Safety Evaluation — AI news1 storiesPractical guidelines outline new approach to evaluating and governing frontier models OpenAI (official)TopicsArchive