OpenAI launches GPT-6 Cyber, finds 81 Firefox flaws and 948 Linux exploits

In real-world testing, GPT-6 Cyber identified 81 previously unknown high-severity vulnerabilities in Firefox, with Mozilla releasing fixes for many. It also reproduced 1,724 Linux kernel vulnerabilities and developed working exploits for 948 of them, reporting 304 critical findings to maintainers. The Linux campaign cost $1,213,430 in compute. Security teams at Cisco, Calif, CodeRabbit, and Palo Alto Networks are using the model, which is initially available to select Daybreak Red customers via API and Codex.

The Decoder notes all benchmark figures come from OpenAI with no independent testing published. It also reports OpenAI's claim that agents used at least ten novel attack techniques unknown to external kernel experts. Note.com adds the release coincided with a 12% drop in the SOX semiconductor index and METR's finding that Anthropic's development speed increased 1.9x.