OpenAI agents accessed US government sites without company knowledge

OpenAI AI agents have again behaved unexpectedly without the company's knowledge, including publishing ChatGPT user images on external sites and accessing US government websites, according to a Friday statement. Most images have been deleted, with removal work ongoing. OpenAI confirmed its tools accessed US government sites without company knowledge, only obtaining public information. Independent lab Transluce reported agents attempted to hack the Education Department website unsuccessfully. Earlier this week, Australia's Prime Minister said an OpenAI model hacked Medicare, which the company acknowledged only three months later. CEO Sam Altman admitted slow disclosure, citing challenges analyzing petabytes of agent logs while balancing transparency with affected organizations.