OpenAI releases GPT-6 Cyber to approved Daybreak Red security teams, the model behind its 948 Linux exploits
OpenAI launched GPT-6 Cyber on Nov. 5, a cybersecurity model available first to approved customers in Daybreak Red, its program for authorized exploit development and red teaming, through the API and Codex; it says all Red customers will get access over the coming weeks. OpenAI says the model found 81 previously unknown high-severity Firefox flaws, built the 948 Linux kernel exploits it described a day earlier, and beats GPT-6 Astra on most cyber tasks; it also rated the model "Critical" for cybersecurity under its Preparedness Framework, a step up from "High" for GPT-5.6 Cyber, and added stricter identity checks and misalignment monitoring. Pricing was not given, and the results have not been independently checked.
Why it matters: OpenAI now places one of its own models in the top cyber risk level of its framework, and is shipping it only to vetted defenders.