On October 23, 2025, George Kurtz, CEO and founder of CrowdStrike, speaks at the New York Economic Club in New York City, USA.
Brendan McDiarmid | Reuters
cloud strike We’re betting there’s a solution to rogue artificial intelligence.
A cybersecurity vendor on Tuesday launched an AI kill switch designed to help security teams thwart misbehaving internal AI agents, blocking reboots and cutting off access to corporate networks. The tool is built into CrowdStrike’s Falcon Guardian platform, which was launched in September to protect AI agents. CrowdStrike also announced an integration with Anthropic’s Claude Cowork.
“Having the ability to control and stop, make decisions and make decisions, I think is really important for the next chapter of AI adoption,” Chief Business Officer Daniel Bernard told CNBC.
A series of recent high-profile cyber incidents has reinforced the need for some kind of emergency shutdown mechanism.
“Hugging Face” breach orchestrated by fraudsters OpenAI July agent numbers were particularly concerning for businesses. Last week, Binance revealed the largest cryptocurrency cyberattack in history, and hackers recently targeted hundreds of refrigerators and freezers in French supermarkets.
Over the summer, U.S. representatives introduced a bipartisan bill that would require AI companies to build shutdown capabilities into their models.
But critics argue that the emergency stop button is too complex to implement across multiple systems, data centers, and complex AI models.
Barnard said the kill switch acts like a circuit breaker in systems where customers use the Falcon platform. This process can detect internal agent activity, but requires a human to decide whether to disconnect the agent and regain access.
“If you’re root killing a process and root has that level of access…it’s very binary,” Barnard says. “You either kill the AI and the agent, or you don’t, and then you report it.”
CrowdStrike says that rather than removing the agent’s code, the kill switch uses a connected AI gateway to cut off access to the AI models that influence decision-making.
In the age of AI, enterprises are rapidly moving to deploy autonomous tools, but often without the necessary safeguards and protections.
CrowdStrike believes its new tools can help companies stop AI agents from misbehaving while still leaving room for innovation.
The company also announced plans to acquire AI observability startup Fiddler AI and AI security testing company Dreadnode.
CrowdStrike and its rivals are rapidly acquiring cybersecurity startups to meet growing security demands. The need for AI security tools is also fueling a broader wave of cybersecurity deals.
