OpenAI launched GPT-6 Cyber, a model designed to help identify and fix software vulnerabilities, and said it had found serious flaws in widely used software.

OpenAI released GPT-6 Cyber on November 5, 2026, making it available initially to a small group of vetted Daybreak Red customers through its API and Codex. The company says researchers used it to identify 81 previously unknown Firefox vulnerabilities, all rated high severity by Mozilla, and to reproduce 1,724 Linux kernel flaws and build working exploits for 948. Of those Linux findings, 588 were previously unknown; OpenAI said it reported 304 critical issues to maintainers with proposed patches. Mozilla has begun issuing fixes. The results matter because browsers and Linux underpin widely used consumer and server systems, while the model can both locate flaws and demonstrate how they may be exploited. OpenAI rated it “Critical” for cybersecurity under its Preparedness Framework, up from “High” for GPT-5.6 Cyber. On ExploitGym, it scored 48.5%, versus 32.1% for its predecessor.