OpenAI proposes safety cases before frontier AI training runs
OpenAI’s early guidelines call for documented risk arguments, technical safeguards and senior review before frontier reinforcement-learning runs continue. The company says implementation...