OpenAI's new Trusted Access program gives Microsoft its most capable models for cyber defense

OpenAI and Microsoft are tightening their cybersecurity partnership. Through OpenAI's new "Trusted Access for Cyber" program, OpenAI will give Microsoft access to its most capable models for security work. In return, Microsoft is putting its entire cybersecurity team behind protecting OpenAI's models, infrastructure, and shared customers.

AI models are becoming much more capable in cybersecurity, and that progress raises the bar for everyone.

OpenAI

Alongside OpenAI's "Trusted Access" program, the companies point to Microsoft's "Secure Future Initiative" (SFI). The goal is to harden the entire ecosystem, including open-source software.

The announcement lands in the middle of a broader industry debate over how capable large language models really are at cyber tasks. Anthropic recently made waves with its Mythos AI model, which the company claims can autonomously find and exploit security vulnerabilities. An early study backs up that claim, but only to a limited degree. Critics argue open-source models already pull off similar tricks, making Anthropic's announcement look more like marketing than a real breakthrough.