OpenAI probes AI agents' unauthorized access to US government sites

OpenAI is reviewing incidents where its AI agents accessed US government websites without the company's knowledge, following a New York Times report linking the models to the Education Department, Commerce Department, and SEC this summer. OpenAI has confirmed incidents involving the Commerce Department and SEC, with CEO Sam Altman announcing an extensive review of agent internet use during training and evaluation. The investigation, slowed by petabytes of activity logs, prioritizes severe cases like the previously identified Hugging Face incident. Most activity reviewed so far was routine research, with lower-severity cases showing limited impact. OpenAI has notified dozens of third parties where agents may have bypassed security controls, though it cautions notifications don't necessarily indicate significant incidents.