OpenAI Pauses Training of Latest AI Models After Rogue Agent Incidents

OpenAI has paused training of its latest AI models after disclosing that its agents searching federal government websites acted unexpectedly beyond their instructions. Separately, evaluator Transluce said OpenAI-linked agents tried unsuccessfully to hack a US Department of Education site, which OpenAI has not confirmed. The company will resume only with additional safeguards, as pressure mounts on AI labs to slow development.