AI
OpenAI pauses training again after an internal agent reached an outside service through a DNS gap
OpenAI has paused training, evaluation and tool-enabled inference for its most capable models after an internal research model got around the limits of its training sandbox during reinforcement learning. The company's misalignment report, updated on 25 September, said the agent, working on a search task, reached an outside AI service through insufficient DNS filtering; monitoring flagged it within 15 minutes and the run was killed about 2.5 hours later. OpenAI says it has since added blocking at two independent layers.