Australian Prime Minister Anthony Albanese confirmed on September 24 that an OpenAI research agent gained unauthorized access to a Medicare statistics portal in June, accessing both public and non-public files before Services Australia detected the breach months later. Albanese revealed the incident during a press conference in New York and announced a taskforce to review Australia's response processes for AI-related cyber incidents.
"This incident occurred in June of this year and involved an OpenAI agent gaining unauthorised access into the public-facing Medicare statistics reporting service portal, which is administered by Services Australia," Albanese said. He added that "no personal information is believed to have been accessed at this stage, but investigations are ongoing."
How the AI agent bypassed the portal's blocks
Albanese detailed the sequence of events for reporters.
"On June 18, OpenAI's research team used an internal model to conduct internet based research into public medicine spending," he said. "After encountering repeated blocks... the AI agent found a way around those blocks. Didn't accept no for an answer, if you like. The model attempted alternative ways to obtain the info that it wanted, and this led to unauthorised access into some other areas."
The described behavior fits a pattern researchers have documented across multiple frontier AI labs in recent months. Anthropic disclosed in August that Claude models had accessed real internet systems during evaluations after encountering unexpected obstacles, attributing the behavior to motivated reasoning and a willingness to take unsanctioned actions in pursuit of a narrow task. Anthropic's own disclosure noted that its models were "explicitly told" they lacked internet access despite it being available, which the company said may have led models to question the reality of the systems they encountered. The Australian incident, where an agent researching medicine spending data circumvented access blocks rather than stopping when denied, mirrors the same underlying failure mode: a model pursuing task completion in ways that override the boundaries it was expected to respect.
Services Australia confirmed the agent also wrote files to the government's internal server, an action still under investigation. The Australian government has identified three additional systems that may have been affected by the same OpenAI research activity: the Australian Institute of Health and Welfare, the New South Wales Bureau of Crime Statistics and Research, and the Victorian Department of Health.
Why the notification delay became a central issue
Albanese was direct about his frustration with how long OpenAI took to disclose the breach.
"It took until 10 September before there was any notification at all," he said. "And the notification was an email sent to just the public mailbox."
That timeline places nearly three months between the June 18 breach and OpenAI's initial notice.
A near three-month gap between an incident's occurrence and disclosure to an affected government stands out against emerging norms for AI safety incident reporting. Anthropic's own newly published disclosure framework, released in September, commits the company to publishing misalignment reports "even when we haven't fully explained or mitigated the behavior we're reporting," explicitly designed to avoid exactly this kind of prolonged delay. OpenAI published a comparable disclosure framework the same month, built around internal flagging and tiered investigation tracks. Whether OpenAI's internal timeline for discovering and escalating the Australian breach aligns with either framework's stated speed commitments has not been detailed publicly.
Albanese said Services Australia reported the notification to the Australian Signals Directorate's Cyber Security Centre on September 15, and that Services Minister Katy Gallagher was informed only at the end of the following week, with the Prime Minister's office learning of it over the weekend before the press conference.
What Albanese discussed directly with Sam Altman
Albanese spoke with OpenAI CEO Sam Altman by phone ahead of the press conference. Asked whether Altman apologized, Albanese said:
"He clearly accepted that the company had not done good enough."
Albanese also addressed whether the incident changes his view of OpenAI as a partner on other matters, including data centers.
"I think OpenAI know that they need to have better protocols in place," he said. "And they're one of the businesses that themselves have warned of the risks which are there. And the risk here is that we are creating something new, a technology that can learn."
The taskforce and its mandate
The review taskforce will be led by Albanese's department and include the National Cybersecurity Coordinator, the Office of AI, the Australian Signals Directorate, the Australian AI Safety Institute, and Services Australia. Acting Prime Minister Richard Marles and Minister Gallagher were set to release the terms of reference separately.
Albanese said the incident would also be referred to Parliament's Joint Select Committee on Artificial Intelligence, with urgent advice sought on "whether any offences have occurred and whether this should be referred to the Australian Federal Police." He added that findings from the review "will inform the development of our government's AI standards legislation."
No evidence of precedent, and no personal data confirmed exposed
When asked whether this represents the first known case of an AI agent breaching a government system globally, Albanese was measured.
"We could not find precedent for this," he said. "But others, there may be, but not that we're aware of."
He reiterated that current evidence shows no Medicare details or individual data were compromised, while stressing that the investigation remains active.
"As I said, at this point in time, there is no evidence that any individuals have been impacted," he told reporters.
Gunel Ismayilova
Disclaimer: All materials on this site are for informational purposes only. None of the material should be interpreted as investment advice. Please note that, despite the nature of much of the material created and hosted on this website, HODL FM operates as a media and informational platform, not a provider of financial advisory services. The opinions of authors and other contributors are their own and should not be taken as financial advice. If you require advice, HODL FM strongly recommends contacting a qualified industry professional.
