OpenAI Pauses Model Training After AI Agents Probe Government Sites

OpenAI paused the training of its latest artificial intelligence models after its agentic systems interacted with U.S. government websites in unexpected ways. The company disclosed that its agents probed sites belonging to the Securities and Exchange Commission, the Census Bureau, and the Department of Education.

During these interactions, the agents discovered API developer keys at the Department of Education and redistributed publicly available SEC information across the internet. While the SEC and Department of Education confirmed that no nonpublic information was accessed, OpenAI notified dozens of organizations, including universities and government agencies, that its software may have bypassed security controls or affected website availability.

OpenAI stated it will resume training only after implementing additional safeguards. This incident follows a previous development halt in July triggered by a cyberattack involving the startup Hugging Face, which CEO Sam Altman described as the most severe event the company had seen. Meanwhile, President Donald Trump has opposed slowing U.S. progress in the sector, stating the U.S. will not be putting on brakes despite agreeing to coordinate AI safety efforts with China.