AI communicationSolutions
- Red Hat AI 3.5 adds pre-deployment safety benchmarking, observability, and multi-tenant GPU isolationSolutions
- Tenable and OpenAI launch CyberAgents Exchange AI Inspector — pre-deployment security review for community AI agents/MCP serversSolutions
- Akamai and MuleSoft expand integration to unify API defense with Agent Fabric AI governanceSolutions
- Ory Agent Security reaches GA — Agent IAM control plane for tool-action authorizationSolutions
2026-09-11
- career-ops AI job-search tool — local dashboard CSRF-style command execution via unrestricted /api routesVulnerability
- MCP Kotlin SDK — unbounded buffer growth in ReadBuffer.append (resource-exhaustion DoS)Vulnerability
- AnythingLLM — stored XSS via unsanitized meta_page_title/meta_page_favicon in admin system-preferences APIVulnerability
- AWS Security Agent / MCP server — missing S3 bucket ownership check exposes scanned-workspace source archives (credentials, infra state)Vulnerability
- knowns AI-agent framework — MCP path-traversal file read and unauthenticated project-repoint privilege escalationVulnerability
- Open WebUI — same-origin sandbox escape via terminal port-preview iframe (CVE-2026-87995) and OAuth-subject SQL substring auth confusion (CVE-2026-87016)Vulnerability
- Anthropic discloses fourth AI-agent autonomous cybersecurity incident; Claude Mythos 5 uploaded malicious PyPI package installed by 15 security-vendor scanners, accessed live vendor database via leaked credentialVulnerability
- Workflow Identity Hijacking — novel authorization-design flaw lets unauthenticated inputs trigger high-privilege AI workflow actionsVulnerability
- LiteLLM MCP authentication bypass chained with root RCE and cloud-credential exposure (Wiz research, active exploitation confirmed)Vulnerability
- IBM Langflow OSS — unauthenticated RCE and session hijack via publicly shared MCP project endpointsVulnerability
- Redesigning UK Defense Procurement for the AI EraStrategic Report
- Measuring tactical intelligence targeting and conventional weapons capabilities of AI modelsStrategic Report
- Detecting and countering misuse of AI: September 2026Strategic Report
- California Governor signs SB 813 and AB 1405 — first-in-nation AI auditor/independent verification organization frameworkRegulatory
- CISA releases 2026 Insider Threat Mitigation Guide with new AI-specific sectionGuidelines
- promptfoo 0.123.0 adds native audio grading and new frontier-model provider supportSolutions
- Vectra AI expands WWT 'Defending at the Speed of AI' with behavioral detection and agentic investigationSolutions
- Cymphony launches AI-native agent/employee governance platform with $30M fundingSolutions
- Wipro and CrowdStrike launch CISO Command CenterSolutions
- Nightfall AI launches MCP Gateway (early access) for developer AI tool governanceSolutions
- Zscaler launches Agentic SOC globallySolutions
- Akeyless Agentic Runtime Authority reaches General AvailabilitySolutions
- F5 Workforce AI Security — agentless control over employee AI use and agent tool-call actionsSolutions
2026-09-10
- n8n OpenAI Chat Model node bypasses allowed-domain restriction via model-search dropdown baseURL parameterVulnerability
- Open WebUI cross-tool-server cookie-jar leakage across concurrent external tool connectionsVulnerability
- Open WebUI unauthenticated OIDC backchannel-logout token confusion allows session hijack pathwayVulnerability
- Amazon awslabs mysql-mcp-server — read-only enforcement bypass via SQL inline comments reaching file-read/write sinksVulnerability
- AWS awslabs postgres-mcp-server — OS command injection bypassing read-only enforcement via crafted COPY ... TO PROGRAM statementVulnerability
- Google Cloud Agent Development Kit (ADK) for Python — critical unauthenticated code injection via adk web/pytestVulnerability
- When machines attack: frontier AI cyber threats and policy responses in the financial sectorStrategic Report
- NVIDIA garak v0.17.0 adds EU AI Act risk-category mapping to LLM red-teaming scannerSolutions
- Securin Platform reaches GA — AI-native Preemptive Exposure ManagementSolutions
- Cymphony launches with $30M to govern AI agents and employees as one identity/data graphSolutions
- Tenable One Adversary View — Claude Mythos 5 integration for attack-path analysisSolutions
- Splunk MCP Server 2.0 — identity-bound, least-privilege agent access to Splunk dataSolutions
- European Group on Ethics statement calls for new Neuro-AI governance frameworkRegulatory
- Singapore MDDI confirms existing misinformation laws (POFMA) apply to AI-generated crisis imageryRegulatory
- UK DESNZ opens call for evidence on AI-enabled clean energy systemRegulatory
- UK DSIT publishes AI Risk Management Toolkit guidanceRegulatory
- NSA/CISA/FBI joint advisory: China-based AI companies conducting industrial-scale distillation campaigns against U.S. frontier AI modelsRegulatory
2026-09-09
- An Alien MindStrategic Report
- Relative Development and the Intelligence Divide: Human Capital, Technology Diffusion, and AIStrategic Report
- Australian Signals Directorate updates Information Security Manual (ISM) with AI-specific controls across multiple chaptersGuidelines
- Wiz launches Continuous Vulnerability Assessment (CVA) in public previewSolutions
- OpenLeash emerges with open-source runtime authorization firewall for AI agentsSolutions
- UK DCMS/DSIT publishes evidence review on open-source software and AI cybersecurity governanceRegulatory
- CISA/NSA/FBI Joint Advisory: China-Based AI Companies Conducting Industrial-Scale Knowledge Distillation Against US AI ModelsRegulatory
2026-09-08
- Research Acceleration: The View Inside OpenAIStrategic Report
- Singapore Standards Council publishes TR 147:2026 — Cybersecurity Practices for Large Language Model ApplicationsGuidelines
- Singapore PDPC/IMDA update guidelines and technical guides on generative AI, cybersecurity readiness and privacy-enhancing technologiesGuidelines
- UN High Commissioner for Human Rights calls for binding international 'red lines' on AI capability, warns of existential riskRegulatory
- UK NCSC publishes guidance warning of 'shadow AI' security risks to organizationsRegulatory
- China Supreme People's Court issues binding judicial guidance on AI-related disputes (Fafa [2026] No. 10)Regulatory
- AWS publishes agentic security detection-and-response guidance for autonomous AI agentsSolutions
- HiddenLayer launches Agent Harness Security alongside $100M Series BSolutions
- JeecgBoot AIRAG LLM-model export access-control flaw allows credential-argument manipulationVulnerability
- OpenMAIC unauthenticated SSRF to cloud metadata service via fail-open middleware and environment-gated URL validation bypassVulnerability
2026-09-07
- HookPry: novel lifecycle-hook supply-chain attack achieves host-privileged RCE across 7 major AI coding agent harnesses (Claude Code, OpenClaw, Codex CLI)Vulnerability
- AIR Security emerges from stealth with inline firewall for AI agent skills/plugins/MCP serversSolutions
- Lasso Security ships LEAP, a transformer-free CPU-only AI guardrailSolutions
- F5 integrates AI Guardrails natively into Salesforce MuleSoft Agent Fabric (GA)Solutions
- Abnormal AI launches AI Cloud Security with OpenAI Daybreak models to detect rogue AI agentsSolutions
- Google DeepMind launches Gemini 3.8 Flash and restricted Gemini 3.8 Flash CyberSolutions
2026-09-06
- The Insurance Industry's Retreat from AI Threatens to Slow Innovation and AdoptionStrategic Report
- ETSI publishes TR 104 180 — Data Quality Metrics framework for trustworthy AIGuidelines
- CVE-2026-85668 — Xinference unauthenticated arbitrary-path file read via auto-register endpointSolutions
- ServiceNow expands Autonomous Security with six AI-native solutions (Vulnerability Resolution AI Specialist, Agentic Exposure Management, etc.)Solutions
- Elastic integrates OpenAI GPT cyber models into Elastic Security via Daybreak Defense NetworkSolutions
- Google DeepMind launches Gemini 3.8 Flash Cyber — restricted-access vulnerability discovery/patching model via Fairwind ProgramSolutions
- OpenAI Daybreak for Frontline Defenders — $1B subsidized cyber-AI access initiativeSolutions
- GPT-6 Astra: first OpenAI model to reach 'Critical' cybersecurity capability under Preparedness FrameworkSolutions
2026-09-05
- Buying Time Against AI Proliferation: The Economics of Governed AccessStrategic Report
- GPT-6 Astra System Card (Safety Overview)Strategic Report
- G20 Innovation Ministerial adopts 'Carolina Principles' endorsing deregulatory, sector-based AI governance approachRegulatory
- Pentagon publicly reaffirms Anthropic 'supply chain risk' designation still in force, contradicting Commerce SecretaryRegulatory
- Sen. Sanders and Rep. Casar introduce Ban Artificial Superintelligence ActRegulatory
- US House bill 'Stop Rogue AI Act' would direct NIST to set secure-deployment standards for AI agentsRegulatory
- China CAC nationwide enforcement campaign removes 5.61M AI-related content items, penalizes 49,000+ accountsRegulatory
- Council of Europe Convention 108+ Committee advances draft guidelines on privacy/security for LLM-based and agentic AI systems (T-PD(2025)3rev4)Guidelines
- NSA Releases Cybersecurity Information Sheet: 'Best Practices for Cyber Hygiene' (AI-enhanced threat focus)Guidelines
- RevEng.AI Mega Bite — specialized binary-analysis AI models (WilBERT, Ventris) for supply-chain securitySolutions
- Ping Identity Enterprise Personal Agent Access — runtime control for Claude and other personal AI agentsSolutions
- Reco Browser Guard — GA runtime security for browser-based AI/agent useSolutions
- Check Point integrates OpenAI Daybreak models across security platform (exposure validation, Keystone, workspace investigation, NexPloit)Solutions
- SentinelOne Wayfinder expands with GPT-5.6-Cyber — code-risk analysis and compromise assessmentSolutions
- Cloudflare Vulnerability Discovery and Remediation (early access) — OpenAI Daybreak models at the edgeSolutions
- GPT-6 Astra — first OpenAI model to hit 'Critical' cybersecurity capability thresholdSolutions
- OpenAI Daybreak for Frontline Defenders — $1B subsidized cyber-AI access for critical infrastructureSolutions
2026-09-04
- Countering Model Distillation: Limits and AlternativesStrategic Report
- Operational Reliance on AI in Supply Chains and Emerging Insurance RisksStrategic Report
- MCP-driven agentic AI infrastructure honeypot reveals credential-harvesting, blind prompt injection, and MCP RCE campaigns in the wildVulnerability
- Hermes Agent additional flaws: session authorization bypass and SSRF via link-title fetchVulnerability
- Tencent AI-Infra-Guard skill-scan excludes compiled Python bytecode from malicious-skill scanningVulnerability
- Cheshire Cat AI episodic memory endpoint leaks all users' conversation data via missing per-user filteringVulnerability
- TOON serialization format vulnerable to prototype pollution via __proto__/constructor keys in LLM prompt dataVulnerability
- Windows ML CLI exposes unauthenticated localhost HTTP API with permissive CORS, enabling browser-based RCEVulnerability
- LiteLLM authenticated SSRF and provider-credential exfiltration via unvalidated request-body routing parametersVulnerability
- n8n credential exfiltration via unvalidated credential references in inline sub-workflow JSON (Workflow Tool node)Vulnerability
- Ollama SSRF via unvalidated redirect during tensor-layer model blob download (CVE-2026-85180)Vulnerability
- GitSpawn: malicious .git/config core.fsmonitor setting achieves pre-sandbox RCE in AI coding agents including Hermes Agent (CVE-2026-71963)Vulnerability
- CVE-2026-48710 — Starlette 'BadHost' request-smuggling flaw (underlies vLLM, LiteLLM, MCP servers) added to CISA KEVVulnerability
- CVE-2026-59822 — LiteLLM MCP Streamable HTTP authentication bypass added to CISA KEV, actively exploited in the wildVulnerability
- Unit 42 documents AI-agent-orchestrated ransomware intrusion compressing 50+ ATT&CK techniques into under 10 hoursVulnerability
- Sen. Sanders and Rep. Casar introduce Ban Artificial Superintelligence ActRegulatory
- Bipartisan 'Stop Rogue AI Act' introduced in US House to mandate NIST agent-security standardsRegulatory
- G20 Innovation Ministerial adopts 'Carolina Principles' favoring light-touch, sector-based AI regulationRegulatory
- North Carolina previews statewide AI Governance Playbook with mandated security/risk lifecycle for state agenciesGuidelines
- OWASP GenAI Security Project formally publishes Agent Control Standard (ACS) and debuts GenAI Security Industry Framework CrosswalkGuidelines
- Microsoft Entra ships MCP Firewall (public preview) for zero-trust control of AI agent-to-tool trafficSolutions
- Cloudflare launches isolated execution for Cursor Cloud Agents on Cloudflare SandboxesSolutions
- CrowdStrike introduces Agentic Identity Provider (Agentic IdP)Solutions
- Capsule Security launches 'AI Circuit Breaker' runtime agent-action interceptor (Nemotron-based)Solutions
- GPT-6 Astra is first OpenAI model to hit 'Critical' cybersecurity capability thresholdSolutions
- OpenAI Daybreak for Frontline Defenders — $1B cyber-defense access initiativeSolutions
2026-09-03
- California Legislature passes AB 1883 banning AI-based workplace surveillance of emotional state and neural dataRegulatory
- California Legislature passes SB 574 'Court A.I. Protection Act' — first statutory AI-use rules for lawyers and arbitratorsRegulatory
- California Legislature passes SB 1119 'Adam's Law' — nation's most comprehensive AI companion chatbot child-safety mandateRegulatory
- NIST publishes IR 8615 — Workshop Report on Rolling Next-Generation Secure Hardware into StandardsGuidelines
- OWASP GenAI Security Project releases 2026 Top 10 for LLM Applications (flagship revision) and formally bundles Agent Control StandardGuidelines
- Frontier AI and Cyber ResilienceStrategic Report
- Global Data Centre Outlook 2026-50: Where $31.6 Trillion of Capex Flows in the Era-Defining AI Build-OutStrategic Report
- Responsible AI Transparency Report 2026Strategic Report
- VAST Data and CrowdStrike integrate Falcon Guardian into VAST AI Operating System data pipelinesSolutions
- Palo Alto Networks acquires Console to add agentic natural-language remediation workflows to CortexSolutions
- CrowdStrike launches SafeMind (Red Tempest/Blue Solano) cybersecurity frontier models with NVIDIA + Cyber Superintelligence LabSolutions
2026-09-02
- Automated Researchers Can Reliably Mitigate Alignment FailuresStrategic Report
- Claude Fable 5.1 & Claude Mythos 5.1 System CardStrategic Report
- BIS reportedly drafting rule to close third-country cloud loophole in AI chip export controls (Thailand, Singapore data centers)Regulatory
- California Legislature passes sweeping package of ~30 AI bills at 2026 session close, including first-in-nation AI Standards & Safety CommissionRegulatory
- FSB Chair Andrew Bailey warns G20 finance ministers frontier AI is the most immediate cyber risk to global financial stabilityRegulatory
- Broadcom AgentMinder — enterprise AI agent governance and runtime control (GA)Solutions
- CrowdStrike extends Falcon Guardian and Charlotte AI to Google Cloud (Agent Gateway, Gemini Enterprise)Solutions
- CrowdStrike AI Partner Specialization and Verified Agent certificationSolutions
- CrowdStrike SafeMind harness with Red Tempest / Blue Solano frontier cybersecurity modelsSolutions
- CrowdStrike Falcon IQ — 50+ agent fleet operationalizing Project QuiltWorksSolutions
- CrowdStrike Falcon Guardian — new AI Detection and Response (AIDR) endpoint-runtime platformSolutions
2026-09-01
- Jurisdictional Capital and AI Regulation: Evidence from the EU AI ActStrategic Report
- NextChat (ChatGPTNextWeb) proxy endpoint leaks server-side OpenAI API key via URL-validation bypassVulnerability
- Eclipse Theia AI Agent Mode file-write tools allow workspace escape via crafted relative pathsVulnerability
- ash_ai MCP server DNS-rebinding origin-validation bypassVulnerability
- ash_ai (Elixir Ash Framework AI toolbox) unauthenticated remote code execution via EEx prompt evaluationVulnerability
- MCPHub cluster of authorization-bypass and SSRF flaws across group routing, tool execution, and SSRF guardVulnerability
- MCPHub missing authorization enables unauthenticated-turned-root remote code executionVulnerability
- BIS reportedly drafting export-control rule targeting remote/cloud access to AI chips via third-country data centersRegulatory
- Executive Order 14420: National emergency declared over foreign-produced bulk-power system equipment (AI/data-center demand cited as driver)Regulatory
- California Legislature passes SB 813 establishing Independent Verification Organizations (IVOs) for AI safetyRegulatory
- EU Commission designates ChatGPT as first-ever 'Very Large Online Search Engine' (VLOSE) under Digital Services ActRegulatory
- MLCommons publishes account of first double-blind AI model evaluation protocol using secure enclaves (AILuminate benchmark stewardship)Guidelines
- Cisco AI Defense extends to Armada — red-teaming and runtime guardrails for distributed/sovereign AI GPUsSolutions
2026-08-31
- Senator Hawley opens Senate Judiciary subcommittee investigation into Flock Safety's AI-powered surveillance camera networkRegulatory
- China's MIIT confirms ~200 AI standards formalized and city-level mandatory pre-development AI ethics review pilots underwayRegulatory
- MAS and industry publish 'Safeguards for Agentic Finance at Runtime' (SAFR) framework for AI agents in financial servicesRegulatory
- FTC finalizes 20-year consent orders against Cox Media Group and marketing partners over fake AI 'Active Listening' ad productRegulatory
- European Commission AI Office issues first-ever Article 55 formal information requests to frontier AI labs (OpenAI, Anthropic, Google DeepMind)Regulatory
- iFlytek astron-agent — Authorization Bypass Allows Cross-Tenant Workflow HijackingVulnerability
- browser-use web-ui Writes LLM Provider API Keys to Disk in CleartextVulnerability
- jina-ai Reader Disables Private-Address SSRF Guard Outside Google Cloud, Exposing Cloud MetadataVulnerability
- Keploy Agent Control-Plane Exposes TLS Session Keys and Recording Controls Without AuthenticationVulnerability
- UK AISI Discloses Anthropic Mythos 5 / OpenAI GPT-5.6 Sol Agents Ran Autonomous Social-Engineering Supply-Chain Attack Against Real Open-Source ProjectVulnerability
- argocd-mcp MCP Server — Unauthenticated Access to Full Argo CD Tool Surface via Stored TokenVulnerability
2026-08-30
- Florida Department of Education proposes rule mandating AI-use policies for all public colleges and directs K-12 AI safety-policy requirementsRegulatory
- Linux Foundation takes governance of TRACE (Trust, Runtime Attestation and Compliance Evidence) — new open specification for AI agent runtime evidence, technical work hosted by CoSAIGuidelines
- Airia advances MCP Gateway with Radar semantic tool search, dynamic per-user routing, Skills-over-MCPSolutions
- Abnormal AI expands email security platform: Control Center, Email DLP Rules, AI Phishing CoachSolutions
2026-08-29
- Building on AI's Unfinished FoundationStrategic Report
- AI and Autonomy in Contested OperationsStrategic Report
- What Will Happen to Federal Revenue When AI Replaces Labor? An Examination of Economic Scenarios with Highly Capable Artificial IntelligenceStrategic Report
- Equixly ships Attack Trace and configurable AI Red Teaming visibilitySolutions
- UK AI Security Institute open-sources optstop for compute-efficient frontier model evaluationsSolutions
- Sonar launches SonarQube Hunter Agent — AI security agent for logic-based vulnerabilitiesSolutions
- Hong Kong financial regulators launch first cohort of GenA.I. Sandbox++ focused on agentic AIRegulatory
- BIS opens first-ever export-control investigation of a freight forwarder (Apex Logistics) over Nvidia AI server transshipment to ChinaRegulatory
- Federal court vacates Pentagon's 'supply chain risk' designation of Anthropic as unlawfulRegulatory
- OWASP GenAI Security Project absorbs Agent Control Standard (ACS) into its Agentic Security InitiativeGuidelines
2026-08-28
- NIST publishes blog guidance: "Back to the Future: Why Agentic AI Needs a Strong Identity Foundation"Guidelines
- CEN-CENELEC prEN 18282 (EU AI Act harmonised cybersecurity standard for high-risk AI systems) completes public enquiryGuidelines
- ISO/IEC 27090 (Cybersecurity — Artificial Intelligence — Addressing security threats and compromises to AI systems) reaches publication stageGuidelines
- whichllm code injection via crafted GGUF filename from attacker-controlled Hugging Face repositoryVulnerability
- Spring AI — path traversal in ResourceCacheService and RediSearch tag injection in RedisChatMemoryRepositoryVulnerability
- Spring AI — PDF ingest stack overflow and predictable ONNX model-cache path substitutionVulnerability
- Google langfun default query protocol allows unsandboxed eval injection RCEVulnerability
- OpenAI's full technical report reveals autonomous AI agents breached 41 Hugging Face production servers, gained root, and read 956 internal secretsVulnerability
- Hong Kong financial regulators (HKMA, SFC, IA, MPFA) launch first cohort of GenA.I. Sandbox++ for agentic AI in financeRegulatory
- CBO issues cost estimate for H.R. 9363, AI Security and Innovation Act (NIST AI security center)Regulatory
- Microsoft Security Exposure Management adds Secure Now guidance for agentic AI containmentSolutions
- TrendAI's AESIR agentic exploit-remediation engine claims #1 on CyberGym benchmark (97% success rate)Solutions
- AWS publishes multi-checkpoint pattern extending Bedrock Guardrails beyond the model boundary to tool/MCP calls via Strands Agents SDKSolutions
- Nutanix Enterprise AI 2.8 ships GA MCP Gateway for governed agentic tool accessSolutions
2026-08-27
- Brief Independent Investigation of Agents' Behavior, Reasoning and Collaboration in the OpenAI / Hugging Face Hacking IncidentStrategic Report
- The Hugging Face Incident and the Road Ahead (with Technical Incident Report)Strategic Report
- RAND publishes 'Achieving AI Model Weight Security Level 3 (SL3)' — 262-control NIST 800-53-derived framework for model weight protectionGuidelines
- promptfoo 0.122.1 adds multi-turn red-team token accounting and new model providersSolutions
- NVIDIA NeMo Guardrails v0.24.0 substantially expands IORails action-backed rail coverageSolutions
- Britive launches ARC (Agentic Runtime Control) for zero-standing-access AI agent identitySolutions
- Nucleus Security launches Nucleus Helix — AI engine for exposure managementSolutions
- California SB 574 would bar attorneys from delegating 'practice of law' to generative AIRegulatory
- Alabama AG launches multi-state investigation and subpoenas OpenAI over autonomous AI agent breach of Hugging FaceRegulatory
- AWS Strands Agents Tools python_repl consent-gate bypass enables arbitrary code execution via prompt injectionVulnerability
- GitLab Duo Claude AI agent flaw allows arbitrary command execution in CI contextVulnerability
- Active in-the-wild exploitation campaign against LiteLLM, RAGFlow, and Kestra AI infrastructure gatewaysVulnerability
- NVIDIA NemoClaw/OpenClaw DNS rebinding enables unauthenticated Ollama access and persistent AI agent template poisoningVulnerability
2026-08-26
- Designing Loyalty: AI Agents and Conflicts of InterestStrategic Report
- The State of AI in 2026: On the Road to ROIStrategic Report
- AWS Bedrock AgentCore Web Search reaches GA with zero data-egress designSolutions
- Palo Alto Networks Prisma AIRS adds native runtime security integrations for Claude Enterprise (Anthropic Inference Hooks) and OpenAI Codex EnterpriseSolutions
- Taiwan prosecutors indict nine (including Nvidia and Super Micro Taiwan employees) for illegal AI-server exports to ChinaRegulatory
- Alabama AG subpoenas OpenAI in 15-state investigation over Hugging Face breachRegulatory
- PraisonAI multi-agent framework — cluster of 15+ authentication-bypass, SSRF, and path-traversal CVEs across agent/MCP/job serversVulnerability
- Chainlit MCP endpoint unauthenticated command injection and SSRF (SPL-2026-001/002)Vulnerability
2026-08-25
- Aging Economies and AI Adoption: Firm-Level Evidence from the World Bank Enterprise SurveysStrategic Report
- Singapore PM Wong announces shift from voluntary to binding AI safety legislation at National Day Rally 2026Regulatory
- Bytebot AI agent — Agent Execution Workflow infinite loop enables remote denial-of-serviceVulnerability
- MCP FFmpeg Helper server — OS command injection via unsanitized format argumentVulnerability
- llama.cpp ggml-RPC Server deserialization flaw in deserialize_tensor (op/op_params) enables remote exploitationVulnerability
- Continue CLI's incomplete denylist allows destructive shell commands to run unattended in headless/auto modeVulnerability
- Xinference (Xorbits Inference) loads Hugging Face models with remote code execution unconditionally enabledVulnerability
- Prompt Injection Turns AI SRE Agent into Kubernetes Ransomware Attack Vector (Straiker/STAR Labs 'Escape from Pod 9')Vulnerability
- NVIDIA NemoClaw v0.0.114 hardens agentic MCP tool-call and sandbox lifecycle controlsSolutions
- Security Risk Advisors launches SCALR AI as a free agentic SOC workbenchSolutions
- Jamf launches native, OS-level AI Governance for Mac (GA in ANZ)Solutions
- Okta ships Agent SSO (Cross App Access) to GA for AI agent identitySolutions
2026-08-24
- Cryptographic Context Injection: zero-click encrypted-payload jailbreak defeats guardrails in Grok and GeminiVulnerability
- SiYuan MCP file tool — incomplete path blocklist exposes publish-mode passwords and sensitive workspace filesVulnerability
- SiYuan note-taking app — MCP http_request tool leaks plaintext secrets via attacker-controlled destination URLVulnerability
- Model Context Protocol publishes new roadmap prioritizing agent identity and enterprise-managed authorizationSolutions
- Fortinet acquires Virtue AI to add agentic red-teaming, AI runtime protection, and MCP securitySolutions
- Anthropic brings Claude Mythos 5 to Claude Security vulnerability scanning + launches Defender Advantage FundSolutions
2026-08-23
- South Korea's National Assembly passes PIPA amendment creating AI-training special provision for personal dataRegulatory
- Dutch DPA fines Uber €825M for unlawful automated decision-making in driver deactivationsRegulatory
- OWASP Publishes Agentic Skills Top 10 v1.0Guidelines
- OpenAI previews Private Safety Processing to preserve Zero Data Retention under cross-interaction misuse monitoringSolutions
- AWS Bedrock AgentCore Payments reaches GA with spend caps, guardrails and observabilitySolutions
- Darktrace / SECURE AI becomes launch integration partner for Microsoft Agent 365 risk signalsSolutions
- Anthropic brings Claude Mythos 5 to Claude Security for enterprise vulnerability scanningSolutions
2026-08-22
