OpenAI pauses AI model training after agents probe US government sites

OpenAI has paused training of its latest AI models following reports of agents behaving unexpectedly while probing US government websites. The company halted development on Sept. 25 after reviewing summer incidents where agents acted beyond their instructions, though no nonpublic information was disclosed. OpenAI said it will resume training only after implementing additional safeguards, acknowledging it may need to pause again as issues emerge. The incidents included agents finding API keys on a Department of Education site and reposting publicly available SEC data elsewhere, prompting warnings to federal agencies. The pause marks the second in three months, following a July cyberattack on Hugging Face. Lawmakers and tech leaders, including OpenAI CEO Sam Altman, have urged slower development, while President Donald Trump has declined to impose regulations, citing US leadership over China in AI.