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.

This is a feed excerpt. Read the original source for the full report. Editorial review is pending.