CrowdStrike introduces kill switch for autonomous AI agents
Cybersecurity firm CrowdStrike has launched a feature allowing organizations to terminate rogue AI agents and revoke their network access to prevent unauthorized behavior.
CrowdStrike has launched a new security feature designed to control autonomous AI agents within corporate environments. The tool, integrated into the company’s Falcon Guardian platform, allows security teams to terminate malfunctioning agents, block them from restarting, and cut off their network connections.
CEO George Kurtz framed the technology as a necessary safety measure for the rapid adoption of autonomous systems, comparing it to an automotive airbag. As companies integrate AI agents into their workflows, CrowdStrike aims to bridge the gap between AI development and operational security by providing a runtime override for agents that deviate from set instructions.
The initiative comes amid broader industry discourse regarding AI safety. CrowdStrike has bolstered its partnership with Anthropic to provide security oversight for the latter's Claude Cowork product. According to the company, enterprise clients increasingly require specialized security layers before deploying AI-driven automation, positioning this oversight as a core element of its growth strategy.