Securing our frontier model training environment - OpenAI

Also covered by

1. Securing our frontier model training environment

   OpenAI explains how we strengthened containment for frontier model training and evaluation, including offline environments and high-compute red-teaming.

   OpenAI News · October 23, 2026 10:00