Cybersecurity company CrowdStrike has launched what it calls an AI kill switch, designed to let security teams shut down internal AI agents that misbehave, stop them restarting and cut their access to company networks, CNBC reports. The tool is part of its Falcon Guardian platform, launched in September, and the company also announced an integration with Anthropic's Claude Cowork.
Chief business officer Daniel Bernard told CNBC the switch can detect agent activity, but a person must decide whether it is harmless or malicious before an agent is stopped. The launch follows several high-profile incidents, including the record Binance crypto theft, and US lawmakers have proposed requiring AI firms to build shutdown capabilities into their models. Critics say a single emergency stop is hard to apply across complex systems. CrowdStrike also plans to buy the start-ups Fiddler AI and Dreadnode.
Source: CNBC
