What happened
On Sept 30, 2026 BlackFog launched ADX Vision 2.0, expanding its anti-data-exfiltration AI governance product on Windows/macOS endpoints. It adds seven layers of prompt protection (context-switching, obfuscation, jailbreak, data-exfiltration, dangerous-pattern, fuzzy-keyword and structural-anomaly detection) applied to both human and agent-generated prompts, redirect-to-sanctioned-LLM enforcement, optional prompt auditing for frontier LLMs, and a unified token-consumption view to flag token theft.
Why it matters
One of the first endpoint-native controls to govern agentic AI prompts (not just employee chat), filling a gap where DLP/CASB/EDR do not inspect what an autonomous agent sends to LLMs. It extends an existing anti-exfiltration vantage point into the agent security layer without requiring new network/API plumbing.
Applicability
Enterprises already running BlackFog ADX, or those wanting endpoint-tier AI governance for employee-built agents and chatbots, can enable agentic prompt protection with no AI-infrastructure changes; evaluate token-visibility and dashboarding against agentic traffic growth.