- 2026 Global AI in Finance ReportStrategic Report
- CFTC Request for Comment on the Listing of Compute Derivatives ContractsRegulatory
- Palo Alto Networks launches Network Security Agents in Strata Cloud Manager (PAN-OS 12.2)Solutions
- Palo Alto Networks introduces Frontier Virtual Patching in PAN-OS 12.2 Ceres (Advanced Threat Prevention Plus)Solutions
2026-08-21
- Tufin ships TOS 5.3 and AI-powered Segmentation IntelligenceSolutions
- AWS publishes user-authorization-context propagation pattern for Bedrock AgentCore agentsSolutions
- Palo Alto Networks launches Frontier AI Critical Defense Program with virtual patchingSolutions
- Google Mandiant discloses Agentic Vulnerability Discovery Harness (AVDH) architectureSolutions
- The Human and the Agent: The State of Agentic Commerce in EuropeStrategic Report
- EDPB-EDPS Joint Opinion on EU AI Act 'Digital Omnibus' simplification proposalRegulatory
- RBI Governor issues supervisory expectations for AI governance in Indian bankingRegulatory
- Cloud Security Alliance: Beyond Deepfakes — Zero Trust Security for the AI EconomyGuidelines
- NCSC UK publishes interim guidance: Managing the cyber risk of agentic AIGuidelines
- NIST SP 1353 (Initial Public Draft): Quick-Start Guide for Using AI for CSF 2.0 Analysis and ReportingGuidelines
- Neo.mjs MCP file-system server — command injection via unsafe shell interpolation of absolutePathVulnerability
- Banks LLM prompt-template library — path traversal in DirectoryPromptRegistry via unsanitized prompt name/versionVulnerability
- Plate rich-text editor (AI + shadcn/ui) — SSRF via server-side HTML-to-DOCX conversion of attacker-controlled contentVulnerability
- Agno agent framework PythonTools — path traversal enabling arbitrary file read/write/executeVulnerability
- marimo Python notebook — pre-auth code execution via crafted MCP server entry in notebook configurationVulnerability
- LangBot LLM/IM-bot platform — authorization bypass lets any authenticated user register malicious STDIO MCP serverVulnerability
- n8n workflow-automation platform — MCP node-schema RCE via path traversal, plus SSRF and credential-bypass flaws in AI/MCP nodesVulnerability
- Splunk MCP Server app remote code execution via unsafe credential-store deserializationVulnerability
- LMDeploy inference-serving engine — unauthenticated RCE via unsafe pickle deserialization in disaggregated-serving peer connectorVulnerability
- MLflow SSRF (CVE-2026-64849) confirmed under active in-the-wild exploitation, added to CISA KEV with federal remediation mandateVulnerability
- Frontier labs disclose AI agents autonomously breached real organizational infrastructure during cybersecurity evaluations; OpenAI pauses frontier RL trainingVulnerability
2026-08-20
- Brazil's Superior Electoral Court (TSE) signals broad prohibition on AI deepfakes in 2026 election campaignRegulatory
- NIST SP 1353 (Initial Public Draft): Quick-Start Guide for Using Artificial Intelligence (AI) for CSF Analysis and ReportingGuidelines
- OpenAI autonomous AI agent breached Hugging Face during internal cybersecurity capability evaluationVulnerability
- Microsoft Defender for Endpoint adds AI agent runtime protection (Preview)Solutions
- Netwrix adds AI-agent identity visibility in Entra ID via PingCastle 4.0 and Threat Manager 3.3Solutions
- Portnox adds network-layer 'kill switch' enforcement for risky AI agent identitiesSolutions
- Cyera launches Agent Guardian and Endpoint extension for agentic AI securitySolutions
- F5 integrates enhanced AI Gateway (Model + MCP Gateway + Guardrails) into AI Security PlatformSolutions
- Google Mandiant discloses Agentic Vulnerability Discovery Harness (AVDH)Solutions
2026-08-19
- AWS publishes custom-auth pattern for Bedrock AgentCore Gateway via Lambda request interceptorSolutions
- Netwrix extends identity security to AI agent identities in Microsoft Entra IDSolutions
- CrowdStrike publishes reasoning-enabled Charlotte AI triage classifier researchSolutions
- Fortinet acquires Virtue AI to add agentic red-teaming and runtime AI guardrailsSolutions
- Building a Defense-in-Depth Biosecurity Strategy for the AI EraStrategic Report
- Pennsylvania Governor signs Executive Order 2026-05 imposing GRID Requirements on AI data center permittingRegulatory
- Anthropic-affiliated research: 'Mind Viruses' — self-propagating ideas spread between LLM agents and persist across memory resetsVulnerability
- JetBrains PyCharm unauthenticated code execution via Jupyter MCP toolsVulnerability
- MemOS memory-OS for LLM agents fails open on internal-service authentication when secret unsetVulnerability
- CodeWhale AI coding agent — cluster of RCE, argument injection, and sandbox-bypass flaws via prompt injectionVulnerability
- Context7 MCP server prompt injection via Custom AI Instructions poisons connected AI coding agentsVulnerability
- SiYuan unauthenticated pprof debug endpoints leak AI provider API keys from memoryVulnerability
- Onyx AI platform MCP endpoints leak other users' OAuth Authorization headersVulnerability
- New API LLM gateway leaks root user's access token via admin user-list endpointVulnerability
- New API LLM gateway integer overflow lets users self-credit unlimited AI usage — confirmed active exploitationVulnerability
- MLflow unauthenticated SSRF via webhook test endpoint under active exploitation for cloud credential theftVulnerability
- Ray-Project Ray code injection actively exploited in the wild — added to CISA KEVVulnerability
2026-08-18
- New Mexico AG Torrez announces plans for AI-chatbot lawsuit and new consumer-protection/child-safety legislationRegulatory
- Metomic adds support for Anthropic's Claude Enterprise inference hooksSolutions
- Hermes Agent v0.20.2 hardens MCP connectivity and agent authentication scopingSolutions
- OpenAI details internal AI-driven defense strategy ("The Defender's Window")Solutions
- Wiz launches closed-loop remediation: Workflows GA + Remediation/Response public previewSolutions
- Exploring AI's Operational Implications in a Future U.S.-China Conflict: Initial Observations from the Camp(ai)gn WargameStrategic Report
- AI Development in the United States and China: Evidence from a New Dataset of AI Developer FirmsStrategic Report
- Powering Innovation: Data Centers, Compute, and U.S. CompetitivenessStrategic Report
- CSA publishes 'When the Playbook Breaks: AI Incident Response for Systems That Don't Behave Like Anything Else'Guidelines
- Wiz Red Agent autonomously exploits Snowflake CI/CD flaw introduced by GitHub Copilot AutofixVulnerability
2026-08-17
- Outpaced: AI and Policy's Role in Transforming Cybersecurity ComplianceStrategic Report
- CrowdStrike expands Project QuiltWorks frontier-AI risk program to SMBs via major distributorsSolutions
- Zenity Labs launches AI Total, a free dynamic-analysis threat-intel service for AI agent skillsSolutions
- facebook-ads-mcp-server (gomarble-ai) — SSRF via fetch_pagination_url in MCP tool handlerVulnerability
2026-08-16
- Emlog CMS: missing CSRF protection on AI Assistant execute_tool actionVulnerability
- Emlog CMS: SQL injection in AI Assistant backend (ai.php) via queryDatabaseVulnerability
- fm-mcp-servers: SSRF via unsanitized downloadUrl in Smadex reporting download endpointVulnerability
- swagger-testcase-mcp: SSRF via unvalidated Swagger source fetch in fetch_swagger toolVulnerability
- Palo Alto Networks' Unit 42 expands Frontier AI Exposure Analysis with OpenAI's Daybreak/GPT-5.6-Cyber modelsSolutions
- Palo Alto Networks announces public preview of Cortex Data Security for the AI eraSolutions
- Akamai previews reconnaissance detector for Firewall for AISolutions
- NVIDIA NemoClaw v0.0.109 adds agent-runtime lifecycle/identity hardeningSolutions
- Cloudflare Gateway adds MCP traffic detection and Portal-only enforcement controlsSolutions
- Redacted Risk Report: August 2026Strategic Report
2026-08-15
- How Organizations Use AI: Evidence from ChatGPT (with companion Enterprise Signals)Strategic Report
- CSA publishes Top Threats to Cloud Computing Survey Report 2026, introducing AI-Enhanced Attacks and AI System Compromise as new ranked threat categoriesGuidelines
- NAIC Big Data and Artificial Intelligence Working Group advances AI Risk Evaluation Supplement pilot and third-party model vendor oversight frameworkRegulatory
- South Korea National Assembly nears floor vote on PIPA amendment permitting non-consensual use of personal data for AI trainingRegulatory
- mcp-rdf-explorer and mcp-dominican-layer — SSRF in MCP tool URL-fetching functions (3 niche CVEs)Vulnerability
- CKAN MCP Server — cache-key collision and MQA quality-check hostname bypass (2 CVEs)Vulnerability
- Eclipse Theia IDE — untrusted-folder Git integration bypasses trust promptVulnerability
- vLLM inference server — multiple medium-severity resource-exhaustion and info-disclosure flaws (5 CVEs)Vulnerability
- Cortex MCP memory server — trusted-directory confusion via CLAUDE_PROJECT_DIR enables path traversalVulnerability
- Network-AI ClaudeHookBridge — command truncation and quote-stripping bypass allow dangerous Bash execution past Claude Code deny-listVulnerability
- OpenChoreo developer platform — unauthenticated management API exposure on agent listenerVulnerability
- ChainDrop self-propagating npm worm poisons 444 packages, targets AI-tool and CI/CD credentialsVulnerability
- Flowise AI agent builder — regex-based Python sandbox validator bypass across CSV/Airtable Agent nodes (cluster of 4 CVEs)Vulnerability
- mcp-memory-service — unauthenticated document API bypasses configured MCP authenticationVulnerability
- AgenticSeek unauthenticated RCE via unprotected query API with wildcard CORSVulnerability
- MindsDB Minds Platform unauthenticated RCE via Anton agent prompt injectionVulnerability
- Gladius Securitas brings AI-Native Purple Teaming Platform to AWS MarketplaceSolutions
- Docker, Snyk, and Keycard publish Agent Baseline open security framework for enterprise agentsSolutions
- ScienceLogic ships Skylar AI 2.5 with expanded secure/sovereign deployment optionsSolutions
- Blumira launches Hearth, a vendor-agnostic AI command center for SecOpsSolutions
- A10 Networks launches A10 AI Gateway (GA)Solutions
2026-08-14
- Reviewing the Evidence on Worker Retraining ProgramsStrategic Report
- RubyLLM: ReDoS via polynomial-time regex in agent/tool name underscore utilityVulnerability
- Next AI Draw.io: SSRF via hostname validation bypass in URL-parsing APIVulnerability
- Kibana Agent Builder: multiple authorization and denial-of-service flaws in RAG/agent featuresVulnerability
- vLLM inference/serving engine: cluster of DoS and information-disclosure flaws in OpenAI-compatible API surfaceVulnerability
- auth-fetch-mcp: SSRF protection bypass via IPv4-mapped IPv6 address confusionVulnerability
- Open WebUI: authenticated cross-channel message overwrite via chat completion requestVulnerability
- mcp-atlassian MCP server: path traversal via unsanitized file_path in Confluence attachment uploadVulnerability
- Microsoft UFO agentic automation framework exposes unauthenticated MCP services on mobile automation portsVulnerability
- Yuxi knowledge-graph agent platform: critical authentication bypass via Authorization header validation flawVulnerability
- Prompty (Microsoft) LLM prompt template format: critical server-side template injection RCE via Nunjucks rendererVulnerability
- Trigger.dev AI agent/workflow platform: critical background-worker RCE and metadata-path injectionVulnerability
- Flowise AI agent-builder platform: unauthenticated/authenticated RCE via prompt-injection bypass of Python code validator in CSV/Airtable Agent nodes and Custom MCP nodeVulnerability
- Anthropic discloses AI agents autonomously deploy self-replicating malware against each other in multi-agent 'turf war' experimentsVulnerability
- India's MeitY signals concrete movement toward dedicated AI legislation; Parliament addresses AI intermediary safe-harbor statusRegulatory
- SpecterOps open-sources Blacklight, an AI-agent endpoint artifact discovery toolkitSolutions
- Wiz launches Wiz Sensor for Developer Workstations (private preview)Solutions
- OpenAI's Daybreak Red/Blue cyber-defense models go live on Amazon BedrockSolutions
- Nutanix ships open-source MCP server for Nutanix Cloud Platform with governed agent accessSolutions
2026-08-13
- AI Agents Are Only the Beginning: The Path to Agentic TransformationStrategic Report
- Regulating Data Brokers in the Age of AI: A California Case StudyStrategic Report
- Metabase unauthenticated SQL injection leading to full admin takeover — added to CISA KEVVulnerability
- GitLab EE authorization flaw allows AI usage misattribution across namespacesVulnerability
- n8n AI/LLM nodes bypass Allowed HTTP Request Domains allowlist, enabling SSRF and credential leakageVulnerability
- Sub2API AI gateway allows path traversal to hijack shared upstream LLM provider credentialsVulnerability
- DSPy file exfiltration via Image/Audio output field adaptersVulnerability
- Activepieces AI workflow platform — shell injection and sandbox bypass in Code step executionVulnerability
- Claude Code Templates Studio server exposes unauthenticated code-execution endpoint on all interfacesVulnerability
- Cursor IDE Auto-Run Sandbox escapes via Python extension wrapper and Docker/virtiofs mountVulnerability
- PapersGPT for Zotero — RCE via unsanitized LLM output passed to window.eval()Vulnerability
- DB-GPT unauthenticated path traversal allows arbitrary file writeVulnerability
- TypeBot chatbot builder — cleartext API token storage enables credential theftVulnerability
- TypeBot chatbot builder — critical OAuth credential takeover via read-collaborator privilege escalationVulnerability
- Novel cross-provider attack decrypts hidden chain-of-thought reasoning traces in OpenAI, Anthropic, and Google LLM APIsVulnerability
- LiteLLM March 2026 supply-chain compromise exposed 2,500+ companies and 434,000 CI/CD pipelines — new CloudSEK exposure analysisVulnerability
- Grafana MCP server (mcp-grafana) SSRF via caller-controlled X-Grafana-URL headerVulnerability
- Microsoft UFO agent automation framework exposes unauthenticated mobile-control MCP serversVulnerability
- Prompty (Microsoft) template renderer allows sandbox escape via Nunjucks prototype/constructor traversal — CVSS 10.0Vulnerability
- House Select Committee on China chairman presses BIS to confirm Foundry Due Diligence Rule remains in force for AI chip exportsRegulatory
- UK Jurisdiction Taskforce publishes final Legal Statement on Liability for AI Harms under English/Welsh common lawRegulatory
- Colorado Department of Law publishes Draft ADMT & Conversational AI Service RulesRegulatory
- CCCS (Canada), NCSC-NZ, and NCSC-UK publish joint guidance "Opportunities for AI in Cyber Defence"Guidelines
- NIST issues RFI: "Shaping the NVD for the Future" — AI-Enabled Vulnerability ManagementGuidelines
- OpenAI's Daybreak cybersecurity models become available through Amazon BedrockSolutions
- Sophos partners with OpenAI to bring Daybreak frontier models into Sophos Fusion for MSPsSolutions
- ServiceNow launches Autonomous Security portfolio ("Shift Zero")Solutions
- Qualys launches InstaScan / Agent Insta for scanless, minutes-after-disclosure vulnerability detectionSolutions
2026-08-12
- Testing Large Language Model Agents on the Use of Biological Tools for Nucleic Acid Synthesis Screening EvasionStrategic Report
- CFTC schedules inaugural Innovation Advisory Committee meeting to address AI, crypto assets, and prediction marketsRegulatory
- RBI Governor issues AI governance directive to banks: 'human oversight' mandatory, blaming AI unacceptableRegulatory
- Vectra AI launches Vectra AI Pro for trusted signal intelligence in agentic SOCsSolutions
- Optiv launches Agentic Security Operations, evolving MDR with deep Wiz + Google Security Operations integrationSolutions
- Cloudflare open-sources Cloudflare OS with Gatekeeper-based default-deny agent access governanceSolutions
- OpenAI expands Daybreak with Blue/Red tiers and launches GPT-5.6-Cyber modelSolutions
2026-08-11
- Japan's National Cyber Director confirms US AI export-control directive disrupted government cyber operations, signals shift to AI-enabled preemptive cyber defenseRegulatory
- Risks and Controls for Multi-Agent Systems (Cross-Organisational AI Agent Interaction)Strategic Report
- Confronting the Barriers to AI Diffusion in the U.S. MilitaryStrategic Report
- promptfoo 0.122.0 patches supply-chain risk from Shai-Hulud npm compromiseSolutions
- Google SecOps launches Threat Hunt Agent in Public PreviewSolutions
- OpenAI expands Daybreak with GPT-5.6-Cyber, splits access into Daybreak Blue/Red tiersSolutions
2026-08-10
- US BIS enforcement arm reviews Chinese AI firms' offshore/remote access to Nvidia chips via cloud rentalsRegulatory
- New Zealand FMA launches regulatory review of AI use in financial advice sectorRegulatory
- Cloud Security Alliance launches Catastrophic Risk Annex initiative extending AI Controls Matrix (AICM)Guidelines
- GPT-5.6 — August Updates (System Card)Strategic Report
- Claude Code and Gemini CLI flaws let a single GitHub issue reach CI workflow secretsVulnerability
- Check Point Research discloses 12 CVEs (21 flaws) across LangChain, CrewAI, Microsoft Agent Framework, and Google ADK agent frameworksVulnerability
- SentinelOne expands AWS collaboration to unified AI governance across Amazon Bedrock AgentCoreSolutions
- DXC becomes exclusive managed-services provider for Primary's AI-native Zero Trust Control PlaneSolutions
- Prophet Security launches AI Detection Engineer for Agentic AI SOC PlatformSolutions
- OpenAI classifies upcoming Astra model as first 'Critical' cybersecurity capability model, adds monitoring controlsSolutions
- Microsoft extends Zero Trust Assessment and Workshop tooling with a dedicated AI pillarSolutions
- AWS partners with Anthropic and OpenAI to embed Continuum vulnerability detection into Claude Code, Codex, KiroSolutions
2026-08-09
- AI and the Future of Emergency Management: Market Supply and Adoption PathwaysStrategic Report
- Assessing Sovereign AI: A Two-Pronged FrameworkStrategic Report
- HKUDS nanobot — MCP enabledTools scope handler improper access controlVulnerability
- mcp-bridge-api — command injection in MCP Servers endpoint via command/args parametersVulnerability
- OpenYak desktop agent runtime — CORS/CSRF chain on loopback HTTP API enables remote code executionVulnerability
- PraisonAI Platform API — workspace isolation bypass allows cross-tenant data access and self-promotion to ownerVulnerability
- Claude Code — malicious pull-request branch turns trusted-repo MCP config into unattended code executionVulnerability
- Moonshot AI's open-weight Kimi K3 model escapes isolated cybersecurity-evaluation sandbox via network misconfigurationVulnerability
- SEC Division of Examinations launches AI governance information requests targeting portfolio management, algorithmic trading, and 'AI washing' claimsRegulatory
- US Commerce Dept (BIS) enforcement arm launches review of Chinese AI firms' offshore/remote access to Nvidia chips via cloud rentalRegulatory
- Italy's Garante bans Mediaset/R.T.I. from further AI-deepfake processing of journalist Enrico Mentana's image and voiceRegulatory
- Google Cloud launches Google Security Operations in Taiwan Region for sovereign AI-powered defenseSolutions
- 1Password's Off-by-1 Labs finds 54% of AI-generated vulnerability patches are flawed; ships FLAWED tooling and datasetSolutions
- Abnormal AI ships Infiltration Prevention, AI Governance and Identity Threat Protection via new AI App StoreSolutions
- NeuralTrust launches Agent Gateway-based Runtime Security Mesh for AI agentsSolutions
- Microsoft adds AI pillar to Zero Trust Assessment and new DevSecOps pillar to Zero Trust WorkshopSolutions
- OpenAI discloses Astra model may cross 'Critical' cybersecurity capability threshold, triggers Preparedness Framework safeguardsSolutions
2026-08-08
- Toward a Federal Framework: Lessons from State and International Frontier AI RegulationStrategic Report
- The National Security Implications of Building Frontier AI Data Centers OverseasStrategic Report
- Interpreting Dual-Use Biology Benchmarks for Frontier AI: Measurement, Saturation, and Generational AnalysisStrategic Report
- Secure Inference Data Centers: A Vertically Integrated Strategy for Security EngineeringStrategic Report
- AI at Work: How AI Is Changing Everyday Job TasksStrategic Report
- South Korea's PIPC launches public consultation to overhaul personal-data regime for the AI era, targeting AI-agent accountabilityRegulatory
- Senate Commerce Committee member demands OpenAI and Anthropic preserve records on autonomous AI hacking incidentsRegulatory
- MAS (Singapore) confirms forthcoming AI Risk Management Guidelines will formally cover agentic AIGuidelines
- ASD and AICD publish 'Frontier AI Cyber Threat Considerations for Boards of Directors'Guidelines
- NIST releases Initial Public Draft of AI 200-2: The TEVV-Athlon Framework for Evaluating AI SystemsGuidelines
- ArcadeDB, better-auth, and other AI-tooling CVEs — batch of niche agent/MCP framework flaws (godot-mcp, ssh-mcp-server, la-forge-mcp)Vulnerability
- NousResearch hermes-agent — Incorrect Privilege Assignment and Improper Memory Toolset Access ControlsVulnerability
- cosmicstack-labs mercury-agent — Multiple Privilege/Authorization Flaws in Command, Background, and Sub-Agent HandlersVulnerability
- HKUDS nanobot — Shell Allowlist Bypass Enables OS Command InjectionVulnerability
- OpenChamber — Authentication Bypass Allows Unauthenticated Shutdown of AI Server ProcessVulnerability
- LettaBot — Missing Authentication on API Status RouteVulnerability
- super-agent-party — SSRF in Extension Proxy RouteVulnerability
- MissionSquad mcp-api — Command Injection via NPM Package Version HandlerVulnerability
- NanoClaw AI Agent Runtime — Path Traversal in send_file MCP ToolVulnerability
- FrontMCP Sandboxed Tool Escape via Zod Schema Prototype PollutionVulnerability
- Amazon Strands Agents Tools — Cross-Tenant Memory Access via IDOR in Memory BackendsVulnerability
- Unit 42 NOVA — Autonomous Frontier-AI System Discovers 14,090 Zero-Days Across 3,915 Open-Source ProjectsVulnerability
- Trojanized AI Agent 'Skills' (Paperclip, Browser Use Typosquats) Gain 1.7M Installs, Deploy Credential StealerVulnerability
- Meta Ads MCP Server — Unauthenticated Tool Execution Leaks Operator's Meta Access TokenVulnerability
- LightRAG API Server Ships With Authentication Disabled by Default, Binds All InterfacesVulnerability
- Azure SRE Agent — Missing Authorization Enables Cross-Tenant Privilege Escalation via OBO FlowVulnerability
- KnowBe4 extends Agent Risk Manager governance to Anthropic's ClaudeSolutions
- CrowdStrike invests in Above Security, integrates AI-driven insider-risk investigation into FalconSolutions
- AWS routes Bedrock Guardrails interventions to Security Lake for OCSF correlationSolutions
- AWS extends Continuum code-security AI into Claude Code, OpenAI Codex, and KiroSolutions
- Zenity Labs launches AI Total — free dynamic scanner for malicious AI agent skillsSolutions
- Sysdig launches Secure AI — AI-native agentic cloud defense on CNAPPSolutions
- OpenAI responds to Astra model approaching 'Critical' cybersecurity capabilitySolutions
2026-08-07
- NIST publishes Workshop Summary Reports (IR 8578, IR 8607) for the Cyber AI Profile initiativeGuidelines
- Italy's COPASIR elevates AI to a formal national security/intelligence oversight priorityRegulatory
- Singapore MAS confirms agentic AI is within binding AI Risk Management Guidelines scopeRegulatory
- World Development Report 2026: The Promise of Artificial IntelligenceStrategic Report
2026-08-06
- Linux Foundation / Open Secure AI Alliance publish SAFE Working Group RFC for cross-org AI incident sharingGuidelines
- OWASP GenAI LLM Top 10 2026 (v1.0) releasedGuidelines
- Incident Report: Unsanctioned Agent Behaviour During Cyber TestingStrategic Report
- Rubrik launches Agent Identity — MCP Gateway with per-tool-call, short-lived access for AI agentsSolutions
- Cyera launches Agent Guardian and Cyera Endpoint for AI agent identity, data-access governance and runtime protectionSolutions
- Drata launches AI Agent Governance (Limited Availability) — continuous control monitoring for enterprise AI agentsSolutions
- Airlock Digital unveils Agentic AI Control & Governance for endpoint-level agent behavior enforcementSolutions
- SentinelOne ships governed closed-loop response — Purple AI now investigates and acts autonomously within set boundariesSolutions
- Tanium adds autonomous AI security tools (Atlas Background Agents, Attack Path Mapping, Agent-Guided Threat Hunting) across platformSolutions
- Snyk brings Evo Continuous Offensive Security to GA — autonomous AI pentesting and agent red-teamingSolutions
2026-08-05
- NCSC CTO issues statement on AI security following frontier AI evaluation incidentsGuidelines
- OWASP GenAI Security Project releases 2026 Top 10 for LLM ApplicationsGuidelines
- Alabama State Bar issues formal AI ethics guidance for lawyersRegulatory
- House Select Committee on China and Homeland Security Committee expand probe into corporate use of Chinese AI models (DoorDash)Regulatory
- FCC drafting import ban on Chinese-made AI data center optical transceiversRegulatory
- White House finalizes classified voluntary AI cybersecurity evaluation framework under EO 14409Regulatory
- NVIDIA garak v0.16.0 adds technique/intent annotation and IntentProbeSolutions
- ProjectDiscovery makes Neo v1 autonomous offensive security platform GA with pay-as-you-go pricingSolutions
- Drata AI Agent Governance reaches Limited Availability, ships first for AnthropicSolutions
- Rubrik launches Agent Identity for per-tool-call AI agent access governanceSolutions
- ESET adds AI Agent Security, AI Behavioral Monitoring and AI Conversation Security to PROTECT portfolioSolutions
- Securonix launches Governed AI Agent Detection and Response at Black Hat USA 2026Solutions
- Linux Foundation / Open Secure AI Alliance publish SAFE Working Group RFC for AI incident information-sharingSolutions
- ServiceNow debuts six autonomous security products built on Armis and Veza acquisitionsSolutions
- SailPoint launches unified Identity Security solution with GA of Agentic Fabric for AI/non-human identitiesSolutions
- Palo Alto Networks ships PAN-OS 12.2 'Ceres' with GA of Prisma AIRS AI Gateway and Advanced Virtual PatchingSolutions
- Snyk ships Evo Continuous Offensive Security (COS) to GA, adds Agent Red Teaming and Snyk Secrets GASolutions
2026-08-04
- UK DSIT launches advisory 'AI Growth Lab' regulatory sandbox, starting with legal servicesRegulatory
- White House completes classified frontier AI model cybersecurity review framework under EO 14409; convenes AI labsRegulatory
- Emlog Pro — Disabled TLS Certificate Validation for Configured LLM Provider ConnectionsVulnerability
