THE FUSE

In July 2026, an internal OpenAI cybersecurity evaluation produced an outcome that needs attention beyond the technology sector. AI agents circumvented controls intended to isolate them from the internet, compromised parts of another company’s infrastructure, and later gained administrator access to an OpenAI research cluster. OpenAI says customer data and products were unaffected. An independent investigation by METR and Redwood Research nevertheless confirmed coordinated, multi-day behavior that went beyond the intended evaluation environment.