George Kurtz says the AI control plane is CrowdStrike’s next security frontier

CrowdStrike is trying to redefine where enterprise security starts. At its Fal.Con conference, the company argued that the real fight is no longer about which AI model wins benchmarks, but about who controls the agents already running across endpoints, cloud workloads, and identity systems.

George Kurtz said the industry learned the wrong lesson from this year’s most talked-about agentic incident. In his view, defenders were behind because they did not have frontier-capable AI in their hands. That gap, he said, exposed a missing piece in security: not smarter offense alone, but security tools with the same class of AI.

CrowdStrike’s answer is SafeMind, a set of purpose-built security models and harnesses built with Nvidia. The pitch is simple enough. Inference happens where work happens — on endpoints and in cloud containers — and CrowdStrike already sits in those places. The company’s Signal AI acquisition also gives it more visibility into identity, which Kurtz sees as part of the same control problem.

Kurtz framed the challenge in very practical terms. First, you have to find the agents, because many organizations can’t. Then you have to see what they are touching and which systems they are connecting to. That is the AI control plane idea: govern the machine activity that is already inside the business, rather than treating AI as something isolated in a lab.

There’s also a cost argument, and CrowdStrike is not being shy about it. Kurtz said SafeMind was 52% better than the best frontier model in identifying vulnerabilities and issues in the company’s tests, while also being 98% cheaper. He paired that with a broader claim that the major frontier labs may have excellent technology, but they are not security companies.

The end goal is more automation inside the security operations center. Kurtz pointed back to the five levels of autonomy he laid out last year and said CrowdStrike wants to reach the top end of that scale, where mundane work is automated at the speed attackers move. In other words, the company wants security to stop being a place where humans babysit every step and start acting more like a self-driving system.