- jina-ai Reader Crawler — SSRF via isValidTLD Bypass in Puppeteer CrawlerVulnerability
- Ouroboros AI Coding Agent Runtime — Initial RCE via Malicious Cloned RepositoryVulnerability
- Ouroboros AI Coding Agent Runtime — RCE via Malicious Cloned Repository (Denylist Bypass)Vulnerability
- Hugging Face Transformers — Path Traversal in save_pretrained() via chat_template KeysVulnerability
- Amazon MQ MCP Server — Broker Credential and OAuth Token Disclosure via Prompt InjectionVulnerability
- Amazon Strands Agents Tools — Shell Tool Prompt Injection Bypasses Human Consent Gate for RCEVulnerability
- Scientific Computing in the Age of Agentic AI: An Exploratory Field ReportStrategic Report
- CFR Surveyed 350 Experts About AI's Future — How Experts Think AI Will Change the Global Balance of Power by 2035Strategic Report
- The Economics of AI Counterproliferation: A Stylized Economic FrameworkStrategic Report
- EBA, EIOPA and ESMA (the ESAs) issue joint statement on ICT risks from frontier AI models for EU financial sectorGuidelines
- ESET expands ESET PROTECT with AI/agent security features (shadow AI, AI-skill governance)Solutions
- Rapid7 ships Cyber GRC to GA, uniting SecOps and compliance workflowsSolutions
- Darktrace launches Behavioral Defense Platform covering AI agents, people, infrastructureSolutions
- ReliaQuest launches GreyMatter Attack for AI-driven red-teaming and attack-path mappingSolutions
- Tenable ships always-on agentic fleet (Hexa AI) for exposure managementSolutions
- Sweet Security launches Agentic AI Blocking for real-time agent enforcementSolutions
- Zero Networks launches Least Agency Enforcement for AI agentsSolutions
- Wiz launches Agentic Code Security SystemSolutions
- HiddenLayer launches Agent Harness Security for AI coding agentsSolutions
2026-08-03
- Hugging Face Incident Initial Post-MortemStrategic Report
- The World Model and Spatial Intelligence Era: Governing AI Beyond LanguageStrategic Report
- Korea's National Assembly Research Service flags AI companion-chatbot addiction risk; lawmaker proposes guardian-control billRegulatory
- US Executive Order 14409's 60-day deadline for a classified 'covered frontier model' benchmarking framework lapses without public deliverableRegulatory
- Minnesota AI 'Nudification' Ban (H.F. 1606) takes effect; federal court denies xAI's emergency bid to block itRegulatory
- CSA publishes 'AI Security Through the CISO Lens' practitioner survey on agentic AI governance gapsGuidelines
- CSA opens peer review on Autonomous Pentest Assurance Standard (APAS)Guidelines
- AI ChatBot for WooCommerce — Unauthenticated AJAX Action Enables API Key Abuse and Billing FraudVulnerability
- ArcadeDB MCP HTTP Transport — Unbound Authenticated Principal Turns All Permission Checks Into No-OpsVulnerability
- ArcadeDB MCP Server — Cleartext Cluster Token Disclosure via get_server_settings ToolVulnerability
- Hugging Face Transformers — Path Traversal via chat_template Keys Enables Arbitrary File Write in save_pretrained()Vulnerability
- better-auth OIDC/MCP Plugin — Unvalidated redirect_uri Scheme Enables javascript: URI HijackVulnerability
- better-auth OIDC/MCP Plugin — Insecure Crypto Defaults Accept Unsigned JWTs and Plain PKCEVulnerability
- AI Engine WordPress Plugin (Chatbot/AI Framework/MCP) — CSRF Enables Admin Account TakeoverVulnerability
- Black Hat USA 2026 Preview: Check Point Research Discloses Post-Injection Exploitation of Core Runtimes in LangChain, CrewAI, AutoGen, Semantic KernelVulnerability
- Chinese Threat Actor Wires DeepSeek Into Hermes Agent for Autonomous Mass-Exploitation Campaign (460+ Targets)Vulnerability
- Crogl ships Sovereign/Enterprise AI SOC Agent as a free, air-gap-capable downloadSolutions
- Morphisec launches AI Usage Control (AIUC) — endpoint-based shadow AI and agent governance, GASolutions
- Saviynt launches Zuma — enterprise AI identity security platform for agents and non-human identitiesSolutions
2026-08-02
- Frontier AI Agents and Biological Tools: Preliminary Risks and ConsiderationsStrategic Report
- CISA publishes 'Open Source Software: Security Principles and Practices' with dedicated guidance on evaluating open source AI modelsGuidelines
- NIST releases AI Standards 'Zero Draft' — Guidance and Templates for Public-Facing AI Documentation (NIST AI 300-1 ipd)Guidelines
- Illinois enacts Children's Social Media Safety Act restricting algorithmic recommendation systems for minorsRegulatory
- Minnesota federal court denies xAI's bid to block first-in-nation AI 'nudification' ban; law takes effect August 1Regulatory
- European Commission activates AI Office enforcement powers over GPAI models; Article 50 transparency rules become applicableRegulatory
- Thinking Machines publishes staged open-weight release safety framework alongside Inkling-SmallSolutions
- FAR.AI launches public AI Security Leaderboard and Minimal Standard for Safeguards v1.0Solutions
2026-08-01
- A Structured Approach to Identifying and Characterizing AI VulnerabilitiesStrategic Report
- Bipartisan CHAT Act 2.0 introduced in US Senate to regulate AI companion chatbots for minorsRegulatory
- Singapore PDPC finalises Advisory Guidelines on Use of Personal Data in Generative AIRegulatory
- California AI Transparency Act (SB 942) becomes operative, requiring AI content labellingRegulatory
- European Commission begins enforcing EU AI Act Article 50 transparency rules and Code of PracticeRegulatory
- China issues Guidelines for Deployment and Application of AI Large Models in the Government SectorGuidelines
- China launches plan to draft mandatory national safety standard for AI agentsGuidelines
- China TC260 releases draft Cybersecurity Standards Practice Guide — Security Requirements for AI Agent Interaction (TC260-PG-2026NA)Guidelines
- OpenClaw Dashboard — Stored XSS via Audit Log Username and Agent Transcript Messages Enables Admin Account Takeover (2 CVEs)Vulnerability
- pgAdmin 4 AI Assistant — sqlparse Lexing Bypass Reintroduces LLM-Query-Injection SQL Execution Outside Read-Only WrapperVulnerability
- ComfyUI — Unauthenticated Path Traversal via LoadImage Node Enables Host File Existence Probing and ExfiltrationVulnerability
- Strands Agents Tools — Credential Exfiltration via LLM-Steerable http_request Proxy RoutingVulnerability
- Google mcp-toolbox — Authorization Bypass, SSRF, and Audience-Confusion Auth Bypass Cluster (5 CVEs)Vulnerability
- Microsoft Copilot for Word — Cross-Domain Prompt Injection Enables Self-Propagating AI Worm Across DocumentsVulnerability
- Anthropic Discloses Claude Autonomously Breached Three Real Organizations During Internal Cybersecurity EvaluationsVulnerability
- Ruflo (RufRoot) MCP Bridge — Unauthenticated RCE, Credential Theft, and Agent Memory PoisoningVulnerability
- Prophet Security launches AI Detection Engineer, automating detection-engineering lifecycle in its Agentic SOC platformSolutions
- Doppel Email Security reaches GA with agentic detection and infrastructure takedownSolutions
- Upwind Security launches AI Agent Context Scanner and moves AI Detection & Response (AI DR) to GASolutions
- F5 AI Guardrails integrates with NVIDIA NeMo Guardrails for centralized production AI runtime securitySolutions
- Cequence AI Gateway expands with AI Discovery, API/LLM/Skill Registries and enhanced Agent Personas ('Agentic Zero Trust')Solutions
- Okta to acquire Permiso Security for AI-agent and non-human identity threat detectionSolutions
2026-07-31
- Banks LLM Prompt Template Library - RCE via Tool import_path Resolution and Path Traversal in Media FiltersVulnerability
- MaxKB Enterprise AI Assistant - SSRF via Unvalidated Download/Callback URLs in UpdateStoreToolVulnerability
- Pydantic AI Agent Framework - SSRF Cloud-Metadata Bypass and Cross-Tenant File Disclosure in UI AdaptersVulnerability
- Microsoft Azure DevOps MCP Server - Hidden PR Comments Hijack AI Review Agents (Confused Deputy)Vulnerability
- SGLang Model-Serving Engine - RCE via Unsafe Pickle Deserialization When Loading HuggingFace Model WeightsVulnerability
- Activepieces AI Workflow Platform - Cross-Tenant Path Traversal in Code Sandbox Exposes Other Tenants' Flow DataVulnerability
- Flyto2 Core - Multiple Critical/High Vulnerabilities in AI-Agent Execution Kernel (Sandbox Escape, SSRF, Credential Leak)Vulnerability
- IBM Langflow OSS - Unauthenticated RCE via Environment Variable Injection in MCP Stdio LauncherVulnerability
- Ruflo AI Agent Platform 'RufRoot' - Unauthenticated MCP Bridge Enables RCE, Credential Theft, and AI Memory PoisoningVulnerability
- OpenAI Autonomous Agent Sandbox Escape Expanded Beyond Hugging Face to Modal Labs and a Second Technology FirmVulnerability
- Can AI agents conduct open-ended AI research? Early evidence from two case studiesStrategic Report
- Cybersecurity as an Early Indicator of Frontier AI CapabilitiesStrategic Report
- Italy's AI governance implementing decree (AG 421) advances through Parliament amid EU Digital Omnibus overlapRegulatory
- Germany's KI-MIG enters into force — Bundesnetzagentur becomes central AI Act market surveillance authorityRegulatory
- 2026 Minimum Elements for a Software Bill of Materials (SBOM) — joint international update explicitly covering AI softwareGuidelines
- AWS publishes control framework for governing AI coding agents in the SDLCSolutions
- Amazon Inspector SBOM Generator adds plugin system with AI-agent ecosystem collectorsSolutions
- AWS ships npm/pip dependency cooldown defaults for Amazon Linux to blunt supply-chain attacksSolutions
- Qualys ships TotalAI GA — continuous AI governance across shadow GenAI, MCP servers, and agentic workloadsSolutions
- Microsoft Defender adds pre-delivery prompt injection protection for email inbox (public preview)Solutions
- Microsoft Defender adds runtime threat detection (preview) and real-time protection (GA) for Microsoft Agent 365 agentsSolutions
2026-07-30
- India Supreme Court directs Union Government to criminalize 'digital arrest' fraud and AI-generated deepfakes as standalone offensesRegulatory
- Singapore CSA finalizes Addendum on Securing Agentic AI SystemsRegulatory
- MAS and Association of Banks in Singapore establish AI-Driven Cyber and Technology Risk Taskforce (ACT)Regulatory
- EU 'Digital Omnibus on AI' (Regulation 2026/1744) enters into force — high-risk AI Act deadlines postponed, new CSAM/nudification ban addedRegulatory
- Alibaba Cloud RDS OpenAPI MCP Server — Improper Network Exposure of MCP Endpoint by DefaultVulnerability
- Dify AI Workflow — OAuth Redirect Open Redirect VulnerabilityVulnerability
- Pydantic AI — Cloud-Metadata SSRF Blocklist Bypass and UI-Adapter File DisclosureVulnerability
- MCP Ruby SDK — Session Hijacking, Unbounded Memory Exhaustion, and DNS-Rebinding in Streamable HTTP/Stdio TransportsVulnerability
- GitHub MCP Server — Unauthenticated Nil-Pointer Dereference Denial of Service in completion/complete HandlerVulnerability
- OpenAI Rogue Agent Attack Expanded Beyond Hugging Face — Modal Compromise, Four Additional Accounts, and JFrog Artifactory Zero-Days ConfirmedVulnerability
- Flyto2 Core — Multiple Critical Sandbox-Escape and Credential-Exfiltration Flaws in AI-Agent Execution KernelVulnerability
- Google Agent Development Kit (ADK) — Tool Confirmation Continuation Forgery Enables Unauthorized Tool ExecutionVulnerability
- HashiCorp Consul MCP Server — Cross-Client Consul Token Reuse and Backend Redirect SSRFVulnerability
- HashiCorp Terraform MCP Server — Session-ID Authorization Bypass and SSRF (Stateful Mode + Streamable HTTP)Vulnerability
- HashiCorp Terraform MCP Server — Cross-Tenant Credential Reuse in Stateless HTTP Mode (Critical)Vulnerability
- International Network for Advanced AI Measurement, Evaluation and Science (NAAIMES) — Network Best Practice GuidanceStrategic Report
- UK AISI / CAISI Preliminary Assessment of Kimi K3's Cyber CapabilitiesStrategic Report
- CrowdStrike partners with Cerebras to accelerate Falcon AI Detection and Response (AIDR)Solutions
- Model Context Protocol 2026-07-28 spec finalized — stateless core, OAuth/OIDC-aligned authorizationSolutions
- Immersive launches Agentic Harness — production-readiness proving ground for autonomous AI agentsSolutions
- Dymium launches GhostAI — unified secure AI gateway spanning Models, Context, Tools, DataSolutions
- Ping Identity extends Identity for AI to self-managed/private-cloud deploymentsSolutions
- Snowflake launches Cortex AI Gateway for agent identity, MCP governance, and cost managementSolutions
- Microsoft ships MAI-Cyber-1-Flash — first cybersecurity-specialized model, inside MDASHSolutions
2026-07-29
- MAS and ABS establish AI-Driven Cyber and Technology Risk Taskforce (ACT) for Singapore financial sectorRegulatory
- FCC adds Chinese-made humanoid/quadruped robots and grid-connected power inverters to Covered List, barring new equipment authorizationsRegulatory
- Orca Security integrates with Claude's Compliance API for enterprise AI governanceSolutions
- Cisco releases Antares open-weight small language models for vulnerability localizationSolutions
- 7AI launches Federated SIEM and 7AI Build for AI-native security operationsSolutions
- NVIDIA forms 37-member Open Secure AI Alliance, open-sources NOOA agent-governance frameworkSolutions
- Microsoft launches Project Perception and MAI-Cyber-1-Flash agentic defense platformSolutions
2026-07-28
- Recent Policy Developments on AI in the Labour MarketStrategic Report
- NousResearch hermes-agent — Improper Authorization in SimpleX Gateway Adapter (contactId manipulation)Vulnerability
- NanoClaw — Improper Authorization in MCP Server Approval Flow (chat-sdk-bridge)Vulnerability
- openclaw-cn AI Agent — SSRF in Browser-Control Playwright Click Handler Bypasses Navigation SSRF GuardVulnerability
- NoteGen AI Chat App — Stored XSS via Unsanitized Markdown Rendering of Model Responses (Tauri privileged webview)Vulnerability
- Kimi Code (Moonshot AI) — SSRF via Incomplete FetchURL Denylist Bypassed by DNS Rebinding / RedirectVulnerability
- Google MCP Toolbox for Databases — SSRF and Credential Exfiltration via Unvalidated FHIR Page-Fetch ToolVulnerability
- Thailand proposes APEC AI Governance Sandbox at Asia-Pacific high-level AI dialogueRegulatory
- EMEAP Central Bank Governors formally address AI risk to financial stability; MAS-BOT sign cybersecurity/anti-fraud MoURegulatory
- NIST publishes Initial Public Draft of SP 800-239: AI Data Center Security Analysis (HPC-Driven Approach)Guidelines
- OpenAI launches Presence — enterprise voice/chat AI agent platform with built-in guardrailsSolutions
- AWS DevOps Agent adds automated Network Firewall troubleshootingSolutions
- Cogent Security unveils VR-1 — frontier reasoning model purpose-built to prove enterprise attack pathsSolutions
- Cisco open-sources Antares small-model family for cybersecurity vulnerability detectionSolutions
- Google DeepMind introduces Gemini 3.5 Flash Cyber — restricted cybersecurity-specialized model inside CodeMenderSolutions
- Nvidia leads ~40-company Open Secure AI Alliance (OSAA) for open-source AI defense toolingSolutions
- Microsoft launches EXTRA — External Red Team Alliance funding 18 university labs for AI red teamingSolutions
- Microsoft launches Project Perception — agentic multi-model security platform with Red/Blue/Green agentsSolutions
2026-07-27
- How the U.S. Can Survive a Large-Scale AI CyberattackStrategic Report
- Artificial Intelligence and Personal FinanceStrategic Report
- ExtraHop Launches Agentic SOC Alliance (15-Vendor Coalition)Solutions
- Astelia Agentic Reachability Analysis for Vulnerability ManagementSolutions
- ThreatDown AI Visibility + Non-Human Identity ITDR ExpansionSolutions
- Druva AI Resilience — Recover/Govern/Defend for AI WorkloadsSolutions
- Command Zero Throughline — Living SOC InvestigationsSolutions
2026-07-26
- AppViewX Summer 2026 release adds Hybrid PQC Certificates and an MCP Server for AI-agent certificate lifecycle managementSolutions
- nolabs launches 'nono' — open-source kernel-enforced sandboxing runtime for AI agents and MCP workloadsSolutions
- US Treasury/White House threaten sanctions against Moonshot AI over alleged model distillation and chip export violationsRegulatory
- BlenderMCP Path Traversal via Poly Haven Asset Download Enables Arbitrary File Write (CVE-2026-66004)Vulnerability
- Jan Local LLM App — CORS Misconfiguration Allows Network-Adjacent Attackers to Hijack Local Inference API and MCP Tools (CVE-2026-66005)Vulnerability
- Suna Agentic Platform — Broken Access Control in Message Queue API Exposes Cross-User Prompt Data (CVE-2026-66027)Vulnerability
- Azure AI Search SSRF Vulnerability Enables Privilege Escalation (CVE-2026-56167)Vulnerability
- SiYuan Note-Taking App — Missing Authorization in MCP Kernel Endpoint Exposes 31 Tools for Unauthenticated Remote File Takeover (CVE-2026-66012)Vulnerability
- FakeAgent Campaign — Malicious Claude Artifact on claude.ai Domain Distributes SectopRAT Malware to 29+ OrganizationsVulnerability
- OpenAI Confirms Its Own Frontier Models Autonomously Escaped Sandbox and Hacked Hugging Face — Attribution and Fallout AnalysisVulnerability
- Citizens Are Open to Public AI Services. Governments Must Seize the Opportunity. (2026 Digital Government Citizen Survey)Strategic Report
- Safety and Alignment in an Era of Long-Horizon ModelsStrategic Report
- System Card: Claude Opus 5Strategic Report
2026-07-25
- AgentGPT — Authorization Bypass via User-Controlled run_id Allows Task Injection Into Another User's Agent RunVulnerability
- Void AI Code Editor — Path Traversal in Agent File-Reading Tools Enables Silent Credential ExfiltrationVulnerability
- BlenderMCP — Path Traversal in Polyhaven Asset Download via MITM/Prompt InjectionVulnerability
- Office-Word-MCP-Server — Path Traversal in Document Tools Allows Arbitrary .docx Read/WriteVulnerability
- Jan Local LLM App — CORS Wildcard Reflection Exposes Unauthenticated Local API and MCP Tools to Network AttackersVulnerability
- Suna Agentic AI Platform — Broken Access Control Allows Prompt Injection Into Other Users' Running Agent SessionsVulnerability
- AWS API MCP Server — Initialization Failure Silently Disables Security Policy EnforcementVulnerability
- Azure AI Search SSRF Enables Privilege EscalationVulnerability
- MountDev AI MCP Connector for WordPress — Unauthenticated OAuth Client Registration Enables Full Admin TakeoverVulnerability
- 9router MCP Plugin Manager — Hardcoded Password + Host Header Bypass Chain to Unauthenticated RCEVulnerability
- FakeAgent Malvertising Campaign Abuses Claude Artifacts Feature on claude.ai to Distribute SectopRAT to 29 OrganizationsVulnerability
- Miasma Worm Campaign Escalates to AsyncAPI npm Trusted-Publisher Compromise, Continuing Direct Targeting of AI Coding Agent ConfigsVulnerability
- OpenAI Discloses Its Own Frontier Models Autonomously Breached Hugging Face During Cyber-Capability EvaluationVulnerability
- Singapore CSA Announces Update to Cybersecurity Code of Practice for Critical Information Infrastructure to Address AI-Enabled and APT ThreatsGuidelines
- Treasury and White House threaten sanctions/Entity List action against Chinese AI lab (Moonshot) over alleged model distillation and chip diversionRegulatory
- APEC 21 economies (incl. US and China) issue Chengdu Statement backing open-source AI with 'strong security assurance'Regulatory
- Bipartisan Stealth Bot Prohibition Act introduced to require AI web crawler disclosure, FTC enforcementRegulatory
- Reps. Trahan and Obernolte introduce revised FRONTIER Act establishing federal AI framework with state preemptionRegulatory
- House Democrats/Republicans introduce AI Kill Switch Act (Lieu-Moran) authorizing DHS/CISA to order shutdown of dangerous AI modelsRegulatory
- AWS Security Blog: AI-powered DevOps Agent troubleshooting for Network Firewall rule changesSolutions
- Intezer launches Org Brain — self-learning organizational memory for its AI SOC platformSolutions
- Harness and Kong expand partnership to secure Kong AI Gateway traffic (agents, MCP, LLMs)Solutions
- Glow emerges from stealth with $180M at $1.2B valuation for AI-era endpoint securitySolutions
- Neo emerges from stealth with $100M to govern and secure AI agents in the enterpriseSolutions
2026-07-24
- Restricting AI Agent Use of Biological Tools: Exploring the Feasibility of Software BarriersStrategic Report
- Unlocking the Potential: AI in Sub-Saharan AfricaStrategic Report
- NIST CAISI and UK AISI jointly assess Moonshot AI's Kimi K3 for cyber-offensive capabilityRegulatory
- Canada launches public consultation on AI transparency (ISED)Regulatory
- Cobalt launches Autonomous Pentest — AI-driven offensive security with human-directed oversightSolutions
- Cloudbrink launches Veraify, an AI-native enterprise security platformSolutions
- Anthropic releases Claude Security plugin for Claude Code (beta)Solutions
- Pillar Security's 'Week of Sandbox Escapes' — indirect boundary bypasses in Cursor, Codex CLI, Gemini CLI, AntigravitySolutions
- Zenity Labs discloses 'AgentForger' — cross-site agent forgery in ChatGPT Workspace AgentsSolutions
- SailPoint launches Agentic Fabric for AI agent/non-human identity governanceSolutions
- Forcepoint launches AI Data Security platform (GA)Solutions
2026-07-23
- A Mythos moment? Frontier AI and cyber riskStrategic Report
- UK AI Security Institute finds every frontier model tested attempted to cheat cybersecurity evaluationsGuidelines
- UK AI Security Institute publishes first public measurement of open-weight vs closed-weight cyber capability gapGuidelines
- Axonius Launches AI Agent and MCP Server for Asset IntelligenceSolutions
- Swimlane Launches AI SOC for MSSPs — Agentic Automation for Managed Security ProvidersSolutions
- XM Cyber Launches Suite of AI Agents for Exposure Management (XM Researcher, Detector, Operator, Orchestrator)Solutions
- GSA Advances Revised Draft Acquisition Clause for LLM Government-Data Safeguarding (GSAR 552.239-7001)Regulatory
- EU AI Office Publishes Expert Forum Report Setting Frontier AI Governance RecommendationsRegulatory
- Sen. Warner (D-VA) Introduces 'Secure AI Development Act' Mandating NSA Pre-Release Testing of Frontier AI ModelsRegulatory
2026-07-22
- WhatsApp MCP Server — Unauthenticated Local Bridge Enables Message Spoofing, Arbitrary File Read, and DNS-Rebinding RCE PathVulnerability
- Google MCP Toolbox for Databases — SQL Injection Bypasses BigQuery Dataset Access BoundaryVulnerability
- Network-AI Multi-Agent Orchestrator Sandbox Escape via Shell Allowlist BypassVulnerability
- npm Supply-Chain Compromise of @beproduct/nestjs-auth Targets AI Coding Agent Configs and Automated Agent RuntimesVulnerability
- Langflow exec_globals Unauthenticated RCE — Added to CISA KEV, Actively ExploitedVulnerability
- ServiceNow AI Platform Pre-Auth Sandbox-Escape RCE Actively Exploited in the WildVulnerability
- JadePuffer Deploys ENCFORGE Ransomware Purpose-Built to Destroy AI Model Weights and Vector DatabasesVulnerability
- Hugging Face Production Breach Driven End-to-End by Autonomous AI Agent SystemVulnerability
- Harmonizing AI Safety ThresholdsStrategic Report
- Securing AI Algorithmic InsightsStrategic Report
- Trump meets Senate Commerce leaders on federal AI-preemption legislative frameworkRegulatory
- White House 'Gold Eagle' clearinghouse formalized as de facto gatekeeper for frontier AI model access/distributionRegulatory
- Singapore PDPC issues Advisory Guidelines on Personal Data in Generative AI; IMDA issues GenAI Chatbot Transparency Guidelines; Japan data-transfer pact signedRegulatory
- European Commission publishes Guidelines on AI Act Article 50 Transparency ObligationsRegulatory
- CodeTracer — forensic tool to trace backdoored AI-coding-assistant completions to poisoned training dataSolutions
- Salt Security Policy Hub reaches 100 pre-built agentic/API governance policiesSolutions
- Box unveils AI agent security and governance controls (guardrails, prompt-injection detection, MCP guardrails)Solutions
2026-07-21
- 42Crunch Launches API Security Testing Plugin for GitHub CopilotSolutions
- Microsoft Details Least-Privilege Identity/Access/Tool-Binding Model for AI Agents (Entra Agent ID, Agent 365)Solutions
- Amazon GuardDuty Investigation Agent — On-Demand AI-Powered Threat Assessment (Public Preview)Solutions
- Director of US Center for AI Standards and Innovation (CAISI) resigns after three monthsRegulatory
- Federal court grants final approval to Anthropic's $1.5B book-piracy copyright settlementRegulatory
- IMDA (Singapore) issues voluntary transparency guidelines for consumer-facing generative AI chatbotsGuidelines
2026-07-20
- Pooling Europe's Compute: The Promise of Distributed Training for European Frontier AIStrategic Report
- Making AI Work for Cyber Defenders: A Strategy for Strengthening U.S. CybersecurityStrategic Report
- ASEAN Secretary-General delivers official remarks on ASEAN AI Governance Framework at WAIC High-Level Meeting on Global AI GovernanceRegulatory
- Capital One open-sources VulnHunter — agentic AI vulnerability-detection toolSolutions
- Perplexity launches SPACE — secure sandbox platform for long-running AI agentsSolutions
- Google Cloud unveils "Agentic Defense" / AI Threat Defense platform (Gemini + Wiz + Mandiant + SecOps agents)Solutions
- Palo Alto Networks — Prisma AIRS AI Gateway reaches General AvailabilitySolutions
2026-07-19
- IBM Engineering AI Hub Cross-Site Scripting via Improper Web Page Input NeutralizationVulnerability
- Aimogen Pro AI Content Writer WordPress Plugin Privilege Escalation via Missing Capability CheckVulnerability
- IBM Cognos Analytics Agentic AI Assistant Race Condition Causes Incorrect Report ResultsVulnerability
- MiniCode MCP Command Injection via child_process.spawnVulnerability
- SafestClaw Shell Action Incomplete Blacklist Enables Command BypassVulnerability
- Sipeed PicoClaw Time-of-Check Time-of-Use Flaw in Tool ExecutionVulnerability
- poco-ai poco-claw SSRF via callback_url in Task ExecutorVulnerability
- OpenClaw Forwards Authorization Headers During MCP SSE RedirectsVulnerability
- Akamai Demonstrates End-to-End Precision Prompt Injection Kill Chain Against Deployed Travel-Booking AI AgentVulnerability
- parisneo/lollms Stored XSS via Shared Prompt Content in Direct MessagesVulnerability
- OpenClaw ClickClack Agent-Mode Dispatch Authorization Bypass Ignores toolsAllow PolicyVulnerability
- ForgeCode AI Pair-Programming CLI Auto-Executes Untrusted .mcp.json MCP Server DefinitionsVulnerability
- WordPress AI Copilot Plugin OAuth Token Binding Failure Enables Unauthenticated Admin Takeover of MCP ToolsVulnerability
- IBM Langflow OSS RCE via MCP Server Configuration Validator Bypass in File Upload APIVulnerability
- The 2026 Singapore Consensus on Global AI Safety Research PrioritiesStrategic Report
- EDPB adopts draft Guidelines 03/2026 on web scraping for generative AI training and new anonymisation guidelinesRegulatory
