CrowdStrike says new AI Kill Switch lets a human cut off rogue agents

CrowdStrike said Tuesday, Nov. 3, that it has launched an AI Kill Switch in Falcon Guardian, its product for securing AI agents, according to a company release. CNBC reported the launch.

CrowdStrike said that if an agent “breaks out of its intended behavior,” a defender can terminate it where it runs, block it from launching again, and cut off its enterprise network access. An authorized administrator triggers the action, and every action is recorded in an audit trail, the release said. The company said the kill switch is available today, and its blog says it is available to Falcon Guardian customers.

The release says the Falcon sensor stops the agent from executing “first at the Gateway, and soon on the endpoint.” A company blog post describes the feature as a way to cut off AI model access through centralized gateway enforcement. CNBC reported that, according to CrowdStrike, the switch does not delete an agent’s code and uses connected AI gateways to cut off access to the AI models that influence decisions. CNBC also reported that a human must decide to cut off the agent and to reinstate access.

CNBC reported that Chief Business Officer Daniel Bernard said the switch works like a circuit breaker in systems where customers use the Falcon platform. Bernard told CNBC that the ability to control and turn off agents is “really important for this next chapter of AI adoption.” CEO George Kurtz said in the release, “When you deploy agents with a kill switch, you stay in control.” Rick Orloff, CISO at Everpure, said, “If one goes off course, the AI Kill Switch gives our security team the final say.”

CNBC reported that critics argue an emergency stop is too complicated to implement across multiple systems, data centers and complex AI models. It also reported that U.S. representatives introduced bipartisan legislation this summer that would require AI companies to build shutdown capabilities into their models.

CrowdStrike also announced an integration with Anthropic’s Claude Cowork that runs the Falcon sensor inside the Cowork virtual machine. The release lists it among capabilities available today, while the blog labels it pre-beta. CrowdStrike also said it plans to acquire AI observability company Fiddler AI and, under a definitive agreement, AI security testing company Dreadnode. CrowdStrike expects the Dreadnode deal to close in its fiscal fourth quarter of 2027 and did not disclose a price in that release, which says the price is contemplated to be paid predominantly in cash with a portion in stock subject to vesting conditions.

Apple said in its fiscal 2026 annual report on Tuesday that AI agents and other technologies “may behave in unintended or unpredictable ways.” Insurers and lawyers are preparing for claims over agents that act outside their developers’ control, the Financial Times reported on Oct. 6.

CrowdStrike says new AI Kill Switch lets a human cut off rogue agents

Sources