At its Fal.Con Europe 2026 conference, CrowdStrike announced an AI Kill Switch for its Falcon Guardian platform. The feature lets security teams terminate unauthorized or misbehaving AI agents, prevent them from relaunching and cut off their network access. It is designed to give people oversight and control as enterprises increasingly adopt autonomous agents with system-level access that could take unauthorized actions.
CrowdStrike also announced a partnership with Anthropic to integrate Falcon Guardian natively into Anthropic’s Claude Cowork platform. The integration will extend Falcon’s security protection to AI agents throughout their lifecycle in the Claude Cowork environment.
The announcements came alongside CrowdStrike’s planned acquisitions of Fiddler AI, which provides agent observability, and Dreadnode, which provides AI security testing.