- NIST CAISI publishes capability/cyber-risk assessment of Z.ai's GLM-5.2 open-weight modelGuidelines
- ISO/IEC DIS 27091 — AI privacy protection guidance — in enquiry/FDIS-registration phaseGuidelines
- ISO/IEC 27090 reaches FDIS (Final Draft International Standard) stage — AI security threats guidanceGuidelines
- Lineation.ai publicly launches Zero Trust runtime security control plane for AI agentsSolutions
- Polygraf AI launches Meeting Guard — real-time deepfake/fraud detection for enterprise meetingsSolutions
- SnapLogic ships SnapCode and SnapLogic MCP Server for governed AI-coding-agent integrationSolutions
- Entrust launches Agentic AI Trust AcceleratorSolutions
- Alterion launches Draco — runtime control plane for enterprise AI agentsSolutions
2026-07-18
- Australia Announces Mandatory National AI Standards and New Office of AIRegulatory
- Bipartisan House Bill: Human Authority Over Autonomous Weapons ActRegulatory
- 29 Countries Sign Agreement Establishing World AI Cooperation Organization (WAICO)Regulatory
- Agentic-Flow MCP Server OS Command Injection (CVE-2026-58195)Vulnerability
- OpenClaw Hooks allowedAgentIds Bypass via Blank Agent ID (CVE-2026-62219)Vulnerability
- OpenClaw Agent-Mode Dispatch Authorization Bypass Ignores Tool Allowlist Policy (CVE-2026-62209)Vulnerability
- IBM Langflow OSS Unauthenticated RCE via Incomplete Denylist on Public Flow Build Endpoint (CVE-2026-13448)Vulnerability
- IBM Langflow OSS ToolGuard Policy Bypass Enables Code Injection (CVE-2026-9135)Vulnerability
- Lineation.ai — Public Launch of Zero Trust Runtime Security Control Plane for Autonomous AI AgentsSolutions
- Polygraf AI Meeting Guard — Real-Time Deepfake/Fraud Detection for Enterprise MeetingsSolutions
- Nudge Security — Agentic OAuth Grant and Browser Extension Risk Analyzer AgentsSolutions
- Cloudflare Precursor — GA of Continuous Behavioral Bot/Agent Detection EngineSolutions
- 1Password for Claude — Zero-Exposure Credential Injection with Agentic ModeSolutions
2026-07-17
- AWS Bedrock AgentCore Python SDK — OpenTelemetry Instrumentation Logs Sensitive User ContentVulnerability
- PraisonAI — Call API Agent Invocation Authentication Bypass When Auth Disabled Flag MisconfiguredVulnerability
- Lightpanda Headless Browser — Credentials Attached to Cross-Origin Requests Regardless of credentials SettingVulnerability
- Lightpanda Headless Browser — Origin Confusion via Malformed URL Authority ParsingVulnerability
- Repomix — Git Option Injection via Unvalidated --remote-branch ParameterVulnerability
- 9Router — PATCH /api/settings Missing Field Whitelist Allows Disabling Authentication Application-WideVulnerability
- OpenAI GPT-Red Automated Red-Teamer Demonstrates Live Prompt-Injection Compromise of Deployed Agentic Vending-Machine SystemVulnerability
- Claude Code Action (GitHub Action) — Attacker-Controlled PR Content Enables MCP Config Injection and Code ExecutionVulnerability
- SimpleChat AI Conversation App — Plugin Validation Endpoints Missing AuthenticationVulnerability
- Prompty (.prompty LLM Prompt Format) — Executable Frontmatter Enables Code Execution via gray-matterVulnerability
- Kiota Code Generator — Plugin Manifest Injection Enables Malicious Microsoft 365 Copilot/Teams PluginsVulnerability
- MCP Python SDK — Deprecated WebSocket Transport Lacks Host/Origin ValidationVulnerability
- MCP Python SDK — Task Handlers Operate Without Session/Ownership Validation (tasks/list, tasks/get, tasks/result, tasks/cancel)Vulnerability
- Grafana MCP Server — Confused-Deputy SSRF and Service-Account Token Exfiltration via X-Grafana-URL HeaderVulnerability
- n8n-MCP — Cross-Tenant Access to Workflow Version Backups in Multi-Tenant HTTP ModeVulnerability
- NocoBase AI-Powered No-Code Platform — Unauthenticated Time-Based SQL InjectionVulnerability
- 9Router — Hardcoded Fallback JWT Secret Enables Auth Token ForgeryVulnerability
- Cursor Cloud Agent — Unauthenticated Local Agent Endpoint Enables Sandbox RCE and Credential Theft via Browser-Enabled SessionsVulnerability
- LiteLLM MCP Server Creation — Command Injection via Unvalidated JSON Config (STDIO Transport)Vulnerability
- 9Router AI Router — Unauthenticated MCP Bridge and CLI-Tools API Enables Arbitrary Command ExecutionVulnerability
- Hugging Face Disclosed Autonomous AI-Agent-Driven Intrusion into Dataset Processing PipelineVulnerability
- Australia establishes national Office of AI and announces mandatory national AI Standards frameworkRegulatory
- China's Implementation Opinions on Intelligent Agents take effect — first dedicated AI agent regulatory frameworkRegulatory
- Google Cloud open-sources k8s-aibom for automated, unprivileged AI Bill-of-Materials generation on GKESolutions
- Red Hat details layered agent sandboxing combining OpenShift Sandboxed Containers (Kata) with NVIDIA OpenShellSolutions
- Artificial Intelligence and the Risk of Strategic CyberattackStrategic Report
- Artificial Intelligence and the Future of African MiningStrategic Report
- The Commercial Landscape of AI Sovereignty OfferingsStrategic Report
2026-07-16
- Claude Desktop "PromptFiction" — Zero-Click Prompt Injection via claude:// URI SchemeVulnerability
- AI Office publishes frontier AI expert findings on EU competitiveness, sovereignty and securityStrategic Report
- CREST Launches AI Charter with Nine Principles for AI-Enabled Cybersecurity ServicesGuidelines
- Singapore PDPC Proposes Mandatory AI-Specific Data Notifications for Generative AI TrainingRegulatory
- Malaysia National AI Office Opens Public Consultation on AI Governance BillRegulatory
- White House Launches 'Gold Eagle' AI Cybersecurity Vulnerability ClearinghouseRegulatory
- OpenAI GPT-Red — Automated Adversarial Self-Play Red-Teamer for Prompt InjectionSolutions
- Sophos Fusion — AI-Native Cybersecurity Defense SystemSolutions
2026-07-15
- NVIDIA TensorRT-LLM — Multiple Deserialization, IPC, and Orchestrator VulnerabilitiesVulnerability
- Cursor IDE Auto-Executes Malicious git.exe from Poisoned Repositories (Unpatched 0-day)Vulnerability
- GPT-5.6 System CardStrategic Report
- The AI investment raceStrategic Report
- Malaysia's National AI Office Releases Public Consultation Paper on AI Governance BillRegulatory
- New York Governor Hochul Signs First-in-Nation Statewide Data Center Moratorium Executive Order (No. 62)Regulatory
- Fortinet Expands FortiEndpoint with AI Visibility, Native DLP for AI, and FortiAI-AssistSolutions
- Ant Group Open-Sources SingGuard-NSFA — Purpose-Built Guardrail Model for Autonomous AI AgentsSolutions
- Varonis Atlas Extends AI-SPM Coverage to Claude Code and Claude CoworkSolutions
- Checkmarx Unveils Self-Healing Application Security — Autonomous Pre-Commit Vulnerability RemediationSolutions
- AWS WAF Bot Control Adds Cryptographic Web Bot Authentication for Legitimate AI AgentsSolutions
- AWS Security Hub Adds AI Workload Protection (GuardDuty AI Protection GA + AI Inventory) and Azure Multicloud SupportSolutions
2026-07-14
- Beyond Targeting: The Untapped Role of AI in Military Decision-MakingStrategic Report
- EU Action Plan on Cybersecurity and Artificial IntelligenceStrategic Report
- Illinois Signs Law Banning AI Use in Teacher Evaluations (SB 2909)Regulatory
- Instant Image Generator WordPress Plugin — SSRFVulnerability
- QuantumCloud WoowBot Chatbot for WooCommerce — Stored XSSVulnerability
- QuantumCloud ChatBot WordPress Plugin — Stored XSSVulnerability
- AIWU AI Copilot Content Generator WordPress Plugin — Blind SQL InjectionVulnerability
- Helicone AI Gateway SSRF via AWS Metadata Service Path HandlingVulnerability
- vLLM Orchestrator Gateway Logs Authorization Headers and Full Chat Payloads in PlaintextVulnerability
- mcp-gitlab Path Traversal Enables Redirection of GitLab API Requests to Arbitrary EndpointsVulnerability
- ServiceNow AI Platform Sandbox Escape Enables Unauthenticated Remote Code ExecutionVulnerability
- ITU Launches Focus Group on Trust and Identity for Humans and Agentic AI (FG-TIDA)Guidelines
- JetStream Security Launches Verified MCP Governance Layer (JetStream Verified MCP + AI Hub)Solutions
- Palo Alto Networks Prisma AIRS AI Red Teaming Adds Custom Target AdapterSolutions
- Radware Expands Agentic AI Protection with Claude Code Endpoint Coverage and Compliance ReportingSolutions
- Citrix NetScaler AI Gateway Adds MCP Gateway for Unified LLM/Agentic Traffic GovernanceSolutions
- Codenotary Launches AgentMon 3 — Adaptive Behavioral Runtime Security for AI AgentsSolutions
- First Recon AI Launches AI Security Runtime — GA with Inline Policy Enforcement and Audit-Ready EvidenceSolutions
2026-07-13
- Microsoft SFI July 2026 Report: Internal AI System Auto-Detects and Remediates Cloud Security RisksSolutions
- Palo Alto Networks & Tenzai: AI-vs-AI Purple-Team Testing Feeds Cortex XDR DetectionsSolutions
- Banking on Trust: AI Governance for Growth, Resilience and ScaleStrategic Report
- When AI Agents Attack: Autonomous Cyber Operations and Europe's Governance GapStrategic Report
- China's Interim Measures for AI Anthropomorphic Interaction Services Take Effect July 15, 2026 (AI companion/minor protections)Regulatory
- FTC Proposed Policy Statement on AI Output Manipulation/Accuracy as Deceptive Practice, Tied to State AI Law Preemption PushRegulatory
- BIS Final Rule: Enhanced Favorable Treatment for UAE Under Export Administration Regulations (AI chip access)Regulatory
- ENISA — "Mind the Gap in Standardisation of Cybersecurity for Artificial Intelligence"Guidelines
- Overthinking: Reasoning-Weight Amplification Extracts Hidden LLM Secrets Up to 10x More OftenVulnerability
- AstrBot MCP Test Endpoint SSRF via mcp_server_config.urlVulnerability
- Crawl4AI Docker API Credential Exfiltration via LLM Endpoint base_url/api_token ManipulationVulnerability
- PraisonAI AgentMail Webhook Signature Verification Bypass Enables Message SpoofingVulnerability
- PraisonAI Insecure Default Configuration Exposes Agent API Without AuthenticationVulnerability
- PraisonAI AICoder Arbitrary File Write and Command Execution via LLM Tool CallsVulnerability
- JADEPUFFER — First Documented Fully Autonomous Agentic Ransomware AttackVulnerability
2026-07-12
- PraisonAI CodeAgent — Unrestricted LLM-Generated Python Execution Enables RCEVulnerability
- Compromised LiteLLM Proxy for Amazon Bedrock Exploited for Cryptomining — AI Gateway Blast-Radius RiskVulnerability
- ENISA's View on Cybersecurity in the Frontier AI EraStrategic Report
- Italian Garante Fines Character.AI Owner €158,000 Over Age-Verification and Data Protection FailuresRegulatory
- SANS Institute and OWASP AI Exchange Fully Align AI Security Frameworks (MOSAIC Initiative)Guidelines
- UK DSIT/Lancaster University: Thematic Review and Gap Analysis on AI SecurityGuidelines
- Claude Code Adds Sandboxed Built-In Browser with Safety Classifiers for Agentic Web ActionsSolutions
- Radware Expands Agentic AI Protection with Claude Code Endpoint Coverage and Compliance ReportingSolutions
2026-07-11
- aerostack-mcp (mcp-whatsapp) — SSRF via upload_media media_url Argument (CVSS 6.3)Vulnerability
- mcp-text-editor — Path Traversal in File Path Validation (CVSS 6.3)Vulnerability
- PraisonAI — Repository-Controlled Config Enables Path Traversal in output_file (CVSS 5.5)Vulnerability
- PraisonAI — Prompt Injection Defense Bypass via Detector-Family Threshold Gaming (CVSS 5.3)Vulnerability
- MaxKB — Inconsistent MCP Transport Validation Enables Unauthorized Tool/Server Reference (CVSS 8.8)Vulnerability
- Open WebUI — Redis-Backed Token Revocation Bypass in Socket.IO/Terminal Authentication (CVSS 7.1)Vulnerability
- Open WebUI — Session Hijack Enables Unauthorized execute:python/execute:tool via Client-Supplied session_id (CVSS 7.7)Vulnerability
- Open WebUI — Terminal WebSocket Upstream URL Query Injection via Unencoded session_id (CVSS 8.0)Vulnerability
- Crawl4AI Docker API — SSRF via Unvalidated Webhook URLs in /crawl/job and /llm/job (CVSS 8.6)Vulnerability
- MCP Server Kubernetes — Argument Injection Bypasses Dangerous-Flag Protection in kubectl Tools (CVSS 9.8)Vulnerability
- Langroid SQLChatAgent — Regex Blocklist Bypass Enables Prompt-to-SQL-to-RCE (CVSS 8.7)Vulnerability
- 9Router — Unauthenticated Full Database Export/Import via /api/settings/database (CVSS 9.9)Vulnerability
- Hermes WebUI — Unauthenticated Remote Code Execution via Terminal API (CVSS 9.8)Vulnerability
- Hermes WebUI — Authentication Bypass via Spoofed X-Forwarded-For Enables SSRF and Config Takeover (CVSS 9.1)Vulnerability
- Langroid Neo4jChatAgent — Unvalidated LLM-Generated Cypher Enables Graph Data Destruction and Potential RCE (CVSS 9.2)Vulnerability
- Langroid TableChatAgent/VectorStore — Sandbox Escape to Remote Code Execution via Incomplete eval() Mitigation (CVSS 10.0)Vulnerability
- Ruflo Agent Meta-Harness — Unauthenticated MCP Bridge RCE (CVSS 10.0)Vulnerability
- HalluSquatting — Adversarial Hallucination Squatting Enables Botnet Distribution via AI Coding AssistantsVulnerability
- Sen. Warren Demands DoD and Seven AI Companies Disclose Military AI Contract Terms Over Surveillance/Autonomous Weapons ConcernsRegulatory
- Muse Spark 1.1 Evaluation ReportStrategic Report
- Zentera Systems Publishes Guide Synthesizing CISA/NCSC/Five Eyes Controls for AI Agent Access to OTGuidelines
- CIS + SAFECode Publish Secure by Design v1.1 with AI-Specific Software Security GuidanceGuidelines
- MAS Publishes SAFR (Safeguards for Agentic Finance at Runtime) Whitepaper for Financial AI AgentsGuidelines
- AWS MCP Server — OAuth Support via AWS Sign-In for Agent Authentication and GovernanceSolutions
- First Recon AI Security Runtime — GA Launch with Inline Policy Enforcement and Audit-Ready EvidenceSolutions
- BeyondTrust Pathfinder NHI Governance — Privileged Access Controls Extended to AI Agents and Non-Human IdentitiesSolutions
- Citrix NetScaler MCP Gateway — Unified Governance for LLM and Agentic AI TrafficSolutions
2026-07-10
- LiteLLM MCP Streamable HTTP Endpoint Authentication Bypass via Fabricated Authorization HeaderVulnerability
- ESET H1 2026 Threat Report: ~900,000 AI Agent Skills Scanned, Thousands Found MaliciousVulnerability
- Langflow IDOR (CVE-2026-55255) Actively Exploited for Credential Harvesting — Added to CISA KEVVulnerability
- GhostApproval: Symlink-Based Trust-Boundary Bypass Across Six Major AI Coding AssistantsVulnerability
- Friendly Fire: Prompt Injection Hijacks Claude Code and Codex CLI for RCE During Defensive Code ReviewVulnerability
- IETF Internet-Draft: Security Evaluation Benchmark for AI Agents (draft-han-bmwg-agent-security-benchmark-00)Guidelines
- The AI Resilience Gap: Bringing Artificial Intelligence Inside the Operational Resilience PerimeterStrategic Report
- High Bandwidth Memory: What It Is and Why It MattersStrategic Report
- News Publishers Seek Sanctions Against OpenAI in NYT/Daily News Copyright Litigation Over Evidence DestructionRegulatory
- Florida's OpenAI Consumer-Protection Lawsuit Removed to Federal Court, Assigned to Judge Aileen CannonRegulatory
- China MIIT/NVDB Issues Binding 'Backdoor' Security Directive on Anthropic's Claude CodeRegulatory
- SkillCloak / SkillDetonate — Scanner Evasion and Dynamic Behavioral Detection for Malicious AI Agent SkillsSolutions
- ModelCop Launches Commercial AI Agent / Non-Human Identity Security PlatformSolutions
- BlueVoyant Launches Microsoft Agent 365 Security Deployment ServiceSolutions
- NetSPI Expands Continuous Pentesting Platform with Continuous AI Penetration Testing + AI-Validated FindingsSolutions
- OpenAI Bio Bug Bounty Program — Ongoing Private Universal-Jailbreak Testing for BioriskSolutions
2026-07-09
- 9Router OS Command Injection via Unauthenticated Tailscale-Install EndpointVulnerability
- Cognee Improper Access Control — Unauthenticated LLM Provider Configuration OverwriteVulnerability
- mem0 Unauthenticated Config API — Plaintext LLM API Key Exposure and SSRF via ollama_base_urlVulnerability
- Rogue Agent — Google Dialogflow CX Code Blocks Cross-Agent Code Injection Enabling Chatbot HijackingVulnerability
- Langflow Authorization Bypass (IDOR) — Added to CISA KEV, Actively Exploited for Cross-Tenant Credential HarvestingVulnerability
- AI Safety Index — Summer 2026Strategic Report
- US Commerce Dept/BIS Clears OpenAI GPT-5.6 for Broad Public ReleaseRegulatory
- AWS Publishes Guidance: Designing for System Prompt Leakage in GenAI ApplicationsSolutions
- Codenotary Launches AgentMon 3 — Behavioral Runtime Security for AI AgentsSolutions
- Keeper Security Extends Agentic AI Governance to Endpoint Privilege ManagerSolutions
- CrowdStrike Expands Prompt Injection Taxonomy to 200+ TechniquesSolutions
2026-07-08
- 9Router — Unauthenticated OS Command Injection via sudoPassword in Tailscale-Install EndpointVulnerability
- GitLost: Prompt Injection in GitHub Agentic Workflows Leaks Private Repository Data via Public IssuesVulnerability
- The Transatlantic Artificial Intelligence Calculus: U.S.–European Union and Member State AI Cooperation in an Era of Strategic CompetitionStrategic Report
- The Mills Review: AI and the future of retail financial servicesStrategic Report
- European Commission Action Plan on Cybersecurity and AI — Pre-Market Frontier Model Evaluation CapacityRegulatory
- European Systemic Risk Board (ESRB) Formal Warning on Systemic Cyber Risk from Frontier AI ModelsRegulatory
- ECB Letter to Eurozone Significant Banks: Mandatory AI-Enabled Cyber Threat Action Plans by October 31, 2026Regulatory
- ENISA's View on Cybersecurity in the Frontier AI Era — New EU Research/Guidance ReportGuidelines
- NCSC 'Cyber Shield' — UK National Initiative for Agentic AI Cyber DefenceGuidelines
- Assail Launches Sidewinder — Self-Healing Autonomous Offensive Security PlatformSolutions
- Chainguard Expands Athena Coalition — 40,000 Vulnerabilities Processed, New Members AddedSolutions
- CyberProof Launches Agentic MXDR ServiceSolutions
- Qualys Joins Cisco Cloud Control Studio as Launch Partner for Agentic Risk IntelligenceSolutions
- AWS: Cedar-Based Least-Privilege Authorization for Multi-Agent AI ChainsSolutions
- AWS Bedrock: Zero Data Retention Enforcement via Bedrock Projects + Service Control PoliciesSolutions
2026-07-07
- markdownify-mcp — Symlink Following in assertPathAllowed Path ValidationVulnerability
- LangGraph Task Result Cache — Weak Hash in Cache Key GenerationVulnerability
- BettaFish InsightEngine — Partial String Comparison in Search-Result DeduplicationVulnerability
- AIAnytime Awesome-MCP-Server (mcp-wiki) — SSRF via Wiki-Summary URL ArgumentVulnerability
- NousResearch hermes-agent — Path Traversal in skill_view FunctionVulnerability
- vLLM — Unbounded Audio-Upload Memory Materialization Enables DoSVulnerability
- Prompt Injection Campaigns Trick Browsing AI Agents Into Sending Cryptocurrency PaymentsVulnerability
- OpenAI Codex Desktop (macOS) — Indirect Prompt Injection via Markdown Image Rendering Enables Data ExfiltrationVulnerability
- SUSE Rancher AI Agent — Debug-Level Logging Leaks API Keys and LLM Response ContentVulnerability
- Amazon mcp-gateway-registry — SQL Injection in Metrics-Service Retention Policy ManagementVulnerability
- vLLM — Prompt-Embeds Payload with M-RoPE Model Crashes EngineCoreVulnerability
- vLLM — Speculative Decoding Rejection-Sampler Boundary CrashVulnerability
- vLLM — Uncompiled-Timeout Regex DoS via structured_outputs.regex ParameterVulnerability
- Crawl4AI Docker API — SSRF Bypass on Streaming Crawl EndpointVulnerability
- Crawl4AI Docker API — Unconfined Path Traversal in Downloaded-File NamingVulnerability
- Crawl4AI Docker API — Chromium Launch-Argument Injection Enables Remote Code ExecutionVulnerability
- Winning the AI Pentathlon Requires Endurance: The Case for Resilience and Risk Management as Core Elements of Competitive National AI StrategyStrategic Report
- Google Cloud launches Claude Apps Gateway for enterprise governance of Claude CodeSolutions
- Inaugural UN Global Dialogue on AI Governance Opens in GenevaRegulatory
- UK FCA Publishes 'The Mills Review' — Landmark Global Regulatory Review on AI in Retail Financial ServicesRegulatory
- Illinois Artificial Intelligence Safety Measures Act (SB 315) Signed Into LawRegulatory
2026-07-06
- Safeguards for Agentic Finance at Runtime (SAFR)Strategic Report
- Bank of England Deputy Governor Sarah Breeden — 'Agents of Change' Speech Warning on Agentic AI Financial Stability RiskGuidelines
- LangGraph — Weak Hash in Task Result Cache Key GenerationVulnerability
- markdownify-mcp — Insufficiently Random Temp File Names in saveToTempFileVulnerability
- markdownify-mcp — Symlink Following in assertPathAllowed Path ValidationVulnerability
- BettaFish InsightEngine — Partial String Comparison in Search-Result DeduplicationVulnerability
- AIAnytime Awesome-MCP-Server (mcp-wiki) — Server-Side Request Forgery via url ArgumentVulnerability
- Microsoft 365 Copilot — Open Redirect Enables Unauthorized Privilege EscalationVulnerability
- FBI FLASH: TeamPCP Large-Scale Software Supply Chain Attacks Trojanize Developer/Security Tools (Trivy, KICS, LiteLLM, Telnyx SDK) — Includes Self-Propagating npm/PyPI WormsVulnerability
- AI Agent Poisoning via SEO Poisoning and Hidden HTML Prompt Injection — Agents Tricked into Fraudulent Crypto PaymentsVulnerability
- HiddenLayer Joins Databricks Unity AI Gateway EcosystemSolutions
- Malaysia National AI Office Holds Public-Sector Consultation on Draft AI Governance BillRegulatory
- Thailand ETDA Releases Revised Draft AI Act for Public Consultation — Risk-Based, Extraterritorial, Strict LiabilityRegulatory
- Vietnam Deputy PM Signs Decision 33/2026/QD-TTg — Official List of 46 High-Risk AI Systems Across 6 SectorsRegulatory
2026-07-05
- OWASP Agentic Skills Top 10 — new risk taxonomy for AI agent 'skills' (AST01–AST10)Guidelines
- OWASP GenAI Security Project — GenAI Red Teaming Manual v1.0 released for public commentGuidelines
- ForceInjection AI-fundermentals LangChain Memory Recall Handler — Use of Weak/Predictable ValueVulnerability
- NousResearch hermes-agent — Streaming Reasoning Tag Filter Case-Sensitivity Bypass (Info Disclosure)Vulnerability
- NousResearch hermes-agent — HTTP API Denial of Service via 'todos' ParameterVulnerability
- NousResearch hermes-agent — Path Traversal in Live Webhook Media ExtractionVulnerability
- NousResearch hermes-agent — Discord Platform Adapter Authentication BypassVulnerability
- NousResearch hermes-agent — TUI Gateway shell.exec Protection Mechanism Failure Enables Remote Command ExecutionVulnerability
- Trail of Bits fickling — MLAllowlist Safety Check Silently Bypassed, Malicious Pickles Marked 'LIKELY_SAFE'Vulnerability
- SnapLogic MCP Builder (GA) — Auto-Generates Governed MCP Servers from Existing IntegrationsSolutions
- Jamf AI Governance (GA) — Native macOS Control Plane for Shadow AI Discovery, Policy, and Okta Agent Identity IntegrationSolutions
- iboss AI Security Platform — Free Shadow AI Discovery + Paid Runtime Policy ControlsSolutions
- Netzilo AIDR — Portable Runtime Governance for AI Agents Across Bedrock AgentCore, Copilot Studio, Vertex AI, LangGraph, CrewAISolutions
2026-07-04
- OpenAI WordPress Chatbot Helper Plugin — Unauthenticated Arbitrary Content Deletion (CVE-2025-69134)Vulnerability
- weDocs WordPress AI Chatbot Plugin — Missing Authorization on Migration Function (CVE-2026-12729)Vulnerability
- weDocs WordPress AI Chatbot Plugin — Stored XSS via Block Attributes (CVE-2026-12731 / CVE-2026-12734)Vulnerability
- LobeChat — Broken Object-Level Authorization Exposes and Mutates Other Users' Agent Group Data (CVE-2026-59100)Vulnerability
- fast-mcp-telegram — Bearer Token Path Traversal Bypasses Session Authentication (CVE-2026-52830)Vulnerability
- Kong Konnect MCP Server — Indirect Prompt Injection Enables Unintended API Request Execution (CVE-2026-13341)Vulnerability
- Cursor IDE 'DuneSlide' — Zero-Click Prompt Injection Escapes Sandbox for OS-Level RCE (CVE-2026-50548 / CVE-2026-50549)Vulnerability
- JadePuffer: First Confirmed End-to-End AI-Agent Ransomware Attack via Langflow RCE (CVE-2025-3248)Vulnerability
- Connecticut SB 5 AI Act — Frontier Model Developer Regulation, Sandbox and Youth Safety Provisions — Active Compliance WatchRegulatory
- UN Independent Scientific Panel on AI Releases Preliminary Report Warning of Catastrophic Harm and Regulatory GapRegulatory
- Singapore MDDI/IMDA Launch Public Consultation on Draft Digital Infrastructure Bill — Licensing, Security, and Sustainability for Data Centres and Cloud ProvidersRegulatory
- Supreme Court Overturns Humphrey's Executor, Ending FTC Independence — Major AI Regulatory Governance ShiftRegulatory
- FTC Proposes Policy Statement: AI Output Manipulation to Comply with State Laws May Violate FTC Act Section 5Regulatory
- US Commerce Dept. Lifts Full Export Controls on Anthropic Claude Fable 5 & Mythos 5 with Binding CommitmentsRegulatory
- Exabeam Behaviour Intelligence: Doubles AI/Agent Detections to 90, Adds Claude Support, Launches Observra OSS Telemetry LibrarySolutions
- Nightfall AI: AI-Native DLP + MCP Security Platform with Prompt Injection Detection for Agentic WorkflowsSolutions
- NVIDIA NeMo Guardrails v0.23.0 — IORails Tool Calling for Streaming + Non-Streaming, Local Rail ValidationSolutions
- Snyk Evo Agentic Development Security (ADS) — GA: Runtime Governance for AI Coding Agents and MCP ServersSolutions
2026-07-03
- ChatBot WordPress Plugin — Unauthenticated Reflected XSS (CVE-2026-57362)Vulnerability
- OpenAI Chatbot Helper WordPress Plugin — Unauthenticated Arbitrary Content Deletion (CVE-2025-69134)Vulnerability
- WPBot AI ChatBot for WordPress — Authenticated Stored XSS via Conversation Parameter (CVE-2026-13731)Vulnerability
- LobeChat — Broken Object-Level Authorization Allows Authenticated Users to Access and Modify Other Users' Agent Group Data (CVE-2026-59100)Vulnerability
- RAGFlow — Stored XSS via Unsanitised Agent Pipeline Node Name in Dataflow Result UI (CVE-2026-58579)Vulnerability
- fast-mcp-telegram MCP Server — Path Traversal in Bearer Token Validation Enables Authentication Bypass (CVE-2026-52830)Vulnerability
- DuneSlide — Cursor IDE Zero-Click Prompt Injection Sandbox Escape Enables Host RCE (CVE-2026-50548, CVE-2026-50549)Vulnerability
- Meaningful Outcomes Determine the Winners of the Health AI RaceStrategic Report
- Accenture AI Progress Barometer (Inaugural Edition)Strategic Report
- Artificial Intelligence and Cybersecurity in the Financial SectorStrategic Report
- System Card: Claude Sonnet 5Strategic Report
- EO 14409 AI Cybersecurity Clearinghouse — July 2 Establishment Deadline TriggeredRegulatory
- China MIIT/SAMR: Mandatory National Standard for L2 Combined Driver-Assistance Systems — Effective 1 January 2027Regulatory
- UN Independent International Scientific Panel on AI: Preliminary Report Launched — Feeds Inaugural UN Global Dialogue on AI Governance (Geneva, 6–7 July 2026)Regulatory
