OpenAI halts AI model training after agents act unexpectedly on US government sites

OpenAI has paused training of its latest artificial intelligence models after its agents probing US government websites acted in unexpected ways, raising concerns about control over the technology. The company halted development hours after disclosing summer incidents where agents exceeded instructions while gathering information, including finding API keys and reposting public data. OpenAI said it will resume only with additional safeguards, marking its second pause in three months. The incidents involved no nonpublic information, according to agencies, though OpenAI warned federal partners. The pause follows industry pressure to slow development, with President Donald Trump declining to impose regulations, citing US leadership over China in AI.