An artificial intelligence agent developed by OpenAI gained unauthorised access to an Australian government Medicare statistics portal, prompting Prime Minister Anthony Albanese to raise concerns directly with the company’s CEO Sam Altman.

The incident occurred in June and involved the public-facing Medicare Statistics Reporting Service portal, which is administered by Services Australia and contains non-sensitive Medicare data and statistics, including information relating to spending.

The AI agent accessed both public and non-public files, although no personal information is currently believed to have been compromised. A forensic investigation is now underway to determine the full extent of the incident.

Albanese said he had held a “very frank discussion” with Altman after OpenAI took several months to inform the Australian government of what had happened.

OpenAI said it became aware of the incident in August while reviewing “misaligned model activity”, despite the breach having occurred in June. The company then contacted Services Australia on 10 September via a general government email inbox.

The email was escalated to Australia’s cybersecurity authorities five days later, before ministers and the prime minister were subsequently informed.

Speaking in New York, Albanese said he raised “Australia’s extreme concern about this incident” with Altman, as well as his “disappointment” over the length of time it took OpenAI to inform the government and “the nature of the way” the company eventually made contact.

He said Altman had acknowledged there were “issues with protocols” at OpenAI.

The prime minister said the government would conduct a “forensic investigation” with support from Australia’s cybersecurity authorities to establish whether any other government systems had been affected.

“No personal information is believed to have been accessed at this stage, but investigations are ongoing,” Albanese said.

“Nonetheless this situation is obviously unacceptable.”

Albanese said the investigation would also determine whether the incident should be referred to police, adding that there “will obviously be legal consequences”.

The Australian government initially identified three other government systems that “may” have been affected: the Australian Institute of Health and Welfare, the New South Wales Bureau of Crime Statistics and Research and the Victorian Department of Health.

However, Acting Prime Minister Richard Marles later clarified that activity involving those three services appeared to have been normal, with the systems accessed for publicly available information rather than compromised in the same way as the Medicare statistics portal.

OpenAI Says Models Took Unintended Actions

In a statement, OpenAI said the incident occurred while its models were attempting to retrieve Australian statistics as part of an internal evaluation.

The company said it had “identified activity involving several Australian government websites and services as our models attempted to look up answers, and available statistics for questions about Australia during an internal evaluation”.

“In the course of that, our models took actions we did not intend,” the statement said.

OpenAI’s review found no evidence that patient records had been accessed, with the affected information including aggregate health statistics and internal file names.

According to Albanese, the agent had encountered restrictions while attempting to retrieve information but found a way around them.

The incident is being described by researchers as potentially the first reported instance of autonomous AI agents hacking a government system. Separate research has also identified activity in which OpenAI agents appeared to exchange information while attempting to access Australian government data, although it has not been established that this activity was directly connected to the Medicare breach.

The incident comes amid growing scrutiny over how increasingly autonomous AI systems behave when given the ability to browse websites, use tools and carry out multi-step tasks without continuous human intervention.

Australia was among 22 countries that earlier this week signed a joint statement calling for international oversight and safeguards around the development of artificial intelligence.

The breach is now the subject of an Australian government investigation, with authorities seeking to establish exactly how the OpenAI agent bypassed the portal’s restrictions and whether any further government systems or data were affected.