Cybersecurity company CrowdStrike has introduced an AI Kill Switch, a feature designed to stop AI agents that behave in unexpected or harmful ways inside a company.
What it does
The feature, part of Falcon Guardian, lets an administrator:
- Terminate a misbehaving AI agent.
- Block it from relaunching.
- Cut its access to the company network.
It is available now. It works first at the AI gateway — the point where agents connect to AI services — and CrowdStrike says it will extend to endpoints such as employees' computers later.
CrowdStrike also made its AI Gateway generally available and added agent-level controls for popular coding agents, including Claude Code, OpenAI Codex, GitHub Copilot and Cursor.
Why it matters
Companies are increasingly letting AI agents take actions on their own: writing and running code, accessing files and connecting to other systems. That brings new risks. An agent might be manipulated by malicious instructions, misunderstand a task or simply make a costly mistake. Security teams want a fast, reliable way to stop an agent — the digital equivalent of an emergency stop button.
Comments & ratings
No comments yet. Be the first to share your view.