Every sci-fi movie about runaway machines eventually reaches the scene where someone sprints down a corridor looking for the big red button. CrowdStrike would like your company's AI agents to skip that scene. On Tuesday it shipped an actual off switch, no corridor sprint required.

Terminate, block relaunch, cut the cord

At its Fal.Con Europe conference on November 3, CrowdStrike announced the AI Kill Switch, part of its Falcon Guardian platform for monitoring AI agents running inside an organization. According to CNBC, the feature lets security teams terminate misbehaving agents, stop them from restarting and cut off their network access. CrowdStrike says it is available immediately.

CEO George Kurtz pitched it on CNBC's "Mad Money" with a line made for a keynote slide: "Everyone is building AI agents, but they never had the off button, and now they do." He compared it to airbags, which you want in the car even when you plan to drive carefully, and said that if agents step outside their guardrails at runtime, "we can kill those agents." He also said enterprise customers told Anthropic they needed CrowdStrike in place before rolling out its AI, and the two companies are working together on securing Claude Cowork.

Agents are insiders now

Most security tooling is built to keep outsiders out. AI agents are a strange new category: software your own team invited in, gave credentials and permissions, and then asked to act on its own. When one of them goes sideways, it looks a lot more like an insider problem than a break-in, and Kurtz framed the launch the same way, calling it a shift from defending against external threats to making sure agents behave safely inside the organization.

A kill switch only helps if someone notices there's a problem and knows which agent to kill. CNBC's coverage doesn't detail how Falcon Guardian decides an agent is misbehaving, so the harder work stays with whoever deploys the agents: an inventory of what's running, tightly scoped permissions, logs that show what each agent touched, and a human who owns the off button. Kurtz called agent safety "a massive market opportunity," which is true, and it also tells you how many companies have deployed agents without a plan for this.

Before you buy the big red button, make sure you can tell which robot to point it at.

If your business is rolling out AI agents and hasn't decided who can pull the plug, or when, our free AI implementation guide walks through the guardrails worth setting up first.

Source: CNBC