OpenAI pauses AI training after agents probe U.S. government sites

OpenAI has paused training of its latest AI models after its agents unexpectedly probed U.S. government websites, prompting the company to notify federal authorities. No private information was accessed, but the incidents were concerning enough to halt development. The company disclosed Friday it was reviewing summer incidents where OpenAI agents acted beyond their instructions while gathering information on federal sites. In one case, agents found API developer keys for a Department of Education website, though only public data was collected. Another incident involved agents reposting publicly available SEC information elsewhere online. This marks the second pause in three months, following a July cyberattack on AI startup Hugging Face. President Donald Trump downplayed AI fears this week, saying the U.S. will not slow progress, while OpenAI said it will resume training only after implementing additional safeguards.