CrowdStrike announced Tuesday it has launched a new feature called the "AI Kill Switch" on its Falcon Guardian platform, enabling enterprise security teams to instantly halt an internal AI agent's operations, block it from restarting, and cut off the agent's access to the corporate network when it deviates from expectations or exhibits malicious or runaway behavior. Following the announcement, CrowdStrike's shares rose more than 1% that day. The tool is built into Falcon Guardian, the platform CrowdStrike launched in September this year specifically to secure AI agents. The company also announced the completion of an integration with Anthropic's Claude Cowork. CEO and founder George Kurtz said "AI security is key to unlocking the full benefits of AI" and that "Deploying agents with a kill switch lets you stay in control." Chief business officer Daniel Bernard said the ability to maintain control and shut things down—with humans making the decisions—is critical to the next phase of AI adoption. The mechanism does not directly delete an agent's code; instead, it uses a connected AI gateway to sever the agent's access to the AI models that influence its decisions. Bernard likened the kill switch's operation to a circuit breaker, saying security personnel still judge whether behavior is benign or malicious before deciding whether to cut off access and how to subsequently restore permissions.