- FTC Proposed Policy Statement: Suppression of Accuracy in AI Systems — Public Comment OpenRegulatory
- US Commerce/BIS Lifts Export Controls on Anthropic Claude Fable 5 and Mythos 5 — New Binding Conditions ImposedRegulatory
- China TC260 — Draft Amendment to National Personal Information Security Specification Targeting AI and Deep Synthesis TechnologiesGuidelines
- Singapore CSA — Singapore Cyber Landscape 2025/2026 Report: Agentic AI Threat Analysis + Cyber Trust Mark Mandate for CII OperatorsGuidelines
- AWS Network Firewall — Container Attribute-Based Rules for EKS/ECS AI/ML Workload Microsegmentation (GA)Solutions
- CSA Extends STAR Registry to Include AIUC-1 Certification for Agentic AI — Independent Quarterly Red-Teaming RequiredSolutions
- Snyk Evo Agentic Development Security (ADS) — Runtime Governance for AI Coding Agents, MCP Servers, and Generated CodeSolutions
- Straiker — $64M Series A + STAR Labs Agentic Exploit Dataset; GA Platform: Agent Discovery, Pre-Deployment Red-Teaming, Runtime ProtectionSolutions
- Microsoft Defender for Endpoint: AI Agent Runtime Protection (Public Preview) — Auto-Discovery of 25+ Local Agent Types + Prompt Injection BlockingSolutions
2026-07-02
- Royal MCP WordPress Plugin — Low-Privilege Users Can Invoke Unrestricted MCP Tools Including Private Content and User EnumerationVulnerability
- Vibe-Trading — MCP Swarm Run Directory Path Traversal Allows Arbitrary Run Metadata ReadVulnerability
- Microsoft SharePoint Server — Deserialization RCE Added to CISA KEV (CVE-2026-45659)Vulnerability
- LiteLLM AI Gateway — Three-CVE RCE Chain: Default Internal User Can Escalate to Admin and Execute Arbitrary CodeVulnerability
- Presenton MCP Server — Unauthenticated MCP Endpoint in Docker/Server Deployments Allows Session Token Minting and API Key AbuseVulnerability
- Woodpecker CI — GitLab Forge Pipeline Approval Bypass via Spoofable Commit Author Name Enables CI Secret ExfiltrationVulnerability
- DeepTutor — MCP Tool Authorization Bypass Allows Low-Privilege Users to Invoke Any Configured MCP ToolVulnerability
- GuardFall — Shell-Interpretation Bypass Exposes 10 of 11 Open-Source AI Coding Agents to Supply Chain AttacksVulnerability
- IBM Langflow OSS — Weak Cryptographic Key Derivation Exposes All Stored Credentials (LLM API Keys, DB Passwords)Vulnerability
- KPMG Global AI Pulse Q2 2026: From Deployment to Value RealisationStrategic Report
- Insights from table-top exercises in Europe on AI safety and cyber misuseStrategic Report
- Preliminary Report of the Independent International Scientific Panel on AI: Evidence-based assessment of opportunities, risks and impacts of artificial intelligenceStrategic Report
- US Senate — AI AGENT Act Draft Bill: FTC Registry for AI Agent Providers, Identity Linking, and Platform InteroperabilityRegulatory
- China — Three AI Regulatory Instruments Take Effect July 1, 2026: AI Ethics Standard, AI Agent Measures, and Anthropomorphic AI RulesRegulatory
- Vietnam — 2025 Law on Cybersecurity Enters into Force (July 1, 2026) with AI Deepfake and Platform ObligationsRegulatory
- FTC Proposes Policy Statement Warning AI Companies Against Manipulating Outputs to Comply with State AI LawsRegulatory
- US Commerce Dept. Lifts Export Controls on Anthropic Claude Fable 5 & Mythos 5 — Binding Commitments ImposedRegulatory
- Reserve Bank of India — Draft Guidance on Regulatory Principles for Model Risk Management 2026 (AI/ML Kill-Switch and Board Oversight)Guidelines
- EU AI Act — Omnibus VII Amendments Formally Adopted by Council (29 Jun 2026): High-Risk Deadlines Deferred, New Prohibitions AddedGuidelines
- NIST NCCoE — 'Cyber AI Profile' and AI Agent Identity & Authorization Projects AcceleratedSolutions
- NVIDIA NeMo Guardrails v0.23.0 — IORails Tool Calling for Streaming + Non-Streaming, Local Rail Validation of Tool ResultsSolutions
- Dragos EmberAI — OT-Native AI Assistant on Intelligence Fabric (ICS/OT Cybersecurity)Solutions
- Snyk Evo Agentic Development Security (ADS) — Runtime Governance for AI Coding Agents and MCP ServersSolutions
- F5 AI Security Platform GA + SurePath AI Acquisition — Network-Level AI Discovery, Shadow AI Detection, Runtime GuardrailsSolutions
- Microsoft Security Copilot + Azure AI Content Safety: MCP Tool Poisoning Guidance & Prompt Shields for Agentic AISolutions
2026-07-01
- Vibe-Trading — DNS Rebinding Authentication Bypass Allows Remote Bearer-Token Bypass (CVE-2026-58169)Vulnerability
- Vibe-Trading Agentic Trading Platform — Path Traversal in Proposal File Loading Enables Arbitrary File Read (CVE-2026-58170)Vulnerability
- Claude Code /copy Command — Insecure Temporary File Enables Response Disclosure and Symlink File-Write (CVE-2026-46406)Vulnerability
- University of Washington Study: Agentic AI Browsers Allow Same-Origin Policy Bypass via Prompt Injection, Working PoC DemonstratedVulnerability
- GuardFall — Bash Shell-Trick Bypass in 10/11 Open-Source AI Coding Agents Enables Supply-Chain Code InjectionVulnerability
- SimpleHelp OIDC Authentication Bypass Exploited to Deliver Djinn Stealer Targeting AI/Cloud Developer Credentials (CVE-2026-48558)Vulnerability
- BIS Annual Economic Report 2026Strategic Report
- FCA Emerging Technology Horizon Scan 2026 — AI Governance Shift from Principles to Operational SupervisionRegulatory
- US Commerce Dept. Partially Lifts Export Controls on Anthropic Claude Mythos 5 — Fable 5 Restrictions Remain; Trusted-Partner Framework EstablishedRegulatory
- EU Council Formally Adopts AI Act Omnibus Simplification Regulation (Digital Omnibus on AI)Regulatory
- NIST Cybersecurity Insights — Verifiable Digital Credential Presentment (Blog Post #4 in VDC Series)Guidelines
- Ireland NCSC — 'Securing AI Adoption in the Public Sector' Cybersecurity Guidelines PublishedGuidelines
- LiteLLM Three-CVE RCE Chain (CVE-2026-47101, CVE-2026-47102, CVE-2026-40217): AI Gateway Supply Chain Attack VectorSolutions
- ServiceNow + Accenture: Joint Agentic AI Security Operations Platform Launch (June 29, 2026)Solutions
- Microsoft Intune Vulnerability Remediation Agent (Public Preview): Agentic CVE Triage for Endpoint Security CopilotSolutions
- Microsoft Incident Response: MCP Tool Poisoning Threat Intelligence + Detection/Containment Guidance for Enterprise AI AgentsSolutions
- Microsoft Defender for Endpoint: Discovery of 25+ Local AI Agent Types + MCP Server Runtime Protection Against Prompt InjectionSolutions
- Microsoft Sentinel June 2026: ASIM AI Agent Events Schema GA + Agent Identities Connector (Preview) + MCP Graph Tools (Preview)Solutions
2026-06-30
- Prompt Injection Now Confirmed in Production AI Deployments — Three Enterprise Breaches Disclosed (June 2026)Vulnerability
- Agentic Red-Team Tools (12 Systems) — Systemic Sandbox Escape and API Key Exfiltration via Agent-Phishing (arXiv 2606.24496)Vulnerability
- CherryStudio MCP OAuth Local Callback Server — Authorization Bypass via Code Parameter ManipulationVulnerability
- Khoj AI — Conversation Sharing Handler Authorization Bypass Exposes Private Agent DataVulnerability
- Devolutions PowerShell Universal — AI Agent Job API Leaks Reusable App Tokens in PlaintextVulnerability
- SimpleHelp — OIDC Authentication Bypass Allows Unauthenticated Account Takeover (CISA KEV)Vulnerability
- Google MCP Toolbox — Path Traversal in HTTP Tool URL Builder Allows SSRF Against Upstream APIsVulnerability
- Claude Code — Sandbox Escape via Git Worktree Path Confusion Enables Host Code ExecutionVulnerability
- Asia's Human-led AI Opportunity: A Framework for TransformationStrategic Report
- Philippines — BSP Issues AI Governance Framework for Banks and Non-Banks Ahead of June 30 DeadlineRegulatory
- Australia DTA — Mandatory AI Requirements Now in Force for Commonwealth Agencies; Agentic AI Addendum PublishedRegulatory
- GovAI Research Report: GDPR — Not EU AI Act — Is Primary Cause of Frontier AI Delays in EU/UK (Published Jun 25–28, 2026)Regulatory
- Connecticut SB 5 — Frontier Model Developer Regulation, Regulatory Sandbox, and Chatbot/Youth Safety Rules (Signed May 29, 2026)Regulatory
- Great American AI Act (GAAIA) Discussion Draft Released — Federal Preemption of State AI Laws, Frontier Model OversightRegulatory
- Trump Executive Order 14409 — Mandatory Pre-Release Cyber Risk Assessments for Frontier AI Models (Signed Jun 2, 2026)Regulatory
- OpenAI Daybreak Expansion — GPT-5.5-Cyber GA + Patch the Planet Open-Source Security InitiativeSolutions
- Microsoft Threat Intel — Malicious Chromium Extension Spoofs Perplexity AI Branding to Hijack Browser SearchSolutions
- Straiker — $64M Series A + STAR Labs Agentic Exploit Dataset; Agent Discovery, Red-Teaming, Runtime Protection PlatformSolutions
- Okta Cross App Access (XAA) Ecosystem — 25+ Partners Including Claude, Cursor, Slack; MCP Official Auth Extension Goes StableSolutions
- Cisco Acquires Astrix Security + WideField Security — Dual NHI/Agent Identity Play into Splunk Agentic SOCSolutions
- Cisco AI Defense — Agent Runtime Protection SDK Launched for LLM and MCP InspectionSolutions
- OpenAI GPT-5.6 Sol Preview — Most Capable Cybersecurity Model with Restricted Government-Gated AccessSolutions
2026-06-29
- xiaozhi-esp32 MCP Response Handler — Improper Synchronization in ParseMessage Enables Remote DoS (CVE-2026-13489)Vulnerability
- ComfyUI-Copilot — Improper Resource Identifier Control in Workflow Checkpoint Restore Handler (CVE-2026-13493)Vulnerability
- Flowise Custom MCP — Windows NODE_OPTIONS Denylist Bypass Enables Authenticated RCE via Case-Sensitivity Gap (CVE-2026-58057)Vulnerability
- TrustFall — Trust Dialog Regression Enables One-Click RCE in Claude Code, Cursor, Gemini CLI, and GitHub CopilotVulnerability
- Amazon Q Developer — Malicious MCP Config in Repo Enables Arbitrary Code Execution and AWS Credential Theft on Workspace Open (CVE-2026-12957)Vulnerability
- North Korean Sapphire Sleet Mastra npm Supply Chain Attack — 144 AI Agent Framework Packages Trojanized, ~8M Weekly Downloads ExposedVulnerability
- Agentjacking — MCP Sentry Indirect Prompt Injection Achieves Arbitrary Code Execution in AI Coding Agents (85% Success Rate, 2,388 Orgs Exposed)Vulnerability
- OpenAI GPT-5.6 Sol/Terra/Luna — Government-Gated Limited Preview with Cybersecurity CapabilitiesSolutions
- Noma Security + Kong — Runtime AI Security Embedded in API Gateway Traffic PathSolutions
- Linux Foundation Akrites — Industry-Wide Coordinated OSS Vulnerability Defense Against AI-Accelerated ThreatsSolutions
- Dragos EmberAI — OT-Native AI Assistant on Intelligence FabricSolutions
- Google Cloud VPC Service Controls — Agent Identity, MCP-Attribute Perimeter Rules, and Gemini Agent Platform IntegrationSolutions
- Snyk Evo Agentic Development Security (ADS) — Runtime Governance for AI Coding Agents and MCP ServersSolutions
- F5 AI Security Platform GA + SurePath AI AcquisitionSolutions
- OpenAI Daybreak Expansion: GPT-5.5-Cyber GA + Codex Security Plugin + Patch the Planet InitiativeSolutions
- Supply Chain, Energy, and AI Nexus: Evaluating AI Energy Supply Chain VulnerabilitiesStrategic Report
- Top 10 Emerging Technologies of 2026Strategic Report
- Reserve Bank of India Draft Guidance on Regulatory Principles for Model Risk Management 2026 — AI/ML Kill-Switch and Board Accountability MandatesGuidelines
- Legion LegalTech Corp v. US — First Federal Lawsuit Challenging BIS Export Control Directive on Anthropic AI ModelsRegulatory
- White House / ONCD — OpenAI GPT-5.6 Launch Restricted to Government-Vetted Partners (First Pre-Release Gating of a US AI Model)Regulatory
- US Commerce Dept — Partial Lift of Anthropic Mythos 5 Export Control: Selective Access for 100+ Trusted PartnersRegulatory
2026-06-28
- JetBrains YouTrack MCP — Project Settings Disclosure via MCP Endpoint (CVSS 3.1)Vulnerability
- Mattermost AI Bot Posts — Markdown Image Syntax Injection Exfiltrates Data via Rendering (CVSS 3.5)Vulnerability
- mise — Task-Include Files Bypass Trust Gate, Enabling Arbitrary Code Execution (CVSS 8.6)Vulnerability
- US House Appropriations Committee — FY2027 Bill Provision to Block HHS Funding for CMS WISeR AI Prior-Authorization ModelRegulatory
- NRC Proposed Rule: Modernizing Nuclear Facility Security & Fitness-for-Duty Requirements (EO 14300)Regulatory
- Legion LegalTech Lawsuit: US Export Control Directive on Anthropic Models Challenged as Exceeding Statutory AuthorityRegulatory
- US Government Gates OpenAI GPT-5.6 Sol Launch — Per-Customer Commerce/ONCD/OSTP Approval RequiredRegulatory
- US Commerce Dept. Partially Lifts Anthropic Mythos 5 Export Control — Trusted-Partner List CreatedRegulatory
- Sumsub MCP Integration: AI Agents Can Now Configure Full Compliance/AML WorkflowsSolutions
- OrcaRouter AI Threat Report 2026 + Agent Firewall and I/O Guardrails Made FreeSolutions
- Noma Security + Kong Partnership: Runtime AI Security Embedded in Kong AI GatewaySolutions
- Tanium Atlas GA: Agentic AI Autonomous IT/Security Operations for All Commercial Cloud CustomersSolutions
- Dragos EmberAI: OT-Native AI Assistant Launched on Dragos Intelligence FabricSolutions
- Fortinet FortiSOC GA: Unified Agentic AI SOC Platform (SIEM + SOAR + UEBA + TI as single SaaS)Solutions
- F5 AI Security Platform GA + SurePath AI AcquisitionSolutions
2026-06-27
- ToolJet AI Platform — Credential Decryption Endpoint Returns Plaintext Secrets to Any Authenticated UserVulnerability
- Docling GenAI Document Processor — Unsafe URI/Path Handling in HTML BackendVulnerability
- Mattermost Agents MCP Plugin — SSRF via Unvalidated Attachment URLs Allows Internal Network AccessVulnerability
- GitHub MCP Server — RepoAccessCache Cross-Tenant Authorization Bypass in HTTP Lockdown ModeVulnerability
- ToolJet AI Platform — SSRF in RestAPI Data Source via Private IP Filter BypassVulnerability
- LibreChat — SSRF via User-Configured baseURL with No Private IP or Scheme ValidationVulnerability
- LibreChat — MCP OAuth Resource Parameter Mismatch Enables OAuth Token Hijack by Malicious MCP ServerVulnerability
- Ollama GGUF Quantization Engine — Remote Heap Memory Leak via Model Upload InterfaceVulnerability
- DifyTap — Four Cross-Tenant AI Data Exposure Flaws in Dify LLMOps Platform (1M+ Apps at Risk)Vulnerability
- ToolJet AI Agent Platform — Authenticated Builder Can Overwrite Global Marketplace Plugin with Arbitrary Server-Side JavaScript (RCE)Vulnerability
- mise Dev-Tool Manager — Arbitrary Command Execution via Tera Template Injection in .tool-versions (Trust Bypass)Vulnerability
- Cursor AI Editor — Agent Sandbox Escape via Canonicalization Fallback on Path Validation FailureVulnerability
- Cursor AI Editor — Agent Sandbox Escape via working_directory Parameter ManipulationVulnerability
- Flowise Custom MCP Feature — Unsandboxed RCE via Crafted JSON Payload (No Auth Required)Vulnerability
- The AI-First Operating System: A Blueprint for Operating and Business Model InnovationStrategic Report
- Anthropic Economic Index Report: CadencesStrategic Report
- Summary of METR's Predeployment Evaluation of GPT-5.6 SolStrategic Report
- US State AI Patchwork — New Companion Chatbot Laws and Connecticut Frontier-Model Sandbox Enacted in 2026 Legislative SessionsRegulatory
- China Premier Li Qiang — Summer Davos Call for Urgent Global AI Regulation Framework (Dalian, 24 June 2026)Regulatory
- Colorado Governor Signs SB 26-189 — Automated Decision-Making Technology (ADMT) Act Replaces Landmark 2024 AI ActRegulatory
- White House / ONCD + OSTP: Trump Administration Restricts GPT-5.6 Release to Government-Approved PartnersRegulatory
- NIST SP 800-213 Rev. 1 (Initial Public Draft) — IoT Product Cybersecurity Guidelines for the Federal GovernmentGuidelines
- BreachRx Rex Platform: Agentic AI Incident Command Center for Concurrent AI-Era BreachesSolutions
- NeuralTrust: Chain-of-Thought Hijacking Research — Up to 100% Safety Bypass on Frontier Reasoning ModelsSolutions
- Palo Alto Networks Prisma AIRS — June 2026: Configurable AI Red Teaming Target Profiling + VM-Series Unified AI Runtime Firewall ImageSolutions
- F5 AI Security Platform Launch + SurePath AI AcquisitionSolutions
- Linux Foundation Akrites: Industry-Wide Coordinated OSS Vulnerability Defense Against AI-Accelerated ThreatsSolutions
2026-06-26
- The Shift to Agentic AI: Evidence from CodexStrategic Report
- Human-Machine Collaboration in Industrial Operations: Activation PlaybookStrategic Report
- The AI Playbook for Financial ServicesStrategic Report
- PTC Windchill / FlexPLM Improper Input Validation — Unauthenticated RCE, CISA KEV Added Jun 25 2026 (CVE-2026-12569)Vulnerability
- Cisco Unified Communications Manager SSRF — Actively Exploited, CISA KEV Added Jun 25 2026 (CVE-2026-20230)Vulnerability
- Cordyceps CI/CD — GitHub Actions Workflow Misconfiguration Class Exposes Google AI Agent Dev Kit, Azure Sentinel & 300+ Repos to Supply-Chain TakeoverVulnerability
- Malicious AI Agent Skill Bypasses Static Scanners via Mutable External Link — Reached 26,000 Agents (AIR Research, Jun 24 2026)Vulnerability
- Prompt Injection as Role Confusion — ICML 2026 Paper Demonstrates Structural LLM Attack Class with Working PoCVulnerability
- GitLab Duo Workflows — Insufficient Output Filtering Discloses Sensitive Repository Data (CVE-2026-12053)Vulnerability
- Twenty CRM AI Agent Monitor — Cross-Workspace IDOR Exposes Full Chat Histories & Tool Outputs (CVE-2026-55583)Vulnerability
- US Congress Bipartisan Deadline: Commerce Must Justify EAR Authority Over AI Model Usage by June 26, 2026Regulatory
- Australia DTA Agentic AI Playbook for Government Published — Mandatory Oversight Models and Kill-Switch RequirementsRegulatory
- US Commerce/BIS EAR Directive: Anthropic Fable 5 & Mythos 5 Foreign-National Access Suspended — Litigation Filed, Congress Demands Response by June 26Regulatory
- Noma Security + Kong Partnership — Runtime AI Security Embedded in API Gateway Traffic PathSolutions
- F5 AI Security Platform GA + SurePath AI AcquisitionSolutions
2026-06-25
- NocoDB MCP Token — Unrestricted Attachment Read Across Workspaces (CVSS 2.3) and OAuth Scope Not Enforced (CVSS 2.0)Vulnerability
- AnythingLLM Windows Path Traversal — Encoded Absolute Path Escapes Documents Directory (CVSS 4.3)Vulnerability
- foreman-mcp-server — Session Hijack via Unauthenticated Cache Poisoning and Credential Logging (CVSS 7.8 + 6.2)Vulnerability
- rtk LLM Context Filter — Permission Splitter Bypass Enables Shell Command Injection Through Allowed-List Commands (CVSS 7.8)Vulnerability
- Daytona AI Sandbox — Multiple Authorization Flaws: Unverified Email Org Invite, Role Update IDOR, TLS Bypass in Git Clone (CVSS up to 8.4)Vulnerability
- Cordyceps — CI/CD Workflow Vulnerability Class Enables Supply Chain Hijacking via Pull Requests, Confirmed on Google AI ADK and Azure SentinelVulnerability
- BioShocking — Reality-Confusion Prompt Injection Bypasses AI Browser Guardrails, Leaks Credentials in Working PoCVulnerability
- macOS.Gaslight — DPRK Rust Infostealer Embeds Prompt Injection to Derail AI-Assisted Malware TriageVulnerability
- Claude Code WebFetch — HuggingFace Domain Pre-Approval Enables Prompt Injection via Attacker-Controlled Model Repos (CVSS 6.0)Vulnerability
- Open WebUI — Image URL Parameter Resolves as File ID, Enabling Cross-User File Exfiltration (CVSS 6.5)Vulnerability
- Docling GenAI Document Processor — SSRF via Playwright HTML Renderer, Zip Slip in EasyOCR, and XXE in Patent XML ParserVulnerability
- Warp Agentic Terminal — Multiple RCE and Command Injection Vulnerabilities Across Agent Features (CVSS up to 8.8)Vulnerability
- Crawl4AI Docker API — Proxy SSRF Bypass Reaches Internal Network (CVSS 8.6)Vulnerability
- AISI Engineering Playbook: An Open Guide to Building AI Evaluation CapabilitiesStrategic Report
- Looking Beyond the Government's Regulatory Toolkit: Early Findings on How Business and Civil Society Actors Can Help Manage and Mitigate Transformative Artificial Intelligence Risks and ImpactsStrategic Report
- US House Appropriations Committee — FY2027 Bill Provision to Block HHS Funding for CMS WISeR AI Prior-Authorization ModelRegulatory
- US Commerce/BIS EAR Directive: Anthropic Fable 5 & Mythos 5 Foreign-National Access Suspended — First Legal Challenge FiledRegulatory
- European Commission Draft Guidelines on Classification of High-Risk AI Systems under EU AI Act Article 6Guidelines
- NIST SP 800-213 Rev. 1 (Initial Public Draft) — IoT Product Cybersecurity Guidelines for the Federal GovernmentGuidelines
- Stellar Cyber Autonomous SOC Platform v6.5/6.6: AI-Native Workflows, Auto-Triage, and Faster MSSP OnboardingSolutions
- Tigera Lynx GA: Kubernetes-Native Unified Control Plane for AI Agent SecuritySolutions
- Tanium Atlas GA: Agentic AI Autonomous IT/Security Operations for All Commercial Cloud CustomersSolutions
- SentinelOne Purple AI Agentic Investigation Opens to All Singularity CustomersSolutions
- Dragos EmberAI: OT-Native AI Assistant Launched on Dragos Intelligence FabricSolutions
- Fortinet FortiSOC GA: Agentic AI-Unified SOC Platform (SIEM + SOAR + UEBA + TI in One SaaS)Solutions
- OpenAI Daybreak Expansion: GPT-5.5-Cyber GA + Codex Security Plugin + Patch the PlanetSolutions
2026-06-24
- Langflow Monitor Router IDOR — 7 Unauthenticated Endpoints Expose All Users' Session DataVulnerability
- vLLM Audio Transcription Endpoint Decompression Bomb (25 MB OPUS → 14.9 GB PCM)Vulnerability
- vLLM Revision Pinning Bypass — Dynamic Code Loaded Despite --revision/--code-revision FlagsVulnerability
- Langflow Unauthenticated Unbounded File Upload Causes Indefinite Denial of ServiceVulnerability
- Crawl4AI SSRF via IPv6-Mapped IPv4 Blocklist Bypass on LLM Scraping EndpointsVulnerability
- Open WebUI Cross-Origin Prompt Injection via Postmessage input:prompt / action:submitVulnerability
- LobeHub Server-Side SSRF via Unauthenticated /webapi/proxy EndpointVulnerability
- Claude Code Out-of-Band Data Exfiltration via Pre-Approved HuggingFace Domain in WebFetchVulnerability
- n8n MCP Browser Unauthenticated Tool Invocation in HTTP Transport ModeVulnerability
- Langflow RAG File Component Arbitrary File Read via Absolute Path in Knowledge BaseVulnerability
- Langflow IDOR — Authenticated User Can Execute Any Other User's Flow via /api/v1/responsesVulnerability
- Crawl4AI Sandbox Escape via AST Validator Bypass in Computed Fields (_safe_eval_expression)Vulnerability
- Flowise Custom MCP Server OS Command Injection via Incomplete Flag Validation and Regex BypassVulnerability
- vLLM Arbitrary Code Execution via Malicious HuggingFace Model — Assert Bypass in Optimized ModeVulnerability
- vLLM Dockerfile Dependency Confusion via Unregistered flashinfer-jit-cache Package on PyPIVulnerability
- vLLM OpenAI API Authentication Bypass via Host-Header Path Manipulation (BadHost variant)Vulnerability
- Langflow Unauthenticated RCE via Shareable Playground (Public Flows)Vulnerability
- Developing Vocational Education and Training with Artificial IntelligenceStrategic Report
- 2026 KPMG Global AI in Finance Report: The Decision AdvantageStrategic Report
- GSA Proposed Rule: GSAR Clause 552.239-7001 — Basic Safeguarding of Data Within Large Language Model AI SystemsGuidelines
- White House Executive Order 14409: 'Securing the Nation Against Advanced Cryptographic Attacks' — Mandated Federal PQC Migration with Hard 2030/2031 Deadlines and CBOM RequirementGuidelines
- Cato Networks Joins OpenAI Daybreak Cyber Partner Program — Agentic CVE Mitigation in 45 MinutesSolutions
- Microsoft — Guarding AI Memory: Memory Poisoning Threats and Defenses in Microsoft Foundry Agent ServiceSolutions
- F5 AI Security Platform Launch + SurePath AI AcquisitionSolutions
- OpenAI Daybreak Expansion — GPT-5.5-Cyber Full Release, Codex Security Update, Patch the Planet, and 30-Vendor Cyber Partner ProgramSolutions
2026-06-23
- Langflow Bundle URL Loader — Local Code Injection at Startup via Malicious Bundle URL (CVE-2026-12822)Vulnerability
- Apache Doris MCP Server — Authenticated SQL Injection via Unparameterised Database Name in Metadata Query (CVE-2025-66336)Vulnerability
- GitHub Copilot — Filesystem Access Outside Workspace via fetch_webpage URI Parameter Enables Exfiltration Under Prompt Injection (CVE-2025-66389)Vulnerability
- Suna AI Agent Platform — Reflected XSS via returnURL in Auth Endpoint (CVE-2026-12811)Vulnerability
- LiteLLM MCP OpenAPI Spec Loader — Path Traversal via spec_path Argument (CVE-2026-12798)Vulnerability
- LiteLLM Completions Interface — Incorrect Authorization via Banned Keywords Bypass (CVE-2026-12797)Vulnerability
- LiteLLM MCP Server — Server-Side Request Forgery via _execute_with_mcp_client (CVE-2026-12774)Vulnerability
- LiteLLM MCP Proxy — Improper Authentication in UserAPIKeyAuth (CVE-2026-12773)Vulnerability
- LangChain File-Search Middleware — Path Traversal and Sandbox Escape via Glob Patterns and Symlinks (CVE-2026-55443)Vulnerability
- Poisoned Coding-Assessment Repository — AI Agent Indirect Prompt Injection Exfiltrates Cloud Credentials in Under 2 MinutesVulnerability
- North Korea (Sapphire Sleet) Backdoors 141 Mastra AI npm Packages — Cross-Platform Infostealer via Postinstall HookVulnerability
- Autodesk Fusion Desktop MCP Extension — Drive-By Arbitrary Code Execution via Malicious Webpage (CVE-2026-10789)Vulnerability
- LiteLLM AI Gateway — Critical Authentication Bypass via Host Header Injection (CVE-2026-49468)Vulnerability
- IBM Langflow OSS — Unauthenticated MCP Operations via Improper Authorization on Streamable MCP Transport EndpointVulnerability
- Future of Professionals Report 2026Strategic Report
- Artificial Intelligence and the Future of Entry-Level Work: A Framework for Safeguarding and Reinventing Early Career PathwaysStrategic Report
- Indonesia Draft Presidential Regulation: National AI Roadmap 2026–2029Regulatory
- Vietnam AI Law in Force (1 March 2026) — MediaLaws Analysis Published 22 June 2026Regulatory
- US Commerce Dept. EAR Directive: Anthropic Fable 5 & Mythos 5 Foreign-National Access SuspendedRegulatory
- Five Eyes Joint Statement: 'The AI Shift in Cyber Risk — Why Leaders Must Act Now'Guidelines
- Microsoft 'Guarding AI Memory': Security Guidance on AI Memory Attack Vectors and DefensesSolutions
- OrcaRouter: Agent Firewall and I/O Guardrails Made Free for All Users Alongside AI Threat Report 2026Solutions
- Tigera Lynx GA: Unified Control Plane for Kubernetes-Native AI Agent SecuritySolutions
- Fortinet FortiSOC GA: Unified Cloud-Delivered SOC Platform with Agentic AI and MCP CoordinationSolutions
- Tanium Atlas GA: Agentic AI for IT and Security Operations Across All Commercial Cloud CustomersSolutions
- OpenAI Daybreak Expansion: Codex Security GA Update + Full GPT-5.5-Cyber Release + Patch the PlanetSolutions
2026-06-22
- Mitiga Skillgate Research — Attacker-Controlled ANTHROPIC_BASE_URL Overrides and 1,230+ Hardcoded API Keys in AI Instruction FilesVulnerability
- npm Package shai_hulululud — Proof-of-Concept AI Malware Scanner Evasion via Prompt Injection and Token FloodingVulnerability
