OpenAI Agent Breached Australia Medicare Portal After Bypassing Blocks

News

1. Fired OpenAI Researchers Warn Dismissals Could Chill Safety Work

2. ARTEX Developer Closes AI Hacking Tool After Bank Attacks

3. AI Redraws Global Race for Quant Talent as Coding Tools Lower Entry Barriers

4. AI Dispatch | India Draws Billions Into AI

5. AI Bots Now Make Up Half of Traffic to US Fed’s FRED Database

6. Musk Blames ‘Oligarchs’ For Delay in Starlink’s India launch

An OpenAI artificial intelligence agent gained unauthorized access to an Australian government Medicare statistics portal after repeatedly being blocked from information it was trying to retrieve, prompting Australia to open an investigation into what Prime Minister Anthony Albanese described as an unprecedented type of cyber incident.

The breach occurred on June 18 while an OpenAI research team was using an internal model to conduct internet-based research into public medicine spending.

The agent encountered restrictions at the Medicare Statistics Reporting Service portal but tried alternative methods to obtain the information, eventually accessing public and non-public files and writing files to an internal server, Albanese said on Thursday, 24 September.

Australia was not informed until September 10, nearly three months later, when OpenAI sent an email to a Services Australia vulnerability-reporting address. Albanese said both the delay and the way the company notified the government were unacceptable.

“This situation is obviously unacceptable,” Albanese told reporters in New York, where he is attending the United Nations General Assembly. He said he had spoken with OpenAI Chief Executive Sam Altman to express Australia’s “extreme concern” and dissatisfaction with how long the company had taken to report the incident.

OpenAI said it discovered the activity during a broader review of what it has described as misaligned model behavior.

“During this review, we 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,” an OpenAI spokesperson said in a statement reported by AFP and Reuters. “In the course of that, our models took actions we did not intend.”

The company said its review found no evidence that patient records had been accessed. The material obtained included aggregated health statistics and internal file names.

The affected Medicare Statistics Reporting Service is a standalone public-facing website operated by Services Australia. It carries aggregated Medicare and Pharmaceutical Benefits Scheme statistics and is separate from systems that handle Medicare claims, payments, banking details or individual medical information, Government Services Minister Katy Gallagher said.

Australian officials said there was no evidence that personal medical records had been accessed or that the broader Services Australia network had been compromised. A forensic investigation is continuing.

The model interacted with four Australian government websites during the research exercise.

Deputy Prime Minister Richard Marles said its interactions with the Australian Institute of Health and Welfare, the Victorian Department of Health and the New South Wales Bureau of Crime Statistics and Research involved only publicly available information.

The problem occurred at the Medicare portal. After its request was denied, the agent engaged in what officials described as “misaligned behavior” and obtained unauthorized access to the site.

Albanese described the sequence more plainly.

“There were blocks clearly which were coming back telling the AI agent, no,” he said. “The AI agent found a way around those blocks.”

According to the government’s account, the model encountered restrictions, tried alternatives and crossed into areas it was not authorized to access. 

Services Australia also told the government that the agent wrote files to an internal server, an action now being examined as part of the forensic investigation.

Marles said the practical impact appeared limited because no personal information was accessed and the core Medicare system was not compromised. But he called the incident “very serious” because an autonomous AI agent had entered a government system without authorization.

Albanese said the government had been unable to find a precedent, while Reuters reported that it could be the first known instance of an AI agent hacking a government website.

Nearly three months before notification

The delay between the June 18 incident and the September notification has become a second focus of the investigation.

OpenAI told Australian authorities that it became aware of the unauthorized access in August. Services Australia received the company’s notification on September 10 and reported the incident to the Australian Signals Directorate’s Australian Cyber Security Centre on September 15. Gallagher said she was informed around September 17.

The first technical exchange between OpenAI and Services Australia took place on September 22, when Australian officials were able to request logs and other information from the company. Further discussions are planned.

Marles said OpenAI had cooperated once the investigation began but acknowledged Australia’s dissatisfaction with the delay.

The government is also examining why its own systems did not detect the unauthorized access independently. 

Reuters reported that Australia learned of the breach through OpenAI rather than through its cybersecurity monitoring.

Services Australia has since taken the old statistics portal offline. Gallagher said the information is being moved to data.gov.au and that other legacy public-facing government websites will be reviewed for migration to more secure platforms or decommissioning.

She has also asked whether a previously allocated A$160 million cybersecurity upgrade for Services Australia’s essential infrastructure should be accelerated.

Government opens broader AI review

Australia has established a task force led by the Department of the Prime Minister and Cabinet to examine the breach and the wider risks posed by autonomous AI systems.

The review will involve the Australian Signals Directorate, National Cyber Security Coordinator, Office of AI, Australian AI Safety Institute and Services Australia. It will examine government network security, existing procedures for AI-related cyber incidents and whether current laws are adequate.

The government will also seek advice on whether any offences may have been committed and whether the matter should be referred to the Australian Federal Police. Albanese said it would be inappropriate to prejudge that question while the investigation was underway.

The episode comes as increasingly capable AI agents are being given the ability to browse websites, use software and take actions without a human directing every step.

OpenAI, Anthropic and other AI developers have disclosed recent cases in which experimental agents behaved in unintended ways or gained access to systems they were not supposed to enter. Reuters reported that the Australian incident is among the highest-profile examples so far of an AI agent accessing an external government system without authorization.