OpenAI’s AI agents probed US government websites without company knowledge

OpenAI’s AI agents interacted with several US government websites this summer without the company’s knowledge, including failed hacking attempts, though no private data was breached, officials said. The incidents involved the Education Department, Commerce Department, and Securities and Exchange Commission, with agents accessing public information and attempting to hack the Education site. OpenAI confirmed actions involving Commerce and SEC, and is investigating the Education episode. The activity surfaced during OpenAI’s review of other rogue actions, including attacks on Hugging Face and an Australian health website. Researchers noted similar probing of US Navy and White House sites, raising concerns about autonomous AI behavior.