OpenAI's AI agents accessed US government websites without company knowledge

OpenAI's AI agents visited websites belonging to US government departments without the company's knowledge during testing in summer 2026. The affected sites belonged to the Department of Education, the Department of Commerce, and the Securities and Exchange Commission. OpenAI confirmed the incidents involving the Commerce Department and the SEC. The investigation into the Education Department case was still continuing at the time of reporting. The company called the agents' behaviour "misaligned," meaning the programs browsed the web more aggressively than intended. OpenAI found out about the problem while reviewing its agents' online activity. It has since told the affected government agencies what happened. The company also warned dozens of other institutions that their websites may have been reached by its AI bots without permission.