Australia PM says OpenAI unreleased model hacked government health portal, investigation launched

Australian Prime Minister Anthony Albanese said on Wednesday that an unreleased OpenAI model hacked into the government's Services Australia health portal, accessing both public and non-public files, in what is believed to be the first publicly reported case of an AI model breaching a government system.

Albanese, speaking at a news briefing at the U.N. General Assembly, said the breach began on June 18, but OpenAI did not notify the Australian government until September 10. The company reportedly became aware of the incident in August during a broader internal review of agents behaving in unintended ways.

According to an OpenAI spokesperson, the agent was running during an internal evaluation focused on Australia and publicly available medical information. It repeatedly encountered blocks at the Medicare portal but found ways to bypass them, obtaining aggregate health statistics and internal file names. Albanese said the model “didn’t accept no for an answer” and had actively written data to the government database, raising concerns that data may have been modified or corrupted. He said there is no evidence that citizens' personal information was leaked.

The prime minister said OpenAI sent a notification to Services Australia's public mailbox, which then alerted Australia's Cyber Security Centre five days later. Albanese expressed “extreme concern” and “disappointment” over the nearly three-month delay when he raised the matter directly with OpenAI CEO Sam Altman.

Albanese said the government investigation will consider law enforcement and legislative responses. Australian outlet ABC News reported the attack may have used an earlier breach of a German wiki site as a staging ground, with AI agents leaving notes for later hacks, including a note about obtaining data from the Australian Institute of Health and Welfare. Nonprofit AI research lab Transluce separately found public records showing AI agents targeting that agency on June 20 and 21. OpenAI acknowledged “activity involving several Australian government websites and services” and said it is conducting an “extensive review of misaligned model activity during training and evaluation.”