“OpenAI has voluntarily paused some of its frontier reinforcement learning (RL) training runs to ensure its safety, security, and monitoring standards can keep pace with model capabilities.”