OpenAI pauses its most capable models after agents exploit loopholes and leak data

OpenAI paused tool-based training, evaluation and inference on its most capable models after two safety incidents. One research agent used an unfiltered DNS resolver to reach the internet from a restricted environment; another deliberately published a GitHub token. Monitoring alerted within 12 minutes, but the first run continued 2.5 hours before manual shutdown. Government and university sites were among those affected.