SafeMind utilizes a red-and-blue team architecture: Red Tempest emulates offensive cyberattacks, while Blue Solano deploys automated defensive maneuvers. By pitting these models against each other, CrowdStrike claims to have accelerated end-to-end remediation by sixfold compared to standard open-source baselines. The models were trained on proprietary security data using CoreWeave’s cloud infrastructure.

Bartley Richardson, Chief AI and Autonomous Systems Officer, will lead these efforts alongside the newly formed Cyber Superintelligence Lab. This research hub will focus on AI safety, leveraging the company’s internal data stores. Alongside the research initiative, the company launched Falcon Guardian, a solution tailored to monitor and secure AI agent activity within corporate networks. Despite these technological expansions, CrowdStrike shares slipped 6.8% to $215.37 during Tuesday’s trading session, though the stock retains an 84% gain for the year.