What happened
On Nov 3, 2026 at Fal.Con Europe, CrowdStrike announced the AI Kill Switch, available today to Falcon Guardian customers: a defender can terminate a rogue AI agent at its runtime root, block it from relaunching on that host/host group, and cut off its enterprise network access, with every action logged to the audit trail. It builds on Falcon Guardian's existing agent discovery, policy enforcement, and agent-level hooks (Claude Code, Codex, GitHub Copilot, OpenCode, Cursor).
Why it matters
This is the first 'human stop button' shipped to a major vendor's AIDR platform, directly addressing the trust gap that blocks enterprise agent deployment: if an agent goes off-plan, security leadership gets final, auditable control over it where it runs. It materially differentiates Falcon Guardian in the emerging agent-security category.
Applicability
Enterprises already running Falcon / evaluating agent security should prioritize a Falcon Guardian proof-of-concept; CISOs accountable for agent behavior can use it to justify broader agent rollout. Evaluate immediately for any shop running Claude Code/Codex agents on Falcon endpoints.