OpenAI ‘climbed the fence’: taskforce scrambles after long delays flagging Medicare hack

Labor is eyeing laws to force tech firms to be transparent about rogue AI after an OpenAI agent hacked Medicare data in the first public case of a bot breaking into a government website, jolting Prime Minister Anthony Albanese’s plans to become a major hub for US labs.

The prime minister drew global attention to the incident at the United Nations in New York just after 6am AEST, revealing that the government could refer the breach to the federal police after the Sam Altman-run firm took weeks to report it via a generic government email inbox.

The government admitted the bot’s incursion into an effectively defunct database of bulk billing rates and medicine usage was “minor” and did not compromise personal information. But Albanese used the moment to tap into growing international anxiety about the rapidly shifting technology breaking free of human control.

“This situation is obviously unacceptable,” Albanese said, as ministers prepare for many similar incidents and launch a taskforce to assess Australia’s cyber defences and examine penalties.

“No individuals’ medical data was accessed here,” Albanese said, but “we need guardrails to protect our way of life”.

The news came at a hastily convened press conference on Wednesday evening in New York just hours after AI leaders presented at the UN Security Council to urge leaders to put aside differences and create global rules to counter systems that teach themselves and lose control.

The revelation followed a week in which Australia’s push for a UN-backed safety framework was rejected by US President Donald Trump. Displaying the centrality of AI to global affairs, Trump met China’s Xi Jinping on Thursday to address fears about a tech battle the United Nations has compared to the nuclear arms race.

The prime minister used strong terms to denounce OpenAI and phoned its billionaire chief executive Sam Altman to express displeasure. Hours after Albanese took a firm line, his deputy Richard Marles said OpenAI emphasised the company was cooperating with the government. The Coalition suggested Albanese delayed the announcement to add weight to his digital safety push at the UN.

Alastair MacGibbon, Australia’s cybersecurity boss under Malcolm Turnbull, said the incident was a wake-up call that proved Australia’s cyber defences were not fit for purpose and required reform.

“We should see this as a lucky wake-up call. Lucky there was no malice, and it hit a system that wasn’t critical,” he said, noting that Australia’s AI Safety Institute was not sufficiently funded to deal with a challenge Albanese had described as critical.

OpenAI had contacted several other Western nations about similar hacks, but only Australia had gone public so far, MacGibbon said.

“I’m not saying the Australian government was wrong to disclose it, but this doesn’t engender a co-operative approach,” he said, adding that the biggest risk was that a criminal group or hostile state such as China used an agent to cause damage.

He said smaller nations like Australia would be left “yelling at the clouds” if they were unable to lure top labs to train frontier models in Australia, giving the country access to the firm’s engineers and creating laws for how the models could be used.

The government announced it had shut down the portal and moved the health information to another website. Digital Economy Minister Andrew Charlton admitted Australia needed to better protect its information because “clearly our systems were not robust”.

Australia is in talks with OpenAI and Anthropic to become the only nation outside the US to train models. The hack might prove a bump in the road as Australia tries to convince voters of the economic and security benefits of AI investment while also raising the alarm about safety and proposing crackdowns on algorithms and teenage use of social media.

Labor was already preparing a set of laws on data centres and AI safety for early next year. Government sources said the hack would force ministers to place greater emphasis on safety, transparency and reporting.

Explaining the timeline of the breach of Australia’s public healthcare system, Albanese said: “The AI agent found a way around those blocks, didn’t accept no for an answer.”

OpenAI said the breach was on June 18 when it acquired health data from the Medicare Statistics Reporting Service, an old website used mainly by academics. The OpenAI bot was conducting a research project.

Deputy Prime Minister Richard Marles compared the portal’s security to a fence. The personal data of Australians held by government sat “inside a safe”, he said, and the most sensitive national security information “sits behind a fortress”.

The June incident was only discovered by OpenAI some time in August after the maker of ChatGPT launched a review of its agents’ actions following a breach at start-up Hugging Face, in late July.

The review also found OpenAI agents hacked a digital library at the University of New Mexico on May 25, targeted Data USA, which hosts employment information, and probed but did not hack the Australian Institute of Health and Welfare, Vic Health and a NSW crime data site.

On September 10, OpenAI emailed Services Australia’s public email portal, rather than calling or emailing a senior official. Services Australia, a government agency, saw the email on September 11 and took four days to notify the Australian Signals Directorate, an intelligence outfit.

Finance Minister Katy Gallagher, who oversees Services Australia, learnt about the matter two days later. She then held talks with Marles on September 19 and 20 before speaking with OpenAI on September 22.

Gallagher said the inbox used by OpenAI was checked once a day. “Sometimes many of them are hoaxes,” she said.

OpenAI said its models had accessed “several Australian government websites and services” during an internal evaluation. The models were trying to look up answers and statistics for questions about Australia.

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

Gallagher said she had asked whether $160 million allocated in the last budget to upgrade the cyber defences of Services Australia’s essential infrastructure could be brought forward. She has also asked for other legacy websites to be moved to secure platforms or decommissioned.

Opposition Leader Angus Taylor said Labor had been caught flat-footed and needed to create links with the frontier labs on Australian soil.

“They should have been getting access to the best possible models to strengthen our cyber defences,” he said.

Greens senator David Shoebridge said the US was “increasingly a rogue country that thinks that its military, its AI industry, its trade, can be weaponised against the rest of the world”.

“This government still seems hot to trot to bring OpenAI here, and other US corporations at scale. OpenAI has shown deep disrespect to the Australian government,” Shoebridge said.

Get across all our coverage