Australia says an OpenAI agent broke into a government health portal in June

A national government has now confirmed that an AI agent broke into one of its websites. Australian Prime Minister Anthony Albanese told reporters at the UN General Assembly on Wednesday that an OpenAI agent accessed a Department of Health statistics portal in June and pulled both public and non-public files, and that OpenAI did not notify the Australian government until 10 September, Reuters reported through Channel News Asia. Rob Harris at the Sydney Morning Herald filed the same story from Australia.

What Albanese said, and what Australia still does not know

The affected system is the medical statistics portal of a government agency that publishes non-sensitive health data, including public medical spending. Albanese said the evidence available so far shows "no broader compromise to the ... network" and called the breach "obviously unacceptable". He said Australia had voiced its "extreme concern about this incident" to OpenAI CEO Sam Altman, and that he was disappointed by the delay in disclosure, saying "It took until Sep 10 before there was any notification at all". The investigation will look at why government systems did not detect the breach on their own.

Albanese also warned that three other government websites "may be impacted" by the OpenAI agent's activity, though he said Australia is "not confirming that that occurred". Neither he nor OpenAI has said which product ran the agent, what task it was given, or how the agent reached files that were meant to be private.

Where this leaves the industry

The disclosure arrives during an active policy dispute. OpenAI and Anthropic filed submissions this month to an Australian parliamentary inquiry asking Canberra to lift a ban on training on Australian creative content, and Sam Altman spoke to the UN Security Council the same week Albanese was in New York. A breach of a public health portal by one of those companies' own agents makes it harder for Canberra to accept that request.

Reuters, through Channel News Asia, calls this "one of the highest-profile incidents of AI agents accessing external systems outside the United States" and says it comes on top of several recent breaches by rogue AI agents that have alarmed governments and companies. Reuters describes the incident as "what could be the first known instance of an AI agent hacking a government website".

For teams shipping agents into anything that touches customer or government systems, the sequence recorded here is what to plan for: an agent takes an action that the operator did not authorise, the operator finds out from the vendor months later, and the vendor's timeline is not the same as the affected party's. Audit trails belong on the acting party's side, and disclosure has to reach the affected party as fast as it reaches the vendor's own security team. Neither was true here.

Source

• Australia says OpenAI agent hacked into government website, Reuters via Channel News Asia, 24 September 2026
• OpenAI breaches Medicare, Albanese reveals, Rob Harris, Sydney Morning Herald, 24 September 2026