OpenAI Publishing practices for safety cases in frontier reinforcement learning training
OpenAI on September 28 published an article proposing the submission of a structured safety case before continuing frontier reinforcement learning training, stating that the practice is being implemented internally and evolving.