OpenAI releases GPT-6 Cyber to a vetted security-research tier

The model is initially available through Daybreak Red, an access tier for approved security work. OpenAI said it will expand access to all approved customers in that tier over the coming weeks.

OpenAI said its researchers used the model to identify previously unknown Firefox and Linux kernel vulnerabilities and supplied findings and proposed patches to maintainers. Those results are company-reported; Reuters said Mozilla had begun releasing fixes.