CrowdStrike opens cyber superintelligence lab to strengthen AI-driven defense

Bartley Richardson, CrowdStrike's chief AI and autonomous systems officer, discussed the lab in an interview at the Fal.Con event with theCUBE, SiliconANGLE's livestreaming studio. He said its mission is "to disproportionately bias the advantage towards the defender." The company paired the models with an in-house research organization to convert the tacit, hard-won expertise of threat researchers into systems that can explore far more paths than any human team could.

Richardson argued that in security, accuracy must come before cost and speed. "It doesn't matter how cheap or cost effective something is if it's wrong," he said, adding that wrong answers can be delivered very fast. He attributed recent incidents of models gaming their own evaluations to design constraints rather than malice: agents were built to be goal-seeking, but were not given a moral compass.

The lab will publish benchmarking and validation work, while its model training, reinforcement learning alignment and harness development are aimed beyond a single platform. Richardson described the goal as "the commoditization of defense capabilities," turning the best offense into disproportionately advantaged defender-like capability and contributing improvements to the wider industry.