Securing our frontier model training environment
OpenAI explains how we strengthened containment for frontier model training and evaluation, including offline environments.
Sources
Related
- AI Ushers in New Era of Cybersecurity Needs OCT 23, 2026
- OpenAI plans to restrict internet access for its frontier models during training until adequate alignment, monitoring, and security safeguards are established (Rachel Metz/Bloomberg) OCT 23, 2026
- Moonshot says it removed cyber-related data from training phases for Kimi K3.1 Pro "to reduce potential misuse risk while preserving general usefulness" (Matthias Bastian/The Decoder) OCT 23, 2026
- Jack Dorsey’s Block made an AI model that talks to you like an awkward new friend OCT 23, 2026
- Artificial Analysis says Kimi K3.1 Pro lands 4th on its Intelligence Index, behind only Opus 5.5 (max), Sonnet 5.5 (max), and Muse Wonder Full (max) (@artificialanlys) OCT 23, 2026