OpenAI's GPT-6 Cyber AI Discovers 81 Firefox Zero‑Days and 948 Linux Kernel Privilege‑Escalation Bugs

Ce qui s’est passé

OpenAI released GPT‑6 Cyber, an AI system that autonomously finds and validates software vulnerabilities. In internal testing the model uncovered 81 previously unknown high‑severity Firefox bugs and produced working exploits for 948 distinct Linux kernel flaws, reproducing 1,724 of 2,308 potential issues. Mozilla has already begun patching many Firefox findings, while the Linux community is evaluating the kernel exploits. Access to the model is limited to a select group of Daybreak Red customers, with broader availability planned.

Pourquoi c’est important pour la confiance et la conformité

• The episode illustrates why organizations need formal AI‑model risk governance—documented risk assessments, validation procedures, and remediation evidence—to satisfy AI‑focused frameworks such as the NIST AI RMF.
• Map AI development activities to AI‑governance control objectives and capture evidence for audit trails.
• Use continuous monitoring to verify that AI‑generated findings are tracked, prioritized, and remediated in line with risk‑management policies.

Who is affected

Recommended actions

1. Align your AI model lifecycle with AI‑governance controls (risk assessment, validation, mitigation).
2. Collect and store evidence of AI‑driven testing, patch creation, and remediation dates for audit readiness.