What happened
Cyera announced (Aug 18, 2026) Agent Guardian, a control plane that discovers agents across cloud/SaaS/endpoint/shadow-AI, maps data/tool access, blocks risky tool calls or strips sensitive response fields at runtime, and runs continuous adversarial red-teaming; Cyera Endpoint extends this to local tools (bash, run-code, MCP) on developer machines including Claude Code and ChatGPT Desktop.
Why it matters
Extends data-security-first AI-SPM into runtime enforcement and endpoint coverage, closing a gap most agent-security tools (prompt/output-only monitoring) leave open, and is already integrated as an Anthropic Inference Hooks partner.
Applicability
Security and data-governance teams with heavy AI-agent/Shadow-AI exposure across SaaS and endpoints should evaluate now; GA per vendor announcement.