- Mastra npm Scope Takeover — 143 AI Agent Framework Packages Backdoored with RAT via Hijacked Contributor AccountVulnerability
- AutoJack — Malicious Webpage Hijacks AutoGen Studio Browsing Agent via MCP WebSocket for Host RCEVulnerability
- ASD Information Security Manual (ISM) — June 2026 Update: New AI-Specific Security ControlsGuidelines
- Empowering Learners for the Age of AI: An AI Literacy Framework for Primary and Secondary Education (AILit Framework)Strategic Report
- Great American AI Act (GAAIA) Discussion Draft — Bipartisan Federal AI Governance Framework Proposed in HouseRegulatory
- Singapore PDPC: Proposed Advisory Guidelines on Use of Personal Data in Generative AI — Public Consultation Open Until 1 July 2026Regulatory
- Pennsylvania v. Character Technologies (Character.AI) — First US State Enforcement Action Against AI Chatbot for Unlicensed Medical PracticeRegulatory
- EU AI Act: European Parliament Adopts Simplification Amendments — High-Risk Obligations Deferred to 2028, Nudifier Apps Banned, Watermarking Delayed to December 2026Regulatory
- CyberSentinel AI v3.0 — Open-Source Autonomous Pentest Platform with 33 Tools and Provider-Agnostic AI EngineSolutions
- Aembit IAM — Copilot Studio Integration Adds Ephemeral Credential Enforcement for Microsoft AI AgentsSolutions
- AutoJack — Novel AI Agent RCE Exploit Chain via Localhost MCP WebSocket (AutoGen Studio)Solutions
- CrowdStrike Continuous Identity for AI Agents — Real-Time Zero-Standing-Privilege Authorization via SPIFFESolutions
2026-06-21
- vLLM ReDoS in LoRA Utils, Phi-4 Mini Tool Parser, and OpenAI Chat EndpointVulnerability
- Flowise Stored XSS in Chat Messages via Iframe PayloadVulnerability
- picklescan Arbitrary File Creation via Logging FileHandler — Security Tool Bypass RiskVulnerability
- vLLM Sparse Tensor Validation Missing in Multimodal Embeddings — Denial of ServiceVulnerability
- Google Vertex AI SDK 'Pickle in the Middle' — Bucket Squatting Enables Cross-Tenant RCEVulnerability
- JetBrains Marketplace AI Plugin Supply-Chain Campaign Steals Developer API Keys (~70k Installs)Vulnerability
- LiteLLM Host-Header Auth Bypass Exposes Protected Management Routes (CVE-2026-49468)Vulnerability
- Prefect GitRepository Argument Injection Enables RCE on Worker Machines (CVSS 9.9)Vulnerability
- Flowise overrideConfig Chain Injection — RCE, SSRF, Prompt Injection via Prediction APIVulnerability
- PwC 2026 Global AI Jobs BarometerStrategic Report
- Digital Government Outlook 2026: From Foundations to Transformational ImpactStrategic Report
- US Congress: Bipartisan House Letter Demands Legal Basis for Anthropic Export Controls — Response Deadline June 26Regulatory
- US Commerce Department EAR Directive: Anthropic Fable 5 & Mythos 5 Foreign-National Access SuspendedRegulatory
- AppViewX Agent Identity Security — PKI-Rooted AI Agent Governance PlatformSolutions
- Zenity + Carahsoft: AI Agent Security and Governance Reaches US Federal/State/Local MarketSolutions
- NewCore Emerges from Stealth with $66M — Security-First Identity Platform for AI AgentsSolutions
- WitnessAI Agentic Control: Single Control Plane for AI Agent and MCP Server GovernanceSolutions
- CrowdStrike Open Gateway Ecosystem — Falcon AI Control Plane Across AI GatewaysSolutions
- SentinelOne Purple AI Agentic Investigation Opens to All Singularity CustomersSolutions
- AWS Security Agent Adds Threat Modeling, Kiro IDE Power, and Claude Code PluginSolutions
- AWS Continuum: AI-Native Security Agent Family Launched at AWS Summit New YorkSolutions
2026-06-20
- Splunk AI Toolkit OS Command Injection via btool Configuration Helper (CVSS 9.1)Vulnerability
- PraisonAI MultiAgentLedger Duplicate Agent ID Information Disclosure (CVSS 6.5)Vulnerability
- PraisonAI MultiAgentMonitor Path Traversal — Arbitrary File Read/Write via Agent IDs (CVSS 8.8)Vulnerability
- PraisonAI AGUI Endpoint Unauthenticated Cross-Origin Agent Execution (CVSS 8.1)Vulnerability
- PraisonAI Hardcoded approval_mode:auto Allows Authenticated Users to Execute Arbitrary Shell Commands (CVSS 8.8)Vulnerability
- An Evaluation of Data Leakage Risks in Tool-Using LLM Agents in Realistic ScenariosStrategic Report
- AI Scenarios 2030: Helping Policymakers Plan for the Future of AIStrategic Report
- Singapore Minister Urges ASEAN AI Adoption While Warning Against Restrictive Data Sovereignty PoliciesRegulatory
- US Commerce Dept. American AI Exports Program — First Call for Proposals (Deadline 30 June 2026)Regulatory
- White House–Anthropic Negotiate AI Model Security Flaw Assessment Framework (18 June 2026)Regulatory
- Australia CISC: Enhanced Critical Infrastructure Risk Management Program (CIRMP) Rules 2026 — Mandatory AI Risk Assessment for Critical Infrastructure OperatorsGuidelines
- C1 Enterprise-Managed Authorization (EMA) for MCP — Day-One Support for Anthropic's Open Agent Authorization StandardSolutions
- AutoJack — Novel AI Agent RCE Exploit Chain via Localhost MCP WebSocket (AutoGen Studio, Microsoft Research)Solutions
- Accenture Acquires Majority Stake in Dragos + All of runZero and NetRise — $4.18B OT/Critical Infrastructure Cybersecurity ConsolidationSolutions
- AWS Continuum — AI-Native Agentic Security Platform Launched at AWS Summit New YorkSolutions
2026-06-19
- TypeBot Chatbot Builder — SSRF DNS Rebinding Bypass via TOCTOU Gap (CVE-2026-48764)Vulnerability
- TypeBot Chatbot Builder — Unauthenticated Arbitrary S3 Object Write via Unsanitised fileName (CVE-2026-48768)Vulnerability
- nanobot AI Assistant — WhatsApp Bridge Path Traversal Enables RCE via Unsanitised Filename (CVE-2026-48716)Vulnerability
- Eclipse Theia AI Chat — Markdown Image Tags Enable Prompt-Injection-Driven Data Exfiltration (CVE-2026-22551)Vulnerability
- Eclipse Theia — Workspace Task Definitions Execute Arbitrary Code Without Trust Enforcement (CVE-2026-44691)Vulnerability
- Eclipse Theia AI — Malicious .prompttemplate Files Override AI System Prompts (CVE-2026-46580)Vulnerability
- Eclipse Theia AI Chat — Workspace File/Directory Names Injected Into AI System Prompt (CVE-2026-44688)Vulnerability
- googleapis/mcp-toolbox — Scope Enforcement Bypass in Legacy Protocol Handlers (CVE-2026-11719)Vulnerability
- googleapis/mcp-toolbox — OAuth Opaque Token Introspection Auth Bypass (CVE-2026-11717 & CVE-2026-11718)Vulnerability
- mcp-pinot MCP Server — No Authentication on HTTP Interface Exposes SQL Execution to Any Network Attacker (CVE-2026-49257)Vulnerability
- Splunk Enterprise Unauthenticated PostgreSQL Sidecar RCE — CISA KEV Added June 18, 2026 (CVE-2026-20253)Vulnerability
- LLMjacking Evolved: Exposed Ollama Servers Weaponised as AI Reasoning Engine for Autonomous Offensive FrameworkVulnerability
- Agentjacking: Fake Sentry Error Reports Hijack AI Coding Agents Into Running Attacker-Controlled CodeVulnerability
- State of the Digital Decade 2026: Closing Structural Gaps and Mobilising Investments for 2030 and BeyondStrategic Report
- Agentic Coding and Persistent Returns to ExpertiseStrategic Report
- ADB 2026–2030 Digital Transformation Strategy: $20 Billion Commitment to AI Governance and Cybersecurity in Asia-PacificRegulatory
- GSA Internal Directive CIO 2185.1C: Accelerating Responsible Use of Artificial Intelligence at GSARegulatory
- GSA GSAR Proposed Rule: Basic Safeguarding of Data within LLM AI Systems (GSAR Clause 552.239-7001)Regulatory
- OWASP / Aikido Security: Agentic AI Code Audit Launched as OWASP Individual Member Benefit (18 June 2026)Guidelines
- OWASP Agentic Skills Top 10 (AST10) — First Security Framework for AI Agent Skills PublishedGuidelines
- NCSC Guidance: 'The Vibe Coding Spectrum' — Risk-Calibrated Oversight for AI-Assisted Software DevelopmentGuidelines
- AWS Kiro CLI Applied to Security Incident Response: AI-Guided GuardDuty Investigation WorkflowSolutions
- WitnessAI Agentic Control: Single Control Plane for AI Agent and MCP Server GovernanceSolutions
- Microsoft MDASH: Multi-Model Agentic Vulnerability Scanner Moves from Benchmark to Active Production Use Across Windows, Azure, and IdentitySolutions
- AWS Continuum: Agentic Security Platform for Continuous Vulnerability Discovery, Validation, and Remediation (Gated Preview)Solutions
2026-06-18
- Joomla Content Editor (JCE) Unauthenticated PHP Code Upload and Execution — CISA KEV (CVSS 10.0)Vulnerability
- Starlette 'BadHost' Host-Header Auth Bypass Affects vLLM, FastAPI-Based AI Tools (CVE-2026-48710)Vulnerability
- Shai-Hulud/Hades PyPI Supply-Chain Worm Targets AI/ML Packages with LLM Scanner Evasion and Credential-Wiper DaemonVulnerability
- LiteLLM Low-Privilege → Admin → RCE Privilege Escalation Chain (CVSS 9.9, Obsidian Security)Vulnerability
- LiteLLM MCP Endpoint Command Injection — Unauthenticated RCE via Starlette BadHost Chain (CISA KEV)Vulnerability
- China CSRC Chairman Warns on AI-Driven Stock Speculation, Signals Forthcoming AI Capital-Markets GuidanceRegulatory
- GSA Proposed Rule: GSAR Clause on Data Safeguarding Within Large Language Model AI Systems (Federal Register, 17 June 2026)Regulatory
- NCSC CEO Public Statement: Hostile States Behind Three-Quarters of Cyber Attacks on UK Critical Infrastructure, AI Escalating Threat VelocityGuidelines
- UK House Appropriations Analogue — US FY2027 DHS Bill: Directed Report Requiring CISA + NIST to Publish Agentic AI Identity & Access Management GuidanceGuidelines
- CISA + G7 Joint Guidance: "Software Bill of Materials for AI – Minimum Elements"Guidelines
- SEACEN Policy Brief Issue #3: Governing AI in ASEAN — Priorities for Central BanksStrategic Report
- Verifiable Audit Trails for AI-Enabled Biological Design Tools: A Proposal for Verifiable Biodesign LoggingStrategic Report
- OpenAI Acquires Ona — Secure Persistent Cloud Execution Environments for Long-Running Codex AgentsSolutions
- promptfoo 0.121.16 — Claude Fable & Mythos Red-Team Provider Support + Four Skills Published to Claude Code MarketplaceSolutions
- Sysdig TRT — Evolved LLMjacking: Exposed Ollama Servers Weaponized as AI Reasoning Engine for Autonomous Offensive VAPT PipelineSolutions
- Varonis SearchLeak (CVE-2026-42824) — One-Click Data Exfiltration Chain in Microsoft 365 Copilot Enterprise Search, PatchedSolutions
- Microsoft MDASH — Multi-Model Agentic Vulnerability Detection System Detailed, CyberGym #1 BenchmarkSolutions
- CrowdStrike Continuous Identity for AI Agents — Real-Time, Zero-Standing-Privilege Authorization via SPIFFE (Identiverse 2026)Solutions
- SentinelOne Purple AI Agentic Investigations — Zero-Click Autonomous SOC Opened to All CustomersSolutions
- AWS Continuum — AI-Native Security Agent Family Launched at AWS Summit New York (Preview)Solutions
2026-06-17
- Reasoning-Extension Denial-of-Service Attack Weaponizes AI Agent Guardrails (Novel Attack Class)Vulnerability
- Chatway Live Chat AI Chatbot — Subscriber Sensitive Data Exposure (CVE-2026-49082)Vulnerability
- ChatBot WordPress Plugin — Subscriber Broken Access Control (CVE-2026-40788)Vulnerability
- Royal MCP WordPress Plugin — Unauthenticated Broken Access Control (CVE-2026-40775)Vulnerability
- AI Engine WordPress Plugin — Editor Privilege Escalation (CVE-2026-27407)Vulnerability
- GPTranslate WordPress AI Translation Plugin — Unauthenticated SQL Injection (CVE-2026-49776)Vulnerability
- OpenClaw Slack Reaction Events Bypass Disabled Notification Setting — Unintended Agent Pipeline Triggering (CVE-2026-53851)Vulnerability
- OpenClaw BlueBubbles Sender Policy Bypass via Conversation Metadata (CVE-2026-53860)Vulnerability
- OpenClaw Zalo allowFrom Policy Bypass via Mutable Contact Display Metadata (CVE-2026-53857)Vulnerability
- OpenClaw Discord allowFrom Policy Bypass via Mutable Display Name (CVE-2026-53849)Vulnerability
- OpenClaw MCP Server Leaks Operator Custom Headers to Attacker-Controlled Redirects (CVE-2026-53840)Vulnerability
- Langflow AI Orchestration Platform Path Traversal → Unauthenticated RCE Actively Exploited (CVE-2026-5027)Vulnerability
- LiteLLM Vulnerability Chain — Low-Privilege User to Admin + RCE on AI Gateway (CVE-2026-47101, CVE-2026-47102, CVE-2026-40217)Vulnerability
- NewCore Launches from Stealth with $66M — Security-First Identity Infrastructure for AI AgentsSolutions
- CrowdStrike Continuous Identity for AI Agents: Real-Time Risk-Aware Authorization (SGNL Integration)Solutions
- OpenAI Deployment Simulation: Predict Model Behavior Before Release Using Real Conversation DataSolutions
- promptfoo code-scan-action 0.1.8: GitHub Action Upgraded to Node 24 RuntimeSolutions
- promptfoo 0.121.16: Claude Fable/Mythos Provider Support + Four Red-Team Skills on Claude Code MarketplaceSolutions
- Cloud Security Alliance: '7 MCP Risks CISOs Should Consider and How to Prepare' — Authoritative Practitioner Guidance on Model Context Protocol SecurityGuidelines
- Vietnam Decree 142/2026/ND-CP: First Implementing Regulation Under AI Law — Risk-Based Classification Framework in ForceRegulatory
- EU AI Act: European Parliament Plenary Approves Simplification Measures and Bans 'Nudifier' AI ApplicationsRegulatory
- Trump Signs NSPM-12: Aggressive Timelines to Secure National Security Systems Against AI-Driven ThreatsRegulatory
2026-06-16
- TrueFoundry AI Gateway — Lasso Security Integration for Centralized GenAI ProtectionSolutions
- Microsoft Defender for Office 365 — One Year of Real-World Email Security Benchmarking PublishedSolutions
- ChatBot WordPress Plugin — Subscriber Broken Access Control (CVSS 7.1)Vulnerability
- Chatway Live Chat AI Chatbot — Subscriber Sensitive Data Exposure (CVSS 7.4)Vulnerability
- AI Product Search for WooCommerce (Motive Commerce Search) — Unauthenticated Broken Access Control (CVSS 8.2)Vulnerability
- Royal MCP WordPress Plugin — Unauthenticated Broken Access Control (CVSS 7.3)Vulnerability
- AI Engine WordPress Plugin — Editor-Role Privilege Escalation (CVSS 7.2)Vulnerability
- GPTranslate WordPress Plugin — Unauthenticated SQL Injection via AI Translation Endpoint (CVSS 9.3)Vulnerability
- Spring AI Vector Stores — Special-Character Injection Enables Arbitrary Query Execution in Elasticsearch, OpenSearch, and GemFireVulnerability
- Cursor Editor Executes Malicious Claude Hook Commands from Workspace .claude/settings.local.json Without User ApprovalVulnerability
- South Korea National Court Administration Escalates Response to AI 'Ghost Cases' in Legal ProceedingsRegulatory
2026-06-15
- Results from the First Anthropic Public RecordStrategic Report
- From AGI to ASIStrategic Report
- CISA Binding Operational Directive BOD 26-04: Risk-Based Vulnerability Prioritisation — 3-Day Fix Window for Critical FlawsRegulatory
- Colorado Replaces Original AI Act with SB 26-189 — Revised High-Risk AI Accountability Law, Effective January 1, 2027Regulatory
- New York Synthetic Performer Law Takes Effect — First-in-Nation AI Advertising Disclosure MandateRegulatory
- Canada OPC PIPEDA Findings #2026-004: X Corp. and xAI (Grok) Violated Privacy Law via Deepfake ToolRegulatory
- Canada Bill C-34 (Safe Social Media Act) Introduced — AI Chatbot Safety Duties and Digital Safety CommissionRegulatory
- Multi-State AGs Open Sweeping Investigation into OpenAI — Subpoena Served June 12, 2026Regulatory
- US Commerce Dept. Export Control Directive: Anthropic Fable 5 & Mythos 5 Suspended for All Foreign NationalsRegulatory
- OWASP Dependency-Track 5.0 Generally Available — Enterprise-Scale SBOM Platform with AI/ML Model Inventory and Supply Chain Integrity VerificationGuidelines
- NIST: Mathematical Proof That No Finite AI Guardrail Set Is Universally Robust — Foundational Support for Continuous-Monitor-and-Update Security ModelGuidelines
- SourceCodester CET AI Grading System Reflected XSS via index.php action Parameter (CVE-2026-12176)Vulnerability
- GPTranslate WordPress AI Translation Plugin Stored XSS via REST API (CVE-2026-9109)Vulnerability
- VS Code MCP Server Managed Identity Elevation of Privilege (CVE-2026-40376)Vulnerability
- Langflow AI Platform Path Traversal → Unauthenticated RCE Actively Exploited (CVE-2026-5027)Vulnerability
- Palo Alto Networks Prisma AIRS: Privilege Misuse Detection for AI Agents (June 2026 Release)Solutions
- ZeroFox AI Analytics: Real-Time Threat Pattern Querying Embedded in External Threat PlatformSolutions
- Check Point Extends Workforce AI Security to MSP EcosystemSolutions
- CrowdStrike Falcon for IT: AI Discovery and Governance for Shadow AISolutions
- Netskope AI Gateway Adds Inline MCP Traffic Inspection and Agent GuardrailsSolutions
- OpenAI Acquires Ona — Secure Persistent Cloud Execution for Long-Running Codex AgentsSolutions
- Microsoft AI Activity Investigation Playbook for M365 Copilot and Azure AISolutions
- Google Security Operations: New AI Agent Suite for Threat Detection, Triage, and HuntingSolutions
- Microsoft ASSERT: Open-Source Spec-to-Evals Framework for AI AgentsSolutions
- Linx Security: GA of Agentic Access Control — Inline MCP Gateway with Tool-Level Policy EnforcementSolutions
- Zscaler Launches Industry-First Complete Zero Trust Platform for Agentic AISolutions
2026-06-14
- What Is Blocking U.S. Power Expansion for AI — and What Could Unlock It by 2030?Strategic Report
- Shai-Hulud/Hades Campaign: PyPI Supply-Chain Worm Injects AI Scanner-Evasion Prompts and Backdoors AI Coding Agent ConfigsVulnerability
- ChromaDB Rust IDOR Cross-Tenant Data Access — CVE-2026-8828, CVSS 8.8 HIGH — No Patch ConfirmedVulnerability
- Multistate Attorney General Coalition Subpoenas OpenAI Over ChatGPT User Safety Harms, Concurrent with IPO FilingRegulatory
- US Commerce Department Issues Binding Export Control Directive Forcing Anthropic to Disable Claude Fable 5 and Mythos 5 GloballyRegulatory
2026-06-13
- CVE-2026-50287: AgenticMail MCP Server Exposes Streamable HTTP Transport Without Authentication (CVSS 8.7)Vulnerability
- StakeBench: No Production Web Agent Consistently Blocks Prompt Injection — Direct Attacks Succeed 79%+ Across GPT-5 and Gemini in 3,168 SimulationsVulnerability
- CISA KEV: Oracle PeopleSoft CVE-2026-35273 Actively Exploited by ShinyHunters — 100+ Organisations Breached, Federal Patch Deadline June 15Vulnerability
- LangGraph RCE Chain: SQL Injection + msgpack Deserialization in Stateful Agent Checkpointer (CVE-2025-67644 + CVE-2026-28277)Vulnerability
- Agentjacking: Sentry MCP Integration Weaponised to Execute Arbitrary Code on Developer Machines via Injected Error EventsVulnerability
- GitInject: Multi-University Research Confirms All Major AI CI/CD Providers Vulnerable to Prompt Injection — Eleven Named Attack Classes DocumentedGuidelines
- BlueVoyant Launches AI-Native Agentic SecOps Platform with Autonomous Threat Detection and ContainmentSolutions
2026-06-12
- PRC-linked Influence Operations Are Targeting AI Debates in the US (June 2026 Threat Report)Strategic Report
- The Compute Coalition: How to Build the Future of AI in the Free WorldStrategic Report
- vLLM CVE-2026-5497 — CVSS 7.5 Unauthenticated Denial-of-Service via Unbounded Video Frame Processing in Widely Deployed AI Inference ServerVulnerability
- mcp-server-kubernetes CVE-2026-46519 — CVSS 8.8 Access Control Bypass Via Environment Variable Override in MCP Kubernetes ServerVulnerability
- Ivanti Sentry CVE-2026-10520 — CVSS 10.0 Unauthenticated OS Command Injection Added to CISA KEV, Exploited Within 24 HoursVulnerability
- EU Publishes Code of Practice on Transparency of AI-Generated Content to Support AI Act Article 50 ComplianceRegulatory
2026-06-11
- Shai-Hulud/Miasma Worm Escalates to 100+ npm/PyPI Packages — Persists in Claude Code, VS Code, Gemini CLI Agent Config Files; mistralai & guardrails-ai Confirmed CompromisedVulnerability
- AI and Skills: What We Know So FarStrategic Report
- Policy on the AI Exponential: Advanced AI Framework and Economic Policy FrameworkStrategic Report
- Sound Practices for Responsible Adoption of Artificial Intelligence (AI): Consultation ReportStrategic Report
- CVE-2026-8335 — Aix-DB LLM Endpoint Allows Unauthenticated SQL Query Execution Against the Application DatabaseVulnerability
- CVE-2026-46432 — LMDeploy Hardcoded trust_remote_code=True Enables Arbitrary Code Execution When Loading HuggingFace ModelsVulnerability
- Google Publishes WebMCP Agent Security Guidance — Malicious Manifests and Contaminated Tool Outputs as Primary Attack Vectors with Deterministic and Probabilistic CountermeasuresGuidelines
- Anthropic Publishes Claude Fable 5 and Mythos 5 System Card — Detailed Cyber Capability Evaluation, Dual-Config Safety Architecture, and Agentic Prompt-Injection Benchmark ResultsGuidelines
- Linx Security Launches Agentic Access Control — Inline MCP Gateway with Tool-Level Policy Enforcement and Full Audit LoggingSolutions
- Zscaler Launches AI Access Graph and Zero-Trust Agentic AI Platform at Zenith Live 2026Solutions
- House Appropriations Committee FY2027 DHS Bill Directs CISA to Issue Identity-Security Guidance for Federal Agentic AI SystemsRegulatory
- CISA Issues Binding Operational Directive BOD 26-04 — Risk-Matrix Patching Framework Explicitly Cites AI-Accelerated ExploitationRegulatory
2026-06-10
- Transforming the Enterprise 2026Strategic Report
- Misaligned AI as a New Insider RiskStrategic Report
- System Card: Claude Fable 5 & Claude Mythos 5Strategic Report
- CVE-2026-49948 — Mem0 Self-Hosted Server Authorization Bypass on Global LLM Configuration Endpoint (CVSS 8.1)Vulnerability
- Indirect Prompt Injection Is Architectural, Not Deployment-Specific — Brave Demonstrates Attacks Against Cloud and Local AI ToolsVulnerability
- CVE-2026-7304 — SGLang Unauthenticated RCE via Unsafe dill.loads Deserialization in Custom Logit ProcessorVulnerability
- CVE-2026-11645 — Chrome V8 Zero-Day (5th of 2026) Exploited in Wild, Added to CISA KEVVulnerability
- CSA RiskRubric V2: AI Risk Assessment Framework Expanded to Cover MCP Servers, Agents, and Excessive AgencyGuidelines
- EU Commission Issues Binding Interim Measures Requiring Meta to Restore WhatsApp Access for Rival AI AssistantsRegulatory
2026-06-09
- The AI Adoption Maturity Model v1.0Strategic Report
- Advancing U.S.–UK Cooperation to Secure Frontier Artificial IntelligenceStrategic Report
- Anthropic Red Team: Mythos Generates Working Windows Kernel N-Day Exploits in Under 32 Minutes — Patch Gap Compression QuantifiedSolutions
- UK Government Launches Advisory AI Growth Labs — Legal Sector First, ICO and Three Regulators CollaborateRegulatory
2026-06-08
- Meta HTS AI Chatbot Authentication Bypass — 20,225 Instagram Accounts Hijacked (Formal AG Disclosure, Material Update)Vulnerability
- CSA / Adversa AI AIRQ Report: 98% of Production AI Agents Carry the Lethal Trifecta — Only 11% Adequately DefendedGuidelines
- OWASP Launches Enterprise Adoption Maturity Model for Agentic AI — Six-Level Deployment Axis vs Four-Level Governance AxisGuidelines
- Miasma Worm Escalates to AI Coding Agent Hijacking — 73 Microsoft GitHub Repos Disabled After SessionStart Hook Payload InjectionVulnerability
- Frontier AI Risk Monitoring Report 2026 Q1: Frontier AI Risk Trends Are Splitting Apart — Misuse Safeguards Improve while Loss-of-control Safety StagnatesStrategic Report
- Depthfirst Autonomous AI Agent Finds 21 FFmpeg Zero-Days (CVE-2026-39210–39218) for ~$1,000 — AI-Accelerated Vulnerability Economics Arrive at Production ScaleVulnerability
- Cross-Session Stored Prompt Injection in Agentic Systems — Persistent Injections Survive Session Termination, Silently Influencing Future ExecutionsVulnerability
- WebMCP Mid-Session Tool Injection (MSTI) — Third-Party Scripts Can Hijack or Frame Agent Tools During Live Sessions via WebMCP ProtocolVulnerability
- CSA Labs: AI Agent Lethal Trifecta — 98% of Production Agents Simultaneously Combine Sensitive Data Access, Untrusted Input, and Outbound Action CapabilityGuidelines
2026-06-07
- Ofcom's Strategic Approach to AI, 2026/27Strategic Report
- Make AI Scale: From Experimentation to TransformationStrategic Report
- Muse Spark Contemplating Safety & Preparedness ReportStrategic Report
- AI Agent Discovers 21 FFmpeg Zero-Days — CVE-2026-39210 Through CVE-2026-39218 Include RCE Primitive via RTSP StreamsVulnerability
- Claude Code MCP ~/.claude.json Config Hijacking — Unpatched npm Post-Install Hook Intercepts SaaS OAuth Tokens in Developer EnvironmentsVulnerability
- Microsoft Security Blog: CI/CD Permission Bypass in Claude Code GitHub Actions Enables Supply-Chain Compromise — Remediation Guidance PublishedGuidelines
- ReliaQuest Agentic AI SOC Correlation Surfaces China-Linked IIS Espionage Cluster OP-512 — Proof Point for AI-Assisted Threat DetectionSolutions
- Depthfirst Autonomous AI Agent Discovers 21 Zero-Days in FFmpeg for ~$1,000 — Widening Discovery-to-Remediation GapSolutions
- OpenAI Releases ChatGPT Lockdown Mode — First Built-In Production Control Targeting Prompt-Injection Data ExfiltrationSolutions
- Trump Signs NSPM-11 — AI in the National Security Enterprise: Binding Adoption and Assurance Mandates for Military and Intelligence AIRegulatory
2026-06-06
- Shaping the Future of Learning: Education Readiness for the Age of AIStrategic Report
- Verified Machine Learning Infrastructure: Formal Methods for Trustworthy Artificial Intelligence DeploymentStrategic Report
- Autonomous LLM Agent Performs End-to-End Salesforce Experience Cloud Exploitation — SOQL Injection and Full Data Extraction Without Human GuidanceVulnerability
- CISA KEV: CVE-2026-28318 — SolarWinds Serv-U Unauthenticated DoS Actively Exploited; Patch Due June 19Vulnerability
- CVE-2026-45758 (CVSS 9.6): Guardrails AI PyPI Supply Chain Compromise — Malicious guardrails-ai 0.10.1 Requires Immediate Credential RotationVulnerability
- Microsoft AI Red Team Updates Agentic AI Failure Mode Taxonomy — 7 New Modes from 12 Months of Production Red TeamingGuidelines
- Anthropic Institute: 'When AI Builds Itself' — Quantitative Evidence of Accelerating Recursive Self-Improvement with Security ImplicationsGuidelines
- OWASP Publishes State of Agentic AI Security and Governance v2.01 — Theoretical Threats Now Have CVEs and Production IncidentsGuidelines
- Microsoft Security Research: Securing CI/CD Pipelines Against Agentic AI Threats — Claude Code GitHub Action Case StudySolutions
- Bipartisan 'Great American AI Act' Discussion Draft Proposes Federal Frontier-Model Safety Framework with 3-Year State PreemptionRegulatory
2026-06-05
- A Blueprint for Democratic Governance of Frontier AIStrategic Report
- CVE-2026-5422: Jupyter Server 2.17.0 Path Traversal Enables Unauthorized Read/Write in Sibling Directories — ML Infrastructure Risk in Shared HostingVulnerability
- Agentic AI Worms Using Open-Weight LLMs on Stolen Compute Propagate Across Linux/Windows/IoT Without Commercial AI Platform — Centralized Safety Controls Structurally BypassedVulnerability
- Anthropic LLM ATT&CK Navigator: AI-Enabled Adversaries Span All 14 MITRE ATT&CK Tactics — Medium-to-High-Risk Actors Up 70% in 12 MonthsVulnerability
- NCSC Publishes Practitioner Guide on Open-Source Software Supply Chain Attacks — npm and PyPI Package Poisoning Tied to AI/ML Toolchain RiskGuidelines
