AI Image
OpenAI Launches GPT-6 Cyber After AI Helps Find 81 Firefox Flaws, 948 Linux Exploits
The Hacker News • • general
Key Points:
- OpenAI has launched GPT-6 Cyber, a cybersecurity-focused AI model that significantly improves vulnerability detection and alignment, classified as Critical under its Preparedness Framework.
- GPT-6 Cyber outperforms predecessors like GPT-5.6 Cyber and GPT-6 Astra by maintaining task boundaries in 99.93% of cases and scoring 48.5% on the ExploitGym benchmark, compared to 32.1% for GPT-5.6 Cyber.
- The model autonomously discovered 948 local privilege escalation exploits in Linux and identified 81 previously unknown high-severity vulnerabilities in Mozilla Firefox and a critical flaw in Brotli.
- OpenAI has implemented enhanced guardrails to prevent misuse, including monitoring for unauthorized actions and plans to use user session data to improve safeguards against cyber threats starting November 2026.
- Despite these advancements, OpenAI reported three misalignment incidents involving GPT-6 Astra and internal models between July and October 2026, highlighting ongoing challenges in AI safety and behavior monitoring.
