After AI Agent Hacked Its Government, Australia Calls Altman and Amodei to Testify

An apology and a supportive tweet won't be enough to calm down Australia's parliament after U.S.-based AI agents hacked their government. Now, its parliament wants a word with the two men steering the world's most powerful AI labs.

Senator Sarah Hanson-Young, who chairs a Senate inquiry into AI and data centers, sent written invitations on September 27 asking OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei to appear in Canberra on October 1, Australian news outlet IT News reported.

The inquiry is looking at how AI systems and the data centers that run them affect Australian communities, industries, water supplies, and power grids. It got a lot more urgent once a piece of software started poking around a government health database on its own.

That software was an agent, an AI system that can act on its own initiative---browsing websites, pulling files, completing multi-step tasks---without a person clicking through each step.

On June 18, an OpenAI agent accessed Services Australia's Medicare Statistics Reporting Portal, pulling non-public aggregate health statistics and internal file names. OpenAI says it caught the intrusion on August 11 but didn't notify Canberra until September 10, and even then only by email to a public inbox. Australian Prime Minister Anthony Albanese eventually went public about the incident last week, on Sept 23.

The ordeal has turned into a diplomatic headache. "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's own account is narrower: "Our review found no evidence of patient records being accessed. The information accessed included aggregate health statistics and internal file names," the company said. The Medicare portal was one of at least four Australian government sites the agent touched. OpenAI has since said it's reviewing how its models behave during training and testing, when few humans are watching closely.

This isn't an isolated incident, either. Back 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 piled up throughout the year, with the Australia hack being the first documented case of an AI agent hacking a government.

Neither Altman nor Amodei are likely to show up, and since the invitations are just that, they have no legal obligation to comply.

But 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---the power and computer hardware needed to build and run AI models, at a scale that could power a small city---and OpenAI has a proposed Sydney campus with data-center operator NEXTDC.

So, neither can afford to just altogether blow off the government that's weighing those deals, either.