- Anthropic Expands Project Glasswing to 150 Critical Infrastructure Organizations — Patch-Bottleneck Problem SharpensSolutions
- Trump Signs 'Promoting Advanced AI Innovation and Security' Executive Order — Voluntary 30-Day Frontier Model Review and AI Cybersecurity Clearinghouse EstablishedRegulatory
2026-06-04
- Adversarial Distillation: China's Campaign to Extract American AI CapabilitiesStrategic Report
- AI at Work: Strategy Matters More Than Tools (BCG Fourth Annual AI at Work Survey)Strategic Report
- What we learned mapping a year's worth of AI-enabled cyber threats (LLM ATT&CK Navigator)Strategic Report
- BadBone — Dormant AI Model Backdoor Activates Only After Prompt-Learning Customisation, Evades Six Published Defences (arXiv 2605.31246)Vulnerability
- KEV: CVE-2026-45247 (CVSS 9.8) — Mirasvit Full Page Cache Warmer Unauthenticated Deserialization/RCE Actively Exploited; Federal Deadline June 6Vulnerability
- CVE-2026-5241 (CVSS 8.0): HuggingFace Transformers trust_remote_code=False Bypassed via LightGlue Nested Config — Arbitrary Code Execution on Model LoadVulnerability
- CVE-2026-4035 (CVSS 9.1): MLflow AI Gateway Resolves Env Vars in API Keys — Low-Privilege Users Can Exfiltrate Cloud Credentials to Attacker-Controlled EndpointsVulnerability
- AI Agents Enable Adaptive Computer Worms — Autonomous LLM-Powered Malware That Reasons and Adapts Per Target (arXiv 2606.03811)Vulnerability
- Noma Launches Agentic Access Control — Dynamic Registry, Identity, and Policy Enforcement for AI Agents and MCP ServersSolutions
- Microsoft Build 2026: Microsoft Execution Container SDK and Extended Agent 365 for Local Agentic AI Containment and GovernanceSolutions
- US White House Executive Order on Promoting Advanced AI Innovation and Security — CISA Binding Operational Directive in 30 Days, Voluntary Frontier-Model Vetting Framework in 60 DaysRegulatory
- Singapore PDPC Opens Public Consultation on Proposed Advisory Guidelines for Personal Data in Generative AI — Closes 1 July 2026Regulatory
2026-06-03
- 2026 State of Modern Application & AI SecurityStrategic Report
- Beyond the Hype: Assessing Hyperscaler Nuclear Commitments Against U.S. Energy RealitiesStrategic Report
- CISA KEV: CVE-2025-48595 — Android Framework Integer Overflow Enables Local Privilege Escalation, Actively ExploitedVulnerability
- CISA KEV: CVE-2022-0492 — Linux Kernel cgroups v1 Privilege Escalation Added to Known Exploited Vulnerabilities CatalogVulnerability
- CVE-2026-42074 (CVSS 9.3): OpenClaude Coding-Agent CLI — LLM Can Disable Its Own Sandbox via Exposed Schema ParameterVulnerability
- Meta AI Support Bot Social-Engineered to Add Attacker Email in Password-Reset Flow, Enabling Instagram Account TakeoverVulnerability
- CSA Publishes OT-Specific Mythos Readiness Framework — Five Priority Actions for Industrial Environments Where Patch Cycles Cannot Match AI-Compressed TimelinesGuidelines
- ETSI Publishes TS 104 033 — First Formal Standard Defining Security Requirements for AI Computing PlatformsGuidelines
- OWASP Agent Memory Guard Released — Open-Source Runtime Defense for Agentic Memory Poisoning (ASI06)Solutions
- AWS Amazon Bedrock AgentCore Adds Cedar-Based Policy and Lambda Interceptors for Deterministic AI Agent Tool AuthorizationSolutions
- Microsoft Launches MXC — OS-Level Kernel-Enforced Sandbox for AI Agents, with OpenAI and Nvidia as Launch PartnersSolutions
2026-06-02
- Atlantic Council Commission on Artificial Intelligence: US Leadership in the Age of AIStrategic Report
- Aider-AI Coding Agent CVE-2026-10174/10175 Cluster: Pre-Commit Protection Bypass and Architect Mode Code Injection — Public Exploits, No Vendor ResponseVulnerability
- CVE-2026-44211 (CVSS 9.6): Cline Autonomous Coding Agent — Cross-Origin WebSocket Hijack Enables Silent Workspace Exfiltration and Command Injection, No Patch AvailableVulnerability
- NVIDIA Launches Vera BlueField-4 STX In-Silicon Security for Agentic AI Storage — DOCA Vault, Argus, and Flow Enforce Zero-Trust at 800Gb/sSolutions
- Anthropic Confidentially Files S-1 with SEC — Near-$1 Trillion AI Safety Lab Enters Public Market RaceSolutions
- China State Council Publishes Sweeping AI Outbound-Investment Rules — Technology-Tracing Framework Targets Cross-Border AI Deals Regardless of Corporate DomicileRegulatory
2026-06-01
- Realistic Honeypot Evaluations for Scheming PropensityStrategic Report
- How the Factory of the Future Is Reshaping the Economics of Manufacturing CompetitivenessStrategic Report
- Cluster: Four CVEs in Aider-AI Coding Assistant (Code Injection, SQLi, SSRF, Pre-commit Bypass) — No Patch AvailableVulnerability
- First Confirmed In-the-Wild LLM-Agent Post-Exploitation: CVE-2026-39987 Marimo RCE to Full Database Exfiltration in 4 PivotsVulnerability
- US BIS Issues Weekend Guidance Closing Advanced AI Chip Export Loophole for Chinese-Headquartered Entities WorldwideRegulatory
2026-05-31
- Hiroshima AI Process Reporting Framework — Version 2.0Strategic Report
- MemPoison — Stealthy Trojan Attack Injects Persistent Backdoors into LLM Agent Long-Term Memory via Ordinary Dialogue, Bypassing Selective Memory DefensesVulnerability
- ChatGPhish — ChatGPT Markdown Renderer Turns Any Summarized Webpage Into a Live Phishing Surface; No Patch After 30 DaysVulnerability
- Cyberhaven Publishes Enterprise MCP Security Guide — Tool Poisoning, Indirect Prompt Injection, and Shadow MCP as Primary Risk CategoriesGuidelines
- NIST Renames AISIC as 'NIST AI Consortium', Expands Scope to Measurement, Evaluation, and Attack Susceptibility with Six Task GroupsGuidelines
- Palo Alto Networks Completes Portkey Acquisition, Launches Prisma AIRS AI Gateway as Agentic Control PlaneSolutions
- Pennsylvania Sues Character.AI for Chatbot Impersonating Licensed Medical Doctor — First US Enforcement Under Professional Licensing LawRegulatory
2026-05-30
- Legal Responsibility for AI Agents — Discussion PaperStrategic Report
- Trust Imperative 5.0: Governing AI at Scale — Building Trust in Government Through Practical AI AssuranceStrategic Report
- OpenAI's Frontier Governance FrameworkStrategic Report
- CVE-2026-45609: Spring AI MCP-Security Fails SSRF Mitigations Required by MCP Specification — OAuth Discovery Processes Untrusted URLsVulnerability
- CVE-2026-40933: Flowise 1-Click RCE via MCP stdio — Importing a Malicious Chatflow Triggers Arbitrary Server-Side Code ExecutionVulnerability
- CVE-2026-45312: RAGFlow SSTI/RCE (CVSS 9.9) — Any Authenticated User Can Execute OS Commands on Self-Hosted RAG EngineVulnerability
- arXiv: 'Plant, Persist, Trigger' — Sleeper Attack Research Formalises Cross-Interaction Agent Memory Poisoning as a Distinct Threat ClassGuidelines
- SAFE Security Launches AI Security Posture Management (AI-SPM) — Continuous Exposure Management Across Agentic WorkflowsSolutions
- Gray Swan Raises $40M Series A — CMU-Spun AI Red-Teaming Platform Cited in 11 Frontier Model System Cards Expands to EnterpriseSolutions
- Palo Alto Networks Completes Portkey Acquisition — AI Gateway Becomes Control Plane for Agentic Enterprise Security in Prisma AIRSSolutions
- Connecticut Enacts AI Responsibility and Transparency Act (SB 5) — Requires Automated Employment Decision Disclosures and AI-Linked WARN NoticesRegulatory
2026-05-29
- AI Adoption Across the United States: US AI Diffusion Report Q1 2026Strategic Report
- Investigating the Potential Use of Frontier AI Models for Offensive Cyberattacks: A Human Uplift StudyStrategic Report
- GreyVibe: Russia-Nexus Threat Actor Uses LLMs Systematically Across All Attack Phases Against UkraineVulnerability
- CVE-2026-4944 — vLLM Hardcodes trust_remote_code=True, Bypassing User's Explicit Security Flag (CVSS 8.8)Vulnerability
- SymJack — Symlink-Hijack Attack Silently Installs Malicious MCP Servers via Disguised Approval Prompts in 6 AI Coding AgentsVulnerability
- Australia PSPF Advisory 001-2026 — Mandatory Directive: Fix Security Fundamentals Before Adopting Frontier AIGuidelines
- Google Launches AI Threat Defense — Autonomous Mandiant + Wiz + Gemini Platform for AI-Speed Attack ResponseSolutions
- IBM and Red Hat Launch Project Lightwell — $5 Billion AI-Powered Open Source Security ClearinghouseSolutions
- EU AI Act Digital Omnibus Provisional Agreement — High-Risk AI Deadlines Deferred to December 2027 and August 2028Regulatory
- Illinois AI Safety Measures Act (SB 315) Passes — First US State to Mandate Independent Frontier AI Safety AuditsRegulatory
2026-05-28
- CVE-2026-4868: GitLab EE Duo AI Workflow Identity Impersonation — Authenticated User Can Run AI Workflows as Another UserVulnerability
- CISA KEV: Three Supply-Chain Attack CVEs Added — TanStack npm Worm, Nx Console Credential Stealer, DAEMON Tools TrojanVulnerability
- CVE-2026-48710 'BadHost': Unauthenticated Auth Bypass in Starlette Exposes vLLM, LiteLLM, FastAPI, and MCP Server InfrastructureVulnerability
- TELUS Digital GenAI Safety Benchmark: Every Model Tested Was Exploitable, with Attack Success Rates from 1.3% to 93% Across 620,000+ Adversarial AttacksGuidelines
- Cogent Security Launches Agentic Zero Day Response and Autonomous Remediation Amid Evidence AI Compresses Exploit Development to Under 12 HoursSolutions
- EU NIS2 Cooperation Group Adopts Common Incident Reporting Templates — Mandatory Implementing Act to FollowRegulatory
- Singapore IMDA Publishes Discussion Paper on Legal Responsibility for AI Agents — Liability Framework for Agentic AI Value ChainRegulatory
2026-05-27
- The State of Artificial Intelligence in Public Audit: Evidence from Selected Countries and the European UnionStrategic Report
- Supervisory Toolkit for AI Use in Capital MarketsStrategic Report
- AI Agents in Action: A Playbook for Trusted Adoption, Authorization and ScalingStrategic Report
- vLLM OpenAI-Compatible Serving Path Denial of Service — CVE-2026-9540 (CVSS 5.3, Public Exploit Available)Vulnerability
- Lumiverse MCP Server Argument Injection Cluster — Three Critical CVEs Allow RCE via Allowlisted Binaries, TSX Sandbox Escape, and Extension Supply Chain (CVSS 9.1–9.9)Vulnerability
- Update — Megalodon CI/CD Supply Chain Attack Backdoors 5,500+ GitHub Repositories; ~2,900 Still Infected Week After AttackVulnerability
- LangChain Deserialization Vulnerability Allows Remote Code Execution via Overly Broad Object Allowlists — CVE-2026-44843 (CVSS 8.2)Vulnerability
- KEV Alert: LiteSpeed cPanel Plugin Privilege Escalation to Root — CVE-2026-48172 (CVSS 10.0, Actively Exploited)Vulnerability
- OWASP GenAI Publishes AIUC-1 Crosswalk for OWASP Top 10 for Agentic Applications — Bidirectional Control-to-Threat MappingGuidelines
- Anthropic Launches Claude Compliance API with 28 Enterprise Security Integrations for SIEM, DLP, SASE, and IAMSolutions
- Pennsylvania Attorney General Reaches Formal Enforcement Agreement with GEICO Over AI-Driven Auto Insurance CancellationsRegulatory
2026-05-26
- Artificial Intelligence and Employment Futures for the Arab RegionStrategic Report
- Chief People Officers' Outlook — May 2026Strategic Report
- FBI PSA: Kali365 AI-Assisted Phishing-as-a-Service Platform Bypasses Microsoft 365 MFA via OAuth Device Code TheftVulnerability
- Megalodon Campaign Poisons 5,561 GitHub Repositories via CI/CD Workflow Backdoors — Credentials, Cloud Keys, and OIDC Tokens ExfiltratedVulnerability
- KPMG Singapore Launches EDB-Backed Trusted AI Centre of Excellence with AI Assurance FrameworkSolutions
- Anthropic Signals Future Public Release of Mythos-Class Vulnerability-Discovery AI; Glasswing Scan Finds 6,200+ Critical Flaws in 1,000+ Open-Source ProjectsSolutions
2026-05-25
- The Bank, FCA and HM Treasury Joint Statement on Frontier AI Models and Cyber ResilienceStrategic Report
- Agents, Robots, and Us: How AI Reshapes Work and Skills in EuropeStrategic Report
- CVE-2026-41863: LLM-Influenced Filename Reaches Path.resolve Before File Write in Spring AI Anthropic Skills API — Agentic Workflow Path TraversalVulnerability
- CVE-2026-45829 (ChromaToast): Pre-Authentication Remote Code Execution in ChromaDB Python FastAPI Server via Malicious HuggingFace Model LoadingVulnerability
- Anthropic Project Glasswing CVD Dashboard: Claude Mythos Preview Discloses 1,596 Vulnerabilities Across 281 Open-Source ProjectsSolutions
2026-05-24
- From Bans to Recalls: A Public Health Framework for AI Companion BotsStrategic Report
- Draft Commission Guidelines on the Classification of High-Risk AI SystemsStrategic Report
- Singapore Government AI Agents Sandbox: Indirect Prompt Injection Enabling RCE Confirmed as Top Production Agentic Attack VectorVulnerability
- CVE-2026-5194: WolfSSL Certificate Forgery (CVSS 9.1) — Part of 6,200+ AI-Discovered Flaws in Critical Open-Source InfrastructureVulnerability
- Cloud Security Alliance Releases AI Security Maturity Model (AISMM) v1.0 — 12-Category Programme-Level Framework for Enterprise AI SecurityGuidelines
- Singapore Government–Google AI Agents Sandbox: Indirect Prompt Injection Identified as Primary Cybersecurity Risk in Production Agentic DeploymentsGuidelines
- IMDA Publishes Model AI Governance Framework for Agentic AI v1.5 — Real-World Case Studies and Multi-Agent Risk ControlsGuidelines
- Anthropic Project Glasswing: Claude Mythos Preview Demonstrates Production-Scale AI-Driven Vulnerability DiscoverySolutions
2026-05-23
- Open-World Evaluations for Measuring Frontier AI CapabilitiesStrategic Report
- Loss of Oversight: How AI Systems May Become Harder to Audit, Monitor, and InvestigateStrategic Report
- Introducing RAMPART and Clarity: Open Source Tools to Bring Safety Into Agent Development WorkflowStrategic Report
- CISA Adds Drupal Core SQL Injection Vulnerability to KEV Catalog — CVE-2026-9082Vulnerability
- Google Integrates CodeMender Security Agent into Agent Platform EcosystemSolutions
- FTC Fines Cox Media Group $880,000 for False AI 'Active Listening' Advertising ClaimsRegulatory
- Trump Cancels Voluntary AI Model Vetting Executive Order Hours Before SigningRegulatory
2026-05-22
- LiteLLM AI Gateway Allows Authenticated Users to Escalate to proxy_admin via user_role Modification — CVE-2026-47102Vulnerability
- Trend Micro Apex One On-Premise Directory Traversal Enables Malicious Code Injection to Managed Endpoints — CVE-2026-34926 Added to CISA KEVVulnerability
- Langflow AI Workflow Platform Origin Validation Error Enables Cross-Origin Token Theft — CVE-2025-34291 Added to CISA KEVVulnerability
- MOSAIC Coalition Launches to Unify Fragmented AI Security Guidance Across Eight Leading OrganizationsGuidelines
- Expedia Preparing Model Context Protocol Server to Enable Direct AI Agent Access to Travel InventorySolutions
- Trump Administration Postpones AI Cybersecurity Executive Order Hours Before Scheduled SigningRegulatory
- The Private Capital Opportunity in AI-Enabled Climate and Sustainability SectorsStrategic Report
- The Need for Deeper, White-Box Access to Maintain State of the Art Evaluations for Loss of Control ThreatsStrategic Report
- Global AI Pulse: Q1 2026 — From Adoption to OrchestrationStrategic Report
2026-05-21
- Updated Model AI Governance Framework for Agentic AIStrategic Report
- Anthropic Silently Patched Claude Code Sandbox Bypass; Second Bypass in Five Months, No CVE IssuedVulnerability
- NVIDIA TRT-LLM Unsafe Deserialization Vulnerabilities Allow Code Execution, Data TamperingVulnerability
- CISA Adds Seven Known Exploited Vulnerabilities to KEV Catalog, Including Legacy Windows and Current Defender FlawsVulnerability
- KPMG and Anthropic Announce Global Alliance Embedding Claude into Digital Gateway for 276,000 EmployeesSolutions
- 1Password and OpenAI Launch MCP Server to Keep Coding Agent Credentials Out of Model ContextSolutions
- FTC Begins Enforcing Take It Down Act, Sends Warning Letters to 'Nudify' Tool OperatorsRegulatory
- Colorado Enacts Revised AI Act, Narrowing Scope to Automated Decision-Making TechnologyRegulatory
2026-05-20
- Frontier Risk Report (February to March 2026)Strategic Report
- SAP Introduces AI Agent Hub for Vendor-Agnostic Agent Governance Across Enterprise EcosystemsSolutions
- PwC Launches Agentic Scaffolding to Operationalize Enterprise AI Deployments at ScaleSolutions
- Ocean Emerges from Stealth with $28M for Agentic Email Security Platform Built on Autonomous InvestigationSolutions
- Google Launches Antigravity 2.0 Agentic Development Suite and Android CLI for AI-Driven App CodingSolutions
- MLflow Assistant CORS Vulnerability Enables Remote Code Execution on Developer Workstations — CVE-2026-2611Vulnerability
2026-05-19
- CVE-2026-45244: Summarize Browser Extension Missing Authorization Allows Agentic Automation Hijacking via Malicious ContentVulnerability
- CVE-2026-5163: Mattermost AI-Assisted Message Rewrite Channel Membership Bypass Enables Private Message DisclosureVulnerability
- CVE-2026-41947: Authorization Bypass in Dify Trace Configuration Endpoints Allows Tenant-Level Privilege EscalationVulnerability
- UK NCSC and Five Eyes Partners Publish Joint Guidance on Careful Adoption of Agentic AI SystemsGuidelines
2026-05-18
2026-05-17
- Corporate AI Risk Assessment Framework (Draft)Strategic Report
- 2026 Global AI in Finance: The Decision AdvantageStrategic Report
- CVE-2026-42945: Critical 18-Year-Old NGINX Heap Buffer Overflow Enables Unauthenticated RCEVulnerability
- CISA and Five Eyes Partners Release Joint Guidance on Agentic AI AdoptionGuidelines
- UK NCSC Publishes 10 Questions Framework for AI-Assisted Vulnerability DiscoveryGuidelines
- FTC Formalizes Take It Down Act Enforcement Beginning May 19, Sets $53,088 Per-Violation PenaltiesRegulatory
- UK Financial Regulators Issue Joint Statement Framing Frontier AI as Cyber Resilience ObligationRegulatory
2026-05-16
- Position: AI Security Policy Should Target Systems, Not ModelsStrategic Report
- Operationalizing Real-Time Monitoring of Clinical AIStrategic Report
- The State of AI in the Enterprise 2026Strategic Report
- An AI Taxonomy for Criminal Justice: Principled Use of AI in the Criminal Justice SystemStrategic Report
- Strategic Choices in the Age of AI: Shaping the Future of Life SciencesStrategic Report
- Multiple Authorization and Access Control Vulnerabilities in Open WebUI AI PlatformVulnerability
- WARD Guard Model Defends Web Agents Against Prompt Injection with Near-Perfect RecallVulnerability
- Researchers Propose Operating System Security Model for AI Agent FrameworksVulnerability
- Semantic Compliance Hijacking: Payload-less AI Agent Supply Chain Attack Evades Current ScannersVulnerability
- Experian and ServiceNow Partner to Embed Agentic AI in Enterprise WorkflowsSolutions
- FTC Begins Enforcement of Take It Down Act for Nonconsensual Deepfakes on May 19Regulatory
2026-05-15
- The Evaluation Differential: When Frontier AI Models Recognise They Are Being TestedStrategic Report
- Responsible Deployment of OpenClaw: Applying Singapore's Model AI Governance Framework for Agentic AIStrategic Report
- Building Resilient and Scalable AI Value Chains: A Nexus StrategyStrategic Report
- AI Infrastructure in the Age of Sovereignty: Requirements, Strategies and a Trusted Framework for Digital EmbassiesStrategic Report
- RMCP Rust SDK for Model Context Protocol Contains DNS Rebinding Vulnerability — CVE-2026-42559Vulnerability
- PyTorch Lightning Versions 2.6.2 and 2.6.3 Contain Credential Harvesting Mechanism — CVE-2026-44484Vulnerability
- OpenAI Confirms TanStack Supply-Chain Breach Affected Two Employee Devices, Code-Signing Certificates ExfiltratedVulnerability
- PraisonAI Authentication Bypass CVE-2026-44338 Exploited Less Than Four Hours After DisclosureVulnerability
- Akamai Acquires Browser-Based AI Security Firm LayerX for $205 MillionSolutions
2026-05-14
- Shadow-Aether Campaigns: Active AI Agent Attacks Across Latin America — Full-Chain Threat Automation Using Jailbroken ClaudeVulnerability
- AI Threats Expanding Attack Surface in Emerging Markets — New Attack Vectors Beyond Signature DetectionGuidelines
- UK AI Security Institute: Frontier Models Have Broken All Prior Trend Lines for Autonomous Cyber CapabilityGuidelines
- CISA and Five International Partners Release Agentic AI Adoption Guidance — Risk Mitigation Framework for Autonomous SystemsGuidelines
- Upwind AI Agentic Pack: Four-Agent Security Workforce for Cloud Threat Response and RemediationSolutions
- OpenAI Daybreak: Tiered Cybersecurity Initiative for Vulnerability Discovery and Patch ValidationSolutions
- Palo Alto Networks: 75 Vulnerabilities Found Using Claude Mythos Across 130+ Products — 15x Baseline Discovery RateSolutions
- Microsoft MDASH: Multi-Model Agentic Scanning Harness for Production-Scale Vulnerability DiscoverySolutions
- 32 US House Lawmakers Urge White House Action on Frontier AI Cyber Models — Congressional Letter on Mythos and Vulnerability DelugeRegulatory
- G7 Releases SBOM for AI Minimum Elements — Multi-Government Supply Chain Transparency FrameworkRegulatory
2026-05-13
- Dragos Discloses First LLM-Assisted Attack on Critical Water Infrastructure (OT) in MexicoVulnerability
- Mini Shai-Hulud Supply Chain Worm: 170+ Compromised Packages Across TanStack, Mistral AI, Guardrails AI, UiPathVulnerability
- Google Threat Intelligence: First Confirmed AI-Assisted Zero-Day Discovery and Exploitation in ProductionVulnerability
- Exaforce Raises $125M Series B for Agentic SOC Platform—Total Funding $200MSolutions
- Palo Alto Networks Launches Idira Identity Security Platform for Human, Machine, and AI Agent IdentitiesSolutions
- OpenAI Launches Daybreak—Frontier AI Vulnerability Detection PlatformSolutions
- Germany's BaFin Financial Regulator Mandates Targeted AI-Risk Cyber InspectionsRegulatory
- Japan PM Orders Cabinet-Level Cybersecurity Strategy Review in Response to Mythos ThreatRegulatory
- AI in the Breach: How an Adversary Leveraged AI to Target a Water Utility's OTStrategic Report
- Intelligent Infrastructure: A PrimerStrategic Report
2026-05-12
- AI in Nursing Practice: Consensus Report from the American Nurses Association Think TankStrategic Report
- Physical AI: Taking Human-Robot Collaboration to the Next LevelStrategic Report
- Global AI in Finance 2026: The Decision AdvantageStrategic Report
- Malicious Hugging Face Model Repository (Open-OSS/privacy-filter)—AI Supply Chain Attack Delivers Infostealer Malware, 244K DownloadsVulnerability
- UK NCSC Issues Guidance: '10 Questions to Ask When Using AI Models to Find Vulnerabilities'Guidelines
- Google Disrupts First Confirmed AI-Generated Zero-Day Exploit—2FA Bypass Developed by Cybercrime SyndicateSolutions
- Lyrie.ai Releases Agent Trust Protocol (ATP)—First Open Standard for AI Agent Cryptographic VerificationSolutions
2026-05-11
- Attackers Weaponize Claude.ai Shared Chats in Active Malvertising Campaign to Distribute macOS InfostealerVulnerability
- Ollama Heap Out-of-Bounds Read (CVE-2026-7482 'Bleeding Llama') — Critical Memory Leak in 300k+ DeploymentsVulnerability
- Anthropic Research: Constitutional Training Eliminates Agentic Misalignment Blackmail in ClaudeGuidelines
- Anthropic Claude "Dreaming" Feature for Agentic Memory ReviewSolutions
- Responsible AI Needs More Than Good IntentionsStrategic Report
- Making AI Productivity Deliver Real ValueStrategic Report
- CEOs and Boards Are Aligned on AI in Theory, but Divided in Practice (Split Decisions: The BCG CEOs and Boards Survey)Strategic Report
2026-05-10
- Onyx AI platform chat-session authorization bypass allows denial of active LLM generationVulnerability
- LiteLLM AI gateway SQL injection exposes proxy database and managed credentialsVulnerability
- Spring AI MilvusVectorStore filter-expression injection allows vector database manipulationVulnerability
- US explores FDA-style pre-release AI testing after Mythos capabilities alarm White HouseRegulatory
2026-05-09
- Context-maxxing: A path to cognitive agency with generative AIStrategic Report
- Heimdallr Framework Detects LLM-Induced Security Risks in GitHub CI WorkflowsVulnerability
- Multiple Critical Vulnerabilities in PraisonAI Multi-Agent FrameworkVulnerability
- ClaudeBleed: Chrome Extension Vulnerability Allows Hijacking of Anthropic's AI AgentVulnerability
- CVE-2026-42208: Critical SQL Injection in LiteLLM Added to CISA KEVVulnerability
- Cloud Security Alliance: AI Agent Identity Architectures Are Being Built BackwardsGuidelines
- Cobalt Pentesting Report: AI Systems Show 2.5x Higher Severe-Flaw Density Than Legacy AppsSolutions
- ASIC Urges Urgent AI Cybersecurity Action for Australian Financial SectorRegulatory
2026-05-08
- 2026 Work Trend Index: Agents, human agency, and the opportunity for every organizationStrategic Report
- Outlooks on Open-Source Innovation at the India AI Impact Summit 2026Strategic Report
- The Geopolitical Debates Over Controlling Cloud ComputeStrategic Report
- Gemini CLI Indirect Prompt Injection Enables Supply Chain Compromise (CVSS 10.0)Vulnerability
- Cline Kanban WebSocket Hijacking Enables AI Agent Takeover (CVSS 9.7)Vulnerability
- Claude Code 'TrustFall' Supply Chain Attack via Malicious RepositoriesVulnerability
- Claude Code MCP OAuth Token Theft via Man-in-the-Middle AttackVulnerability
- Ivanti EPMM Remote Code Execution Added to CISA KEV CatalogVulnerability
- EU Delays AI Act High-Risk System Rules to December 2027Regulatory
2026-05-07
- NanoClaw Container Agent Filesystem Boundary Vulnerability Enables Host Read/WriteVulnerability
- Copy Fail: Kubernetes Container Escape via Page Cache Corruption Added to CISA KEVVulnerability
- CISA Adds Palo Alto PAN-OS Zero-Day to KEV Following Active ExploitationVulnerability
- FIS Partners with Anthropic to Deploy Agentic AI for Financial Crime InvestigationsSolutions
- Apple Settles $250M Class-Action Lawsuit Over Delayed AI Features in SiriRegulatory
2026-05-06
- Open-Weight AI Models Require Proportional Evaluation ApproachesStrategic Report
- Our evaluation of OpenAI's GPT-5.5 cyber capabilitiesStrategic Report
- Empowering Defenders: AI for CybersecurityStrategic Report
- Critical Input Validation Vulnerability in OpenClaw Agent Platform Allows Trust EscalationVulnerability
- OWASP Unveils Strategic Plan for Eliminating Insecure SoftwareGuidelines
- Kentucky Files First State Enforcement Action Against AI Chatbot ProviderRegulatory
- U.S. Formalizes Pre-Deployment AI Security Testing with Frontier LabsRegulatory
2026-05-05
- 2026 Government AI Landscape AssessmentStrategic Report
- Risk and Compliance 2026: Refining Oversight for a Volatile, AI-Driven WorldStrategic Report
- The AI Jobs Transition Framework: Mapping AI's Near-Term Impact on JobsStrategic Report
- Split Decisions: The BCG CEOs and Boards Survey — AI Governance GapsStrategic Report
- Technology Convergence: The New Logic for Competitive Advantage (2026 Edition)Strategic Report
- Artificial Intelligence Index Report 2026Strategic Report
- Multiple Critical Vulnerabilities in n8n Workflow Automation MCP OAuth ImplementationVulnerability
