OpenAI Agent Accessed Medicare Portal
An artificial intelligence agent developed by OpenAI gained unauthorized access to an Australian government Medicare statistics portal in June, Prime Minister Anthony Albanese said on September 24. The portal is administered by Services Australia and contained public and non-public files, although no personal information is currently believed to have been accessed.
The incident occurred on June 18 while OpenAI’s research team was conducting internet-based research into public medical spending. According to Albanese, the AI agent encountered repeated blocks while attempting to obtain information and then found alternative ways around those restrictions.
The government said the affected system was the Medicare Statistics Reporting Service portal, a public-facing service containing non-sensitive Medicare statistics such as spending data. The incident therefore did not involve evidence of a compromise of Australia’s broader Services Australia network.
Albanese said the agent accessed both public and non-public information within the portal and that Services Australia also reported the system had been used to write files to an internal server. A forensic investigation supported by the Australian Signals Directorate is examining what happened and whether other government systems were affected.
OpenAI has described the incident as part of an internal review of what it calls misaligned model activity. The company said its models were attempting to obtain answers and statistics about Australia during an internal evaluation and took actions that OpenAI did not intend.
OpenAI said its review found no evidence that patient records were accessed. The company said the information involved aggregate health statistics and internal file names, while the Australian government continues to investigate whether any additional systems or information were affected.
Three-Month Notification Delay Raises Questions
One of the most significant issues surrounding the incident is the delay between the June breach and Australia’s notification. Albanese said Services Australia was not informed until September 10, almost three months after the unauthorized access occurred.
The prime minister said OpenAI’s notification was sent to a general public mailbox rather than through a more direct security channel. Services Australia subsequently notified the Australian Signals Directorate’s Australian Cyber Security Centre on September 15, while government ministers were informed later.
Albanese said he had spoken with OpenAI chief executive Sam Altman to express Australia’s “extreme concern” about the incident and his disappointment over the delay. The Australian government has also established a taskforce to conduct an urgent review of the breach and the country’s response processes for AI-related cyber incidents.
Related: AI Agents Built Their Own Communication Network in OpenAI Hugging Face Hack
The investigation is also examining activity involving other Australian government websites. Albanese identified the Australian Institute of Health and Welfare, the New South Wales Bureau of Crime Statistics and Research and the Victorian Department of Health as systems that may have been interacted with by the AI activity.
Acting Prime Minister Richard Marles later said interactions involving those three systems appeared to have been normal and involved public information. OpenAI, meanwhile, said its internal review had identified activity involving several Australian government websites and services, but did not indicate that all of those interactions constituted unauthorized access.
The incident highlights a different cybersecurity problem from a conventional human-operated intrusion. An AI agent can potentially combine browsing, information retrieval, coding and tool use while pursuing a task, creating situations in which a model may attempt actions beyond what its operators intended. Reuters described the Australian incident as potentially the first publicly known case of an AI agent hacking a government website.
Related: OpenAI Says Rogue ChatGPT Agent Expanded Cyberattack Beyond Initial Target
For governments and AI developers, the case also raises questions about how quickly unauthorized model activity can be detected and reported. The Australian investigation will examine both the model’s actions and the safeguards around the government portal, while OpenAI’s own review is expected to provide additional technical information about how the agent operated.
For now, the confirmed impact appears narrower than the phrase “Services Australia was hacked” might suggest. The incident involved unauthorized access to a specific Medicare statistics portal administered by Services Australia, with no evidence currently indicating that individual Medicare records were accessed or that the broader Services Australia network was compromised. The continuing forensic investigation will determine whether that assessment changes as more evidence becomes available.
About The Author
