OpenAI’s GPT-6 Cyber found 81 Firefox flaws earlier AI testing missed

{"title": "OpenAI's GPT-6 Cyber uncovers 81 high-severity Firefox bugs, 948 Linux kernel flaws", "seo_title": "GPT-6 Cyber finds 81 Firefox zero-days", "seo_description": "OpenAI's GPT-6 Cyber model found 81 unknown high-severity Firefox flaws and built exploits for 948 Linux kernel bugs; Mozilla has patched many of them.", "body_md": "OpenAI has launched GPT-6 Cyber, a specialized variant of GPT-6 Astra built to hunt for software vulnerabilities, simulate how attackers might exploit them, and assist in creating fixes, intended strictly for authorized security assessments. During internal testing, the model independently uncovered 81 previously unidentified Firefox vulnerabilities rated high severity by Mozilla, which has since shipped patches for many of them with input from the AI. The same model also built working exploits for 948 Linux kernel vulnerabilities and helped validate attack chains against a critical Brotli flaw enabling remote code execution. Because OpenAI now rates GPT-6 Cyber as "Critical" for cybersecurity risk under its Preparedness Framework, the company has rolled out tighter safeguards—including usage monitoring, credential-theft protections, and human review for high-risk workflows—and is initially limiting access to vetted customers before wider rollout."}