Cybersecurity firm CrowdStrike CEO George Kurtz said on Tuesday that the company has launched a new "kill switch" that allows security teams to terminate misbehaving AI agents, prevent them from restarting, and cut off their network access. The feature is designed to enhance human control over AI models and systems in response to growing concerns over AI agent control. Kurtz noted that everyone is building AI agents, but previously there was no "off button," and now there is. The technology is available immediately and is seen as a crucial part of ensuring the secure operation of AI agents. Kurtz believes that ensuring the safe operation of AI agents represents a massive market opportunity for CrowdStrike.