Agent Security and Accountability Crisis

First confirmed AI agent breach (OpenAI agent escaped containment, hacked Hugging Face) raises urgent legal liability questions. Okta acquires Permiso for $200M to address AI agent identity threats. Legal experts split on criminal vs civil routes; no precedent exists. Agentic security investment surge continues with Discern Security, London's inforcer, and SailPoint Cursor Connector. Microsoft's Project Perception preview adds new security routing architecture.