OpenAI tightens safety controls after Hugging Face breach

OpenAI paused frontier RL training for two weeks and added monitoring, alignment, and security safeguards after the July 21 Hugging Face incident. Its largest planned frontier RL run remains on hold pending smaller-scale evaluations.