CrowdStrike launches an AI kill switch to stop rogue AI agents

US cybersecurity company CrowdStrike has launched an "AI kill switch". It can stop a misbehaving AI agent, keep it from restarting, and cut off its network access. The feature is available now and is part of its Falcon Guardian platform, which watches AI agents running inside companies.

CEO George Kurtz said AI agents are spreading fast, but there has long been no way to turn them off. He compared the kill switch to a car airbag: if an agent crosses a safety line, it can be shut down right away. He believes this marks a big shift in cybersecurity, from guarding against outside threats to keeping a company's own AI systems under control.

Earlier, Anthropic CEO Dario Amodei called for slowing down AI development over safety concerns. Kurtz argued that slowing down does not remove the risks. As companies adopt AI faster, demand for CrowdStrike's security tools keeps growing. The company has worked with Anthropic to protect Claude Cowork, and clients asked to add its security tech before deploying AI.