product update CAI v1.0.4: smarter routing, stronger prompt security and a more responsive CLI CAI v1.0.4 introduces the Selection Agent, stronger prompt-layer security, improved wait UX, and key CLI reliability upgrades—making day-to-day cybersecurity workflows more predictable and controllable.
Agentic AI When Documents Become Attack Vectors When data and instructions are not clearly separated, documents themselves become attack surfaces. This post analyzes indirect prompt injection as an architectural failure in agentic AI systems.
