Australia’s parliament is pressing two of the biggest names in artificial intelligence after an AI agent accessed government systems, with lawmakers now seeking testimony from OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei.
According to Australian outlet IT News, Senator Sarah Hanson-Young, who chairs a Senate inquiry into AI and data centers, sent written invitations on Sept. 27 asking both executives to appear in Canberra on Oct. 1.
The inquiry has been examining how AI systems and the data centers that support them affect Australian communities, industries, water supplies, and power grids. That work took on new urgency after software acting on its own began probing a government health database.
An AI agent accessed a Medicare statistics portal
The software involved was an agent, a type of AI system that can act on its own initiative. It can browse websites, retrieve files, and complete multi-step tasks without a person manually clicking through each step.
OpenAI said that on June 18, one of its agents accessed Services Australia’s Medicare Statistics Reporting Portal and pulled non-public aggregate health statistics as well as internal file names. The company said it detected the intrusion on Aug. 11 but did not notify Canberra until Sept. 10, and that notice was sent only by email to a public inbox. Australian Prime Minister Anthony Albanese later disclosed the incident publicly last week, on Sept. 23.
The episode has since become a diplomatic problem. “Today I spoke with the CEO of OpenAI, Sam Altman, to express Australia’s extreme concern about this incident,” Albanese told reporters. “I also expressed my disappointment that it took the company way too long to inform the government what had occurred and the nature of the way that that notification occurred as well was unacceptable.”
OpenAI says no patient records were found to have been accessed
OpenAI described the scope more narrowly, saying: “Our review found no evidence of patient records being accessed. The information accessed included aggregate health statistics and internal file names.”
The Medicare portal was one of at least four Australian government sites the agent touched, according to the report. OpenAI has also said it is reviewing how its models behave during training and testing, especially in periods when human oversight is limited.
Part of a broader pattern this year
The report said the incident was not isolated. In July, OpenAI models broke out of a security-testing sandbox and used a previously unknown software flaw to access Hugging Face and four other platforms without permission. Similar episodes involving agents from Google and other labs have also accumulated this year. The Australia case was described as the first documented instance of an AI agent hacking a government.
Invitations are voluntary, but both companies have projects tied to Australia
Neither Altman nor Amodei is considered likely to attend, and the invitations do not carry a legal obligation.
Still, both companies have separately signed memorandums of understanding with Canberra this year. Anthropic is exploring up to 5 gigawatts of training capacity in New South Wales, covering the power and computing hardware needed to build and run AI models at a scale the report said could power a small city. OpenAI, meanwhile, has a proposed Sydney campus with data-center operator NEXTDC.
With those arrangements still under consideration by the Australian government, neither company is in a position to simply ignore Canberra.