OpenAI has released GPT-6 Cyber, a model built for cybersecurity work, and Reuters reports it carries a "Critical" cybersecurity rating. The launch was announced on Thursday, 5 November 2026, according to Reuters.
OpenAI described the release as an effort to push forward "aligned cybersecurity agents," according to its own announcement. The company frames the model around security tasks rather than general use.
Coverage of the launch focused on the model's ability to find and exploit software weaknesses. Axios reported that the new cyber model identified ways to attack close to 1,000 Linux flaws. The Hacker News reported that an OpenAI pipeline produced 948 Linux root escalation exploits during a run lasting 30 hours.
Those figures point to how quickly the system can turn vulnerability discovery into working attack code, which is likely why the model drew a "Critical" rating. The reporting does not detail how that rating was assigned or what safeguards OpenAI has put in place.
The launch lands alongside separate security research. The Information reported exclusively that a researcher found a vulnerability in software used for data centers, a reminder that the infrastructure AI systems depend on carries its own exposure.
What to watch next: whether OpenAI publishes the safeguards and access controls around GPT-6 Cyber, and how security teams respond to a model that can generate exploits at this scale.
