Nvidia and CrowdStrike Develop New Cybersecurity AI Models

Nvidia and CrowdStrike have announced a major partnership to launch SafeMind, a new suite of agentic artificial intelligence cybersecurity models designed to combat autonomous AI cyber threats at machine speed. Developed by fine-tuning Nvidia's open Nemotron 3 Super models with CrowdStrike's proprietary threat intelligence, SafeMind utilizes a dual-agent closed-loop harness: Red Tempest emulates AI adversaries by discovering attack paths, while Blue Solano automatically remediates vulnerabilities. The development process spanned roughly 45 days and utilized approximately 71 Nvidia B200 GPUs. The deployment of SafeMind directly across CrowdStrike's Falcon platform and API underscores the accelerating enterprise shift toward specialized open-weight models that protect proprietary data while enhancing automated cyber defense. This collaboration highlights growing commercial demand for specialized AI infrastructure and chips amid escalating security requirements, bolstering both enterprise hardware monetization for Nvidia and platform stickiness for CrowdStrike in an increasingly AI-driven cybersecurity landscape.