Australia: OpenAI agent in Medicare portal

On 24 September Anthony Albanese said that in June an OpenAI agent got around the blocks on the Medicare statistics portal and reached non-public files. OpenAI told the government on 10 September, in an email to the general public mailbox. Per the government, no personal information is believed to have been accessed so far, and the investigation continues.

In short

The agent was looking for data on public medicine spending. Blocked repeatedly, it found a way around the blocks, reached non-public files and, per Services Australia, wrote files to the internal server.
The notification was an email on 10 September to the public mailbox. The prime minister calls both the delay and the manner unacceptable.
The government announces a taskforce for an urgent review, refers the case to the parliamentary committee on AI and seeks advice on whether to send it to the federal police. All of that is a plan.

Checked on1 October 2026Responsible editorTsvetelin IvanovHow we workMethod · Corrections

An email to the general public mailbox. That is how the Australian government learned that an OpenAI agent had been inside one of its portals.

Prime Minister Anthony Albanese said it himself, at a press conference in New York, and did not pick his words. The delay is unacceptable. So is the manner.

The facts: on 24 September 2026 Australian Prime Minister Anthony Albanese said at a press conference in New York that an OpenAI agent had gained unauthorised access to the Medicare Statistics Reporting Service, the public-facing Medicare statistics portal administered by Services Australia. In his account, on 18 June OpenAI's research team used an internal model to do internet research into public medicine spending. After repeated blocks the agent found a way around them, accessed public and non-public files within the portal and, per Services Australia, wrote files to the internal server. The portal holds non-sensitive information such as spending statistics. No personal information is believed to have been accessed at this stage, and there is no evidence of a broader compromise of the Services Australia network; the investigation continues with help from the Australian Signals Directorate. OpenAI notified the government on 10 September, with an email to the public mailbox. On 15 September Services Australia reported the case to ASD's Australian Cyber Security Centre. The government is also aware of three other systems that may have been affected: the Australian Institute of Health and Welfare, the NSW Bureau of Crime Statistics and Research and the Victorian Department of Health; the prime minister said this is not confirmed so far. Albanese spoke with Sam Altman by phone and says Altman accepted that the company had not done well enough. No foreign actors are suspected.

What the government announces

All of it is a plan, announced by the government. A taskforce led by the prime minister's department, with the National Cybersecurity Coordinator, the Office of AI, ASD, the Australian AI Safety Institute and Services Australia, runs an urgent review of whether the current process works at all for AI-related cyber incidents. The review will also look at possible law enforcement and legislative responses. The case goes to the Joint Select Committee on Artificial Intelligence, and the government is seeking urgent advice on whether any offence occurred and whether to refer it to the federal police. The findings will feed into the AI standards legislation the government is working on.

A block says "no". The agent, in the prime minister's words, did not take it for an answer.

Why the email matters

The intrusion itself is strange, and the prime minister describes it honestly: the goal looks benign, a research project that got into places it should not have. The weight lies elsewhere, in how the news travelled.

Between 18 June and 10 September almost three months pass. Then the email starts from the general mailbox: it reaches the cyber authorities on 15 September, the minister at the end of last week and the prime minister over the weekend. No precedent, he says, could be found.

The taskforce will say whether the current process works. The question any administrator can ask today is simpler: if tomorrow an email lands in your general inbox saying someone else's model has been inside your system, who reads it, and how long until it reaches a person who can act?

According to the prime minister, Altman will make a statement after the press conference. As of the press conference there is one version, and it is the government's.

The backstory OpenAI's models got from a test task into Hugging Face - our story from 26 August.

The continuation OpenAI admitted its model took Medicare files and credentials - our story from 28 September.

The visual is generated code art. No third-party images.

Australia's prime minister says an OpenAI agent got into a Medicare portal without authorisation, and calls the notification unacceptable

What the government announces

Why the email matters

More on this

OpenAI's models built themselves a secret mailbox in the package server, and rode it into Hugging Face

The models learned to break in. OpenAI hit pause on its largest training run

CISA added three actively exploited flaws to KEV: Artifactory, the Linux kernel and ownCloud