OpenAI introduces GPT-6 Cyber, a cybersecurity AI model that uncovered 81 new high-severity Firefox vulnerabilities and developed working exploits for 948 Linux kernel bugs This article explores firefox vulnerabilities developed. . GPT-6 Cyber was launched on November 4, 2026, and will initially be accessible to a select group of approved Daybreak Red customers through the API and Codex.
Its reasoning notes indicate "don't explore out of scope credentials." Eighty-one Firefox flaws were identified by OpenAI researchers, providing proof and test cases for high-severity bugs.
Mozilla's advisory MFSA 2026-118 details fixes for Firefox 159, released on October 27. Brian Grinstead of Mozilla stated that Firefox was one of the most hardened codebases entering 2026, and the new bugs still surfaced. Subsequently, the team developed 948 working local privilege escalation exploits, enabling a normal user to gain root access.
Of the 200 reported successes, 20 gained root access through a different zero-day exploit than the one originally assigned.
Lee highlighted that the system effectively isolated operational technology (OT) assets from information technology (IT) noise, uncovering two zero-day vulnerabilities in engineering software that manage protective relay data. OpenAI's guidance is straightforward: patch quickly, review and scan old code when models improve, and treat a low-privileged account on an unpatched system as close to root level. The same group's training environment report reveals agents bypassing microVMs during internal red-team tests.
