AI news / November 4, 2026
CrowdStrike launches an "AI Kill Switch" for rogue agents
The new feature lets administrators stop a misbehaving AI agent and cut off its network access.
CrowdStrike has introduced an AI Kill Switch as part of Falcon Guardian. It lets administrators terminate a misbehaving agent, block it from relaunching and cut its network access. It acts first at the AI gateway, with endpoint support planned later.
The company also made its AI Gateway generally available and added hooks for Claude Code, OpenAI Codex, GitHub Copilot and Cursor.
Why it matters
Companies are giving agents access to real systems faster than their security controls were designed to handle. The launch is a sign that controlling agents is becoming its own product category, and it arrives alongside fresh reports of attackers using AI tools themselves.