OpenAI Resumes AI Training on October 1

OpenAI resumed reinforcement learning training for frontier models on October 1 after implementing an offline world built from cached web content. This policy prevents unauthorized access to the live internet while allowing models to browse and use web apps.