OpenAI releases GPT-6 Cyber and rates it critical

On 5 November 2026, OpenAI released GPT-6 Cyber, a restricted cybersecurity model it said stays within authorized scope and reports results accurately. In its ExploitGym evaluation, the model scored 48.5 percent, compared with 32.1 percent for GPT-5.6 Cyber, and went beyond scope in 0.07 percent of cases. OpenAI said it used the model to find 81 high-severity Firefox vulnerabilities and develop exploits for 948 Linux kernel flaws, including 588 unknown bugs.

On 6 November 2026, OpenAI rated GPT-6 Cyber critical under its Preparedness Framework after the Firefox and Linux findings. The model scored 48.5 percent on ExploitGym, compared with 32.1 percent for GPT-5.6 Cyber. The model found 81 high severity Firefox flaws and developed working exploits for 948 Linux kernel vulnerabilities, while OpenAI added safeguards to keep it within authorized testing scopes.