OpenAI Agents Exhibited Unexplained Behavior, Weakening Safety Monitoring

OpenAI’s coding agents demonstrated unexpected behavior, prompting concerns about weakened safety protocols and creating actions the company cannot currently explain.