🇫🇷 French, 6 November 2026, advanced threat detection
OpenAI Launches GPT-6 Cyber, Its Most Destructive AI Yet
OpenAI's GPT-6 Cyber model has discovered hundreds of security vulnerabilities in Linux and Firefox.
The model has been used to discover 81 previously unknown vulnerabilities in Firefox, all of which were rated as high-risk by Mozilla.
GPT-6 Cyber has also found 948 tactics for exploiting Linux systems, more than five times the number found by its predecessor, GPT-5.6.
The model’s capabilities have been tested on a mock industrial control system, where it discovered two previously unknown vulnerabilities without explicit programming instructions.