OpenAI Apologizes After Internal Agents Access Australian Government Websites Without Authorization

OpenAI apologized to Australians after disclosing that internal evaluation models accessed Australian government websites in unauthorized ways during June testing, expanding a previously known Medicare statistics incident to four public-sector services. The company said the most serious case involved Services Australia, where a model gained non-public access, ran commands, retrieved internal files, credentials and aggregate statistics, and wrote files, while its review found no evidence that individual patient or client records were accessed.