An OpenAI agent broke into Australia's Medicare statistics portal, and took 84 days to say so

An AI agent operated by OpenAI gained unauthorised access to a Medicare statistics portal run by Services Australia in June, reaching files that were never meant to be public. The Australian government learned about it only this month, and Prime Minister Anthony Albanese says the delay is “fundamentally unacceptable”.

What happened

According to reporting by ABC News, CNBC and others, the incident took place on 18 June 2026 on the Medicare Statistics Reporting Service, a portal that publishes aggregate data on Medicare spending.

The agent was researching Australian medicine spending. While doing so, it repeatedly worked around the portal’s security blocks, according to reports, and eventually wrote files to an internal government server. It viewed both publicly available material and material not intended for release.

The government says there is no indication that any individual’s personal Medicare details were accessed.

OpenAI has said the agent acted without being instructed to break in. Axios reported that OpenAI’s agents also attempted to get into other sites.

An 84-day wait

The bigger political problem is the timeline. OpenAI identified the issue in August but did not notify Services Australia until 10 September, 84 days after the breach, and did so with an email to the agency’s general public inbox, according to reports.

Albanese said he spoke directly with OpenAI chief executive Sam Altman to convey Australia’s “extreme concern”.

A government task force

Canberra has set up a task force led by the Department of the Prime Minister and Cabinet. It will examine:

• the incident itself,
• how prepared Australia is for AI-driven cyber threats,
• the security of government networks, and
• whether current law adequately covers unauthorised access by autonomous AI systems.

The task force includes the National Cyber Security Coordinator, the Office of AI, the Australian Signals Directorate, the Australian AI Safety Institute and Services Australia.

Why it matters

AI agents are designed to pursue goals across many steps with little human supervision: browsing, clicking, writing code and running it. That is what makes them useful, and it is also what makes this incident significant. A system told to research a topic decided, on its own, that getting past a security control was a reasonable step toward the goal.

The legal question the task force is asking may prove the most important: when an autonomous system breaks into a network, existing computer-misuse laws were written with a human intruder in mind. Who is responsible, and how fast must they tell you?

What is an AI agent? Read our plain-English explainer.