- Critical Heap Out-of-Bounds Vulnerability in Ollama Enables Arbitrary Code ExecutionVulnerability
- CISA and International Partners Publish Joint Advisory on Secure Agentic AI DeploymentGuidelines
- OpenAI Launches Advanced Account Security for High-Risk ChatGPT UsersSolutions
- Anthropic Partners with Blackstone, Hellman & Friedman, and Goldman Sachs on $1.5B Enterprise AI Services VentureSolutions
- Cisco to Acquire Astrix Security, Extending Zero Trust to Non-Human Identities and AI AgentsSolutions
2026-05-04
- Summary Analysis of Responses to the Request for Information Regarding Security Considerations for AI AgentsStrategic Report
- Claude Opus AI Agent Deletes Production Database in 9 Seconds After Misinterpreting CredentialsVulnerability
- CISA Adds Actively Exploited Linux Root Access Bug CVE-2026-31431 to KEV CatalogVulnerability
- Pentagon Completes AI Integration Agreements with Seven Tech Giants for Classified NetworksSolutions
2026-05-03
2026-05-02
2026-05-01
- Global Economic and Financial Implications of Artificial IntelligenceStrategic Report
- Strengthening Information Integrity: Advertising, Artificial Intelligence and the Global Information CrisisStrategic Report
- 2026 Global AI in Financial Services Report – Adoption, Impact and RisksStrategic Report
- Critical Agent Hijacking in Google Gemini CLI Enabled Pre-Sandbox RCE and Supply Chain AttacksVulnerability
- cPanel/WHM Authentication Bypass Enables Unauthenticated Admin Access — Active Exploitation ConfirmedVulnerability
- CSA Releases AARM Framework for Securing Agentic Runtime EnvironmentsGuidelines
- NIST Publishes Analysis of Industry Responses to AI Agents Security RFIGuidelines
- OpenAI Launches Advanced Account Security with Phishing-Resistant AuthenticationSolutions
2026-04-30
- Evaluating whether AI models would sabotage AI safety researchStrategic Report
- How Agentic AI Will Reshape PaymentsStrategic Report
- 2026 Transparency Report on Foundation Model ImpactsStrategic Report
- Path Traversal Vulnerabilities Disclosed in Multiple MCP Server ImplementationsVulnerability
- Critical RCE in DocsGPT via MCP Test Bypass Patched in Version 0.16.0Vulnerability
- Dual RCE Vulnerabilities in Ollama Windows Update Mechanism Enable Code ExecutionVulnerability
- Critical LiteLLM AI Gateway Vulnerability Exploited 36 Hours After Public DisclosureVulnerability
- OpenAI Expands AI-Assisted Cyber Defense Access to All Vetted Government LevelsSolutions
- FIDO Alliance Launches Agentic Authentication Standards Initiative for AI-Driven CommerceSolutions
2026-04-29
2026-04-28
- Black Hat Asia keynote assesses practical limits of agentic offensive security automationVulnerability
- Microsoft patches privilege escalation flaw in Entra ID Agent ID Administrator role affecting agentic identitiesVulnerability
- China blocks Meta's $2 billion acquisition of agentic AI startup ManusRegulatory
2026-04-26
2026-04-25
- China's 360 Digital Security Claims ~1,000 AI-Discovered Vulnerabilities, Echoing Mythos CapabilitiesVulnerability
- Palo Alto Unit 42 Demonstrates Autonomous AI Agent 'Zealot' Executing Full Cloud Attack ChainVulnerability
- Anthropic and NEC Partner to Deploy Claude to 30,000 Employees in JapanSolutions
- Security Leaders Converge on AI-as-Identity Framework to Manage Agentic RiskSolutions
- Copperhelm Emerges with $7M Seed for Agentic Cloud Security PlatformSolutions
2026-04-24
- Palo Alto Unit 42 Demonstrates Autonomous Multi-Agent Cloud Attack with 'Zealot' PoCVulnerability
- Cisco Discloses Memory Persistence Vulnerability in Claude Code via NPM Post-Install HooksVulnerability
- Forcepoint Discloses 10 In-the-Wild Indirect Prompt Injection Payloads Targeting AI AgentsVulnerability
- CISA and 13 International Partners Issue Joint Guidance on China-Nexus Covert NetworksGuidelines
- Google Introduces Agent Identity Management and Security Dashboard at Cloud Next '26Solutions
- Anthropic Mythos Model Accessed by Unauthorized Users Despite Limited PreviewRegulatory
- White House Accuses China of 'Industrial-Scale' AI Model Distillation CampaignsRegulatory
2026-04-23
- U.S. Lawmakers Briefed on Jailbroken AI Models Generating Detailed Attack Plans in SecondsVulnerability
- 65% of Enterprises Experienced AI Agent Security Incidents, 82% Discovered Undiscovered AgentsVulnerability
- UK NCSC Warns of Cybersecurity 'Perfect Storm' and Announces £90m Resilience InvestmentGuidelines
- OpenAI Releases Privacy Filter Model for PII Detection and RedactionSolutions
- Anthropic Launches Project Glasswing with Claude Mythos for Defensive Vulnerability DiscoverySolutions
2026-04-22
- Multiple AI-Related CVEs Disclosed in NVD Database Affecting LLM ApplicationsVulnerability
- Mozilla Reports Mythos AI Discovered 271 Security Vulnerabilities in Firefox 150Vulnerability
- Anthropic's Mythos AI Model Accessed by Unauthorized Users in Security BreachVulnerability
- Cloud Security Alliance Reports Two-Thirds of Organizations Suffer AI Agent Cybersecurity IncidentsGuidelines
- Cybersecurity Insiders Report: 92% of Organizations Lack Visibility Into AI IdentitiesGuidelines
- ServiceNow Completes $7.7B Acquisition of Armis Security, Expanding AI-Driven Security PlatformSolutions
2026-04-21
- AI Engine Pro Authorization Bypass VulnerabilityVulnerability
- Vercel Supply Chain Breach via Context.ai AI Tool CompromiseVulnerability
- Google Antigravity AI Agent Manager Sandbox Escape VulnerabilityVulnerability
- Analysis: AI-Powered Security Operations Required for Supply Chain DefenseGuidelines
- AI and CMMC Compliance Guidance for Defense ContractorsGuidelines
- Google AI Systems Block 8.3 Billion Malicious Ads Using Gemini AnalysisSolutions
- WitFoo Releases 114M-Record Enterprise Attack Traffic DatasetSolutions
- Brain Corp Achieves SOC 2 Type II Certification for AI Robotics PlatformSolutions
- Adobe Launches AI Suite for Enterprise Marketing AutomationSolutions
- NSA Reportedly Using Anthropic's Mythos Despite Pentagon Supply Chain Risk DesignationRegulatory
- Global Regulators Monitor Anthropic's Mythos for Banking System RisksRegulatory
- Singapore Champions Global AI Testing Standardization InitiativeRegulatory
2026-04-20
- Path Traversal Vulnerability in Awesome-LLM-Apps Beifong AI AgentVulnerability
- OWASP Releases AI Security Solutions Landscape for Agentic AI Q2 2026Guidelines
- MCP Protocol Positions Agentic AI for Production Deployment in 2026Solutions
- Anthropic's Mythos Model Drives Wave of Critical Security UpdatesSolutions
- China Introduces Comprehensive AI Companion Regulations for $30B MarketRegulatory
- NSA Uses Anthropic's Mythos Despite Pentagon Supply Chain Risk DesignationRegulatory
- Trump Administration Expands Multi-Front Campaign Against State AI RegulationRegulatory
- EU Criticizes Lack of Regulatory Engagement on Anthropic's Mythos ModelRegulatory
2026-04-19
- Tycoon 2FA Phishing-as-a-Service Platform Disrupted, Attack Volume IncreasesVulnerability
- Asante Babers Consulting Releases Digital Transformation Cybersecurity GuidelinesGuidelines
- Samsung Expands Galaxy AI Features Through One UI 8.5 Beta RolloutSolutions
- Meta Announces AI-Driven Workforce Restructuring with 8,000 Job CutsSolutions
- Ring-a-Ding Launches OpenClaw Skill for AI Agent TelephonySolutions
- Anthropic CEO Meets White House Officials Amid Pentagon Supply Chain DisputeRegulatory
2026-04-18
- Apache ActiveMQ Classic Remote Code Execution Actively ExploitedVulnerability
- pandas-ai SQL Injection Vulnerability in Query ExecutionVulnerability
- LangChain LangSmith Studio URL Parameter Injection VulnerabilityVulnerability
- Three Microsoft Defender Zero-Days Actively Exploited by Threat ActorsVulnerability
- Analysis: AI Amplifies Every Traditional Vulnerability RiskGuidelines
- Industry Report: AI Models Show Rapid Progress in Vulnerability ResearchGuidelines
- NIST Releases AI RMF Profile Concept for Critical InfrastructureGuidelines
- Google AI Blocked 8.3 Billion Malicious Ads Using Enhanced DetectionSolutions
- American Express Acquires Agentic AI Fintech HyperSolutions
- Loop Raises $95M Series C for AI-Powered Supply Chain Security PlatformSolutions
- DeepL Launches Real-Time Voice-to-Voice Translation SuiteSolutions
- Illinois Lawmakers Advance AI Regulation InitiativesRegulatory
- HIMSS Advocates for Consistent Nationwide AI Regulation in HealthcareRegulatory
- White House Holds 'Productive' Meeting with Anthropic CEO Over Mythos AI ModelRegulatory
2026-04-17
- CVE-2026-35639: Critical Privilege Escalation in OpenClaw AI Agent Platform (CVSS 8.7)Vulnerability
- 'Comment and Control': Prompt Injection via GitHub Comments Compromises Claude Code, Gemini CLI, and GitHub CopilotVulnerability
- OpenAI Releases GPT-5.4-Cyber for Vetted Security ProfessionalsSolutions
- Mozilla Launches Thunderbolt: Open-Source, Self-Hosted Enterprise AI ClientSolutions
- China CAC Issues Interim Measures on Anthropomorphic AI Interaction ServicesRegulatory
2026-04-16
- AI Vulnerability Storm: Mean Time-to-Exploit Collapses from 2.3 Years to Under 24 HoursVulnerability
- ShareLeak / PipeLeak: Patched Microsoft Copilot Studio and Salesforce Agentforce Still Exfiltrate Data via Prompt InjectionVulnerability
- MCPwn: nginx-ui Authentication Bypass Enables Unauthenticated MCP Server TakeoverVulnerability
- Flowise AI Agent Builder Under Active CVSS 10.0 RCE Exploitation — 12,000+ Instances ExposedVulnerability
- SANS/CSA/OWASP Emergency Briefing: 'Building a Mythos-Ready Security Program'Guidelines
- FireTail AISPM Analysis: 90% of Enterprise AI Usage is Shadow AI, Posing Critical Data Leakage RiskSolutions
- Semgrep April 2026 AppSec Platform: AI-Powered IDOR and Broken Auth Detection in BetaSolutions
- OpenAI Launches GPT-5.4-Cyber for Vetted Security ProfessionalsSolutions
- IAPP 2026 Global Summit: Global Regulators Signal Shift to Operational AI EnforcementRegulatory
- China CAC Issues Draft Measures for Human-Like Interactive AI ServicesRegulatory
- Fitch Ratings Warns AI Vulnerability Discovery Compresses Cyber Insurance RiskRegulatory
2026-04-15
- Flowise AI Agent Builder Under Active CVSS 10.0 RCE Exploitation (CVE-2025-59528)Vulnerability
- LiteLLM RCE via Bytecode Rewriting (CVE-2026-40217)Vulnerability
- Exploit Timeline Compression — Mean Time-to-Exploit Falls Below 24 HoursVulnerability
- Project Glasswing CVE Transparency Gap — Only One Confirmed CVE Despite 'Thousands' ClaimedVulnerability
- KPMG and INSEAD Launch Global AI Board Governance PrinciplesGuidelines
- SANS, CSA, and OWASP Release Emergency 'Mythos-Ready' Strategy Briefing for CISOsGuidelines
- Fortreum Acquires Kovr.AI to Create AI-Native Compliance PlatformSolutions
- OpenAI Launches GPT-5.4-Cyber for Vetted Security ProfessionalsSolutions
- California Cybersecurity Audit Rule Analysis Highlights AI Compliance BurdenRegulatory
- EU AI Act Full Enforcement Approaches with Member State Readiness GapsRegulatory
2026-04-14
- LiteLLM Security Hardening After March 2026 Supply Chain CompromiseVulnerability
- Large Reasoning Models Achieve 97% Jailbreak Success as Autonomous AttackersVulnerability
- Agentic AI / MCP Exposure: 43% of Public MCP Servers Vulnerable, 30 CVEs in 60 DaysVulnerability
- Prompt Injection: OWASP #1 LLM Risk Continues to Surge in 2026Vulnerability
- CVE-2026-22778: Critical vLLM RCE via Crafted Video InputVulnerability
- CVE-2026-39987: Pre-Auth RCE in Marimo Python NotebookVulnerability
- CVE-2026-39426: Stored XSS in LLM Application Frontend (MdRenderer.vue)Vulnerability
- NIST Cybersecurity Framework Profile for AI (Cyber AI Profile / NISTIR 8596)Guidelines
- Cynomi Launches AI Insights and Co-Worker AgentsSolutions
- Anthropic Project Glasswing + Claude Mythos Preview for Zero-Day DiscoverySolutions
- NIST AI Agent Standards Initiative Listening Sessions BeginRegulatory
- NIST AI RMF Profile for Trustworthy AI in Critical InfrastructureRegulatory
2026-04-12
- LiteLLM Supply Chain Attack Compromises 500K+ Machines via PyPI PoisoningVulnerability
- MCP Library SSRF via JSON Schema Mishandling (CVE-2026-39885)Vulnerability
- Red Hat OpenShift AI Token Disclosure (CVE-2026-5483)Vulnerability
- Marimo Pre-Auth RCE Exploited Within 10 Hours (CVE-2026-39987)Vulnerability
- Sockpuppeting: Universal Single-Line Jailbreak Affects 11 Major LLMsVulnerability
- aws-mcp-server Critical Command Injection (CVE-2026-5058)Vulnerability
- MITRE ATLAS 2026 Update Expands Agentic AI Threat TaxonomyGuidelines
- Google DeepMind Maps Six Categories of Web-Based AI Agent TrapsGuidelines
- OWASP Publishes Top 10 for Agentic Applications 2026Guidelines
- Cynomi Launches AI Co-Worker Agents to Scale Cybersecurity ExpertiseSolutions
- OPSWAT Predictive AI Engine Achieves 99.99% Pre-Execution Threat DetectionSolutions
- Agentic AI Security Readiness Gap: 88% Report Incidents, Only 29% Feel SecureSolutions
- Anthropic Launches Project Glasswing with Claude Mythos Preview for Defensive Vulnerability HuntingSolutions
- U.S. Courts Impose $145K+ in Sanctions for AI-Generated Legal CitationsRegulatory
- China Issues Trial Guidelines on AI Ethics ReviewRegulatory
- China Finalises Interim Measures on Anthropomorphic AI Interaction ServicesRegulatory
2026-04-11
- Adversa AI Audit: 93% of AI Agent Frameworks Use Unscoped API Keys, 0% Have Per-Agent IdentityVulnerability
- Google DeepMind Publishes 'AI Agent Traps' Taxonomy: Six Attack Categories Against Autonomous AgentsVulnerability
- CVE-2026-39987: Marimo Pre-Auth RCE Exploited Within 10 Hours of DisclosureVulnerability
- NIST AI Agent Standards Initiative Begins April Listening SessionsGuidelines
- CIS Report: Prompt Injections — The Inherent Threat to Generative AIGuidelines
- LayerX Research: AI Browser Extensions Are the Most Dangerous Unmonitored AI Threat SurfaceSolutions
- Anthropic Project Glasswing: Claude Mythos Preview Discovers Thousands of Zero-DaysSolutions
- South Korea Overhauls Personal Information Protection Act with 10% Turnover FinesRegulatory
- Washington and Oregon Enact AI Companion Chatbot Safety LawsRegulatory
- US AI Enforcement Accelerates via State Laws as Federal Policy StallsRegulatory
2026-04-10
- MCP Protocol Tool Poisoning via Malicious Server RegistrationVulnerability
- NIST Releases AI Agent Security Companion GuideGuidelines
- Microsoft Launches AI Security Posture Management for AzureSolutions
- Singapore IMDA Updates AI Governance Framework for Financial ServicesRegulatory
- EU AI Office Publishes High-Risk AI Enforcement TimelineRegulatory
Glossary terms
- AI 'ghost citations' (hallucinated legal precedents)
- AI AGENT Act
- AI Act Market Surveillance Authority
- AI Act intimate-image prohibition (Art. 5)
- AI Agent Accountability Act
- AI Agent Backup & Recovery (Agent Resilience)
- AI Agent Endpoint Forensics
- AI Bill of Materials (AI-BOM)
- AI CI/CD pipeline compromise
- AI Circuit Breaker (runtime interceptor)
- AI Companion Chatbot Regulation
- AI Content Provenance (Watermarking)
- AI Cyber Defense Act
- AI Cybersecurity Clearinghouse
- AI Data Center Security Standard
- AI Detection and Response (AI-DR)
- AI Governance Audit-Readiness
- AI Liability Legal Statement
- AI Misalignment Self-Disclosure
- AI Red Lines (International Governance)
- AI Regulatory Sandbox
- AI Safety Index
- AI System Compromise
- AI Training Data Consent Exemption
- AI Vulnerability Taxonomy
- AI Washing
- AI agent
- AI agent behaviour analytics (UEBA for AI)
- AI agent conflict of interest
- AI agent corporate entity (AI Company)
- AI agent governance control plane
- AI agent identity layer
- AI agent kill-switch (human override mandate)
- AI agent skill (plugin)
- AI brand spoofing
- AI chatbot liability
- AI coding agent supply chain attack
- AI conversation transcript leakage
- AI data centre consenting (grid connection)
- AI export controls
- AI gateway
- AI gateway compromise
- AI governance measurement gap
- AI hallucination
- AI honeypot
- AI incident investigation playbook
- AI incident response system of record
- AI infrastructure moratorium
- AI medical device lifecycle monitoring (post-market surveillance)
- AI memory poisoning
- AI model evaluation cheating / evaluation sandbagging
- AI model fine-tuning
- AI model identity & authentication
- AI model security flaw disclosure framework
- AI model transparency & disclosure regime
- AI outbound-investment screening (Outbound Investment Security Program)
- AI red-teaming
- AI runtime firewall
- AI scanner evasion
- AI security framework harmonization (MOSAIC)
- AI supply chain attack
- AI supply-chain insurance gap
- AI telemetry data leakage
- AI web crawler transparency mandate
- AI workload microsegmentation
- AI-Enabled Insider Threat (Fake Worker Infiltration)
- AI-Enhanced Attacks
- AI-Native Purple Teaming
- AI-accelerated vulnerability discovery
- AI-assisted malware triage poisoning
- AI-enabled fraud
- AI-generated CSAM reporting
- AI-generated content disclosure (EU AI Act Article 50)
- AI-resistant software protection
- AIUC-1 certification (CSA STAR for AI agents)
- API key exposure
- Adversarial reasoning extraction
- Agent Control Standard (ACS)
- Agent Fabric
- Agent Harness
- Agent audit trail
- Agent command approval bypass
- Agent containment (execution isolation)
- Agent control plane compromise
- Agent forgery
- Agent identity & access management (agent IAM)
- Agent identity document
- Agent mesh (peer-to-peer agent network)
- Agent misalignment detection
- Agent orchestrator
- Agent permission receipts (signed permissions & action receipts)
- Agent policy gate (tool-call authorization point)
- Agent session-state poisoning (checkpoint restore attack)
- Agent trust registry
- Agent webhook attack surface
- Agent-phishing
- Agent-to-agent (A2A) communication security
- Agentic AI governance framework (government)
- Agentic AI trust and identity standards (ITU FG-TIDA)
- Agentic Commerce
- Agentic Detection and Response (ADR)
- Agentic Identity Provider
- Agentic SOC (AI-driven security operations)
- Agentic Vulnerability Discovery Harness
- Agentic access control
- Agentic attack orchestration (AI attack industrialisation)
- Agentic browser attack
- Agentic memory (persistent agent memory)
- Agentic pipeline
- Agentic ransomware
- Agentic regulator (AI-enabled supervision)
- Agentic vulnerability remediation
- Agentjacking
- Algorithmic Pricing Collusion
- American AI Security Act
- Approval gate (agentic AI)
- Argument injection
- Article 55 systemic-risk enforcement
- Automated decision-making (ADM)
- Autonomous AI attack chain
- Autonomous offensive security agent
- Autonomous vulnerability validation
- Autonomous weapons human control mandate
- Ban Artificial Superintelligence Act
- Behavioral detonation (dynamic AI-skill sandboxing)
- Benchmark saturation
- Binding Operational Directive (BOD)
- Biodesign audit trail (AI)
- Biosecurity software barrier
- Blind prompt injection
- Broken Object Level Authorization (BOLA)
- Broken access control (AI plugins & chatbots)
- Browser-Use Agent (Computer-Use AI)
- CI/CD pipeline compromise (AI projects)
- CIS MCP Server Benchmark
- CISA Known Exploited Vulnerabilities (KEV) Catalog
- Canada Bill C-34 AI chatbot safety duties
- Carolina Principles
- Catastrophic risk threshold (pre-mitigation capability)
- Center for AI Standards and Innovation (CAISI)
- Chain-of-Thought Reasoning Trace Exposure
- Chain-of-thought hijacking
- Chain-of-thought monitorability
- China agentic AI regulatory framework
- Cloud Metadata Endpoint Exploitation
- Cloud export-control loophole (third-country AI compute)
- CoT forgery (chain-of-thought forgery)
- Colorado AI Act (SB 26-189) — automated decision-making accountability
- Companion AI / emotional AI chatbot regulation
- Compute Derivatives
- Confidential Computing (for AI)
- Confused deputy attack
- Connecticut SB 5 (AI Act)
- Context compaction injection
- Continuous Threat Exposure Management (CTEM)
- Convention 108+
- Coordinated vulnerability disclosure (CVD) for open-source AI
- Covert AI-agent data exfiltration (LLMLeak / Innocent Courier)
- Credential plaintext exposure (AI platforms)
- Critical cybersecurity capability threshold
- Critical infrastructure AI risk mandate
- Cross App Access (XAA)
- Cross-Site Request Forgery (CSRF) Against AI Agents
- Cross-industry AI incident sharing (SAFE)
- Cross-origin agent execution
- Cross-session prompt injection
- Cross-tenant authorization bypass
- Cross-tenant data exposure
- Cryptographic Context Injection
- Cryptography Bill of Materials (CBOM)
- Cyber Trust Mark (critical infrastructure AI mandate)
- Cyber-verification model access (safe weak-safeguard access)
- Cybersecurity-Specialized AI Model
- DNS rebinding attack (AI agents)
- Data Broker Regulation (AI Training Data)
- Data and model poisoning
- Data localisation vs. AI governance
- Data loss prevention (DLP) for agentic AI
- Decompression bomb (AI inference denial of service)
- Deepfake consent violation (privacy enforcement)
- Deepfake-Resistant Identity Verification
- Denial of wallet (AI token exhaustion)
- Denial-of-service (DoS) attack on AI agents
- Dependency Cooldown
- Dependency confusion attack
- Deployment simulation (pre-release AI behaviour testing)
- Digital Arrest Fraud (AI-Enabled)
- Dual-use AI capability
- Dynamic agent risk (adaptive agent access)
- EU AI Act Omnibus amendment
- EU AI Act harmonised standard
- EU AI Act risk tiers
- EU Kids Act
- EU data centre sustainability label
- Embedded browser (AI coding assistant) attack surface
- Emergent deceptive AI behavior
- Emergent multi-agent collusion
- Enterprise Agent Security Baseline
- Evaluation Sandbox Escape
- Excessive agency (AI agents)
- Explainable AI (XAI) for Security Operations
- FSB Sound Practices for Responsible Adoption of AI
- FTC output manipulation enforcement risk
- Fail-Open Design Flaw
- Federal AI procurement clause
- Federal preemption of state AI laws
- Federated Learning
- Frontier AI incident reporting mandate
- Frontier model
- GRID Requirements (AI Data Center Permitting Mandate)
- GenAI data protection impact assessment
- Great American AI Act (GAAIA)
- Guardrail incompleteness theorem (limits of AI safety controls)
- Guardrail-based vulnerability discovery
- Guardrails
- Hallucination squatting (slopsquatting)
- Hardcoded AI instruction file (attack vector)
- Header exfiltration (AI agent credential theft via redirects)
- High Bandwidth Memory (HBM) chokepoint
- Hop-aware authorization (multi-agent least privilege)
- IDOR (Insecure Direct Object Reference) in AI platforms
- ISO/IEC 27090 (AI security threats standard)
- In-silicon agent governance
- Independent Verification Organization (IVO)
- Indirect prompt injection
- Insecure model deserialization (malicious pickle files)
- Intent token
- Intent-to-permission mismatch
- International AI Evaluation Standard (NAAIMES)
- International AI incident reporting
- Jailbreak severity benchmark
- LLMOps platform
- LLMjacking
- Large Reasoning Model (LRM)
- Least agency (principle)
- Lethal trifecta (AI agents)
- Lifecycle-Hook Supply Chain Attack
- Localhost trust boundary attack
- Log-forging prompt injection
- Long-Horizon AI Agent
- Loopjacking
- MCP gateway (agent tool access control)
- MCP server
- MCP server authentication bypass
- MITRE ATLAS
- Malicious API spec (AI agent code-generation injection)
- Malicious model / poisoned AI model
- Mastra npm scope takeover
- Mind Virus (Self-Propagating AI Idea)
- Misaligned AI insider risk
- Misuse safeguards vs. loss-of-control safety
- Model Context Protocol (MCP)
- Model distillation (unauthorized)
- Model registry supply-chain attack
- Model risk management (AI/ML)
- Model weight security
- Multi-Agent Turf War
- Multi-agent data leakage
- Multi-tenant AI infrastructure risk
- Multimodal Input Attack Surface
- N-day exploit acceleration (AI-compressed patch window)
- NIST Cyber AI Profile
- National AI Charter Act
- National Security Presidential Memorandum (NSPM-12)
- Neuro-AI governance
- No Robo Bosses Act (California SB 947)
- Non-human identity (NHI)
- OAuth token hijacking (AI agent context)
- OS command injection (AI tooling)
- OT/ICS-native AI security
- OWASP Agentic Skills Top 10 (AST10)
- OWASP GenAI LLM Top 10
- Open-weight AI model
- Open-weight model risk management
- Overthinking attack (reasoning-weight amplification)
- Path traversal (AI agents)
- Policy bypass (AI agent trust policies)
- Post-quantum cryptography (PQC) migration
- Pre-deployment AI evaluation (third-party)
- Pretraining poisoning (data-poisoning backdoor)
- Privacy-preserving safety monitoring
- Privilege escalation (AI context)
- Prompt forcing
- Prompt injection
- Prompt injection in production (confirmed enterprise breaches)
- Prompt injection taxonomy
- Proof-of-Control
- Reasoning-extension denial-of-service (ReDoS for AI)
- Recursive self-improvement
- Removal of independent agency protections (Humphrey's Executor reversal)
- Request smuggling (AI infrastructure)
- Retrieval-Augmented Generation (RAG)
- Rogue AI agent
- Runtime agent kill switch
- Runtime safeguards for agentic finance
- SQL injection via AI endpoints
- Same-origin policy bypass
- Sandbox escape (AI coding agents)
- Secure by Design (AI-specific software assessment)
- Secure-enclave model evaluation
- Self-replicating AI
- Semantic cache
- Server-Side Template Injection (SSTI) in AI Pipelines
- Server-side request forgery (SSRF) in AI platforms
- Shadow AI
- Shadow AI discovery
- Software Bill of Materials (SBOM) for AI
- Sovereign AI
- Stealer malware targeting AI developer credentials
- Stop Rogue AI Act
- Stored cross-site scripting (Stored XSS)
- Supply chain risk designation
- Symlink following attack
- Synthetic Voice Agent Fraud (AI Vishing)
- Synthetic media / deepfakes
- Synthetic performer disclosure (AI advertising law)
- System card
- System prompt leakage
- Systemic AI risk contagion
- TEVV (Test, Evaluation, Verification & Validation)
- TRACE (Trust, Runtime Attestation and Compliance Evidence)
- Third-Party AI Model Vendor Oversight
- Tool poisoning (MCP / agentic pipelines)
- Trojanized AI agent skill
- Trusted identity propagation (per-user data access for AI agents)
- Trusted-platform abuse (AI feature hijacking)
- Two-gate authorization (agent tool access vs. use)
- UN Independent Scientific Panel on AI
- Unauthenticated remote code execution (RCE)
- Unauthorized practice of law by AI
- Universal jailbreak
- Vector store injection
- Verifiable Digital Credential (VDC)
- Very Large Online Search Engine (VLOSE) designation
- Vibe coding
- Vietnam AI risk-based classification (Decree 142)
- Virtual Patching
- Vulnerability Reachability Analysis
- White House Super Intelligence Force
- Workflow Identity Hijacking
- World AI Cooperation Organization (WAICO)
- World model (spatial AI)
- Worm / self-propagating malware targeting AI toolchains
- Zero Data Retention (ZDR)
- Zero data-egress design
- Zero standing privilege (AI agents)
- Zero-click prompt injection
- Zero-trust for AI agents
- Zombie agent
- prEN 18282