Securing our frontier model training environment
This story was prepared automatically by software.

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





