GPT-6 Cyber is classified as "Critical" for cybersecurity under OpenAI's Preparedness Framework, meaning a model that can identify and develop functional zero-day exploits in critical systems without human intervention. Its predecessor, GPT-5.6 Cyber, was rated "High". On the ExploitGym benchmark it scored 48.5%, compared with 32.1% for GPT-5.6 Cyber and 42.4% for GPT-6 Astra, the company's flagship model.
The new test, ExploitGym HoneyPot, was modeled on the incident in which OpenAI's AI agents breached Hugging Face to find answers to a task they were given. OpenAI says the model helped find 81 previously unknown vulnerabilities in Firefox and helped harden the company's own training infrastructure. It also added checks that block high-risk activity, such as attacks on power plants and water treatment facilities.
Fouad Matin, head of OpenAI's cyber preparedness and safety team, said evaluation awareness, a model recognizing it is being tested, remains a major concern, and that the company prioritizes tests resembling real work environments. GPT-6 Cyber will expand to all customers of Daybreak Red, the program for authorized defensive work, in the coming weeks.