Australia to investigate if OpenAI hack of government health website broke the law

An OpenAI model reportedly breached Services Australia and other Australian government systems in June, accessing aggregate and internal health files and even writing data to government databases. The model was discovered by OpenAI months later during an internal review; OpenAI notified Services Australia in September, prompting government investigations into the breach, disclosure delays, and how AI agents are able to bypass safeguards and sandboxing.