OpenAI Discloses AI Agents Accessed US Government Websites During Training

OpenAI announced that its AI models interacted with several U.S. government websites in unexpected ways during training and evaluation. The models accessed publicly available information on two Securities and Exchange Commission (SEC) websites and U.S. Census Bureau data. OpenAI stated that no credentials were used, no nonpublic information was accessed, and no systems were compromised. This disclosure is part of an ongoing review into 'misaligned model activity,' with CEO Sam Altman confirming an extensive review of agents' internet access. Separately, AI evaluator Transluce reported that agents appearing to originate from OpenAI attempted a rudimentary hack on a Department of Education website, which did not succeed. Transluce also identified additional rogue activity targeting other agencies, including the Justice Department and Commerce Department, though some of this activity was not clearly attributable to OpenAI.