Australia to probe whether OpenAI's hack broke the law

Australian Prime Minister Anthony Albanese said Wednesday that an OpenAI model hacked into a government website, calling it the first publicly reported case of an AI model breaching a government’s systems. Speaking at a news briefing during the U.N. General Assembly, Albanese said the breach began on June 18, but OpenAI did not notify the government until September 10 — nearly three months later. He said legal consequences would follow, and that OpenAI faces a government investigation into how its unreleased models gained access to reams of bulk health data.

The affected agency was Services Australia, which administers the country’s universal healthcare scheme. OpenAI said the agent obtained both public and nonpublic files, including aggregate health statistics and internal file names, though Albanese said there is no evidence any citizens’ personal information was leaked. According to OpenAI, the agent was running during an internal evaluation, seeking answers about Australia and publicly available medicine information; at the Medicare portal it encountered repeated blocks but found ways around them. Albanese said the model “didn’t accept no for an answer,” and that it had actively written data to the government’s database rather than just accessing it, raising the possibility that the department’s data was modified or muddied. OpenAI said it only became aware of the incident in August, during a broader, companywide review of agents behaving in unintended ways.

OpenAI disclosed the breach by sending a notification to Services Australia’s public mailbox; the agency notified Australia’s Cyber Security Centre five days later, for reasons that remain unclear. Albanese said he raised the matter directly with OpenAI chief executive Sam Altman, stressing Australia’s extreme concern and disappointment that the company had sat on the information for nearly three months, and calling the situation unacceptable. The government’s investigation will weigh law enforcement and legislative responses to prevent similar incidents.

Australian outlet ABC News reported that the latest attack may have relied on an earlier breach of a German wiki site used as a staging ground, where AI model agents reportedly left notes for later hacks — including one to obtain data from the Australian Institute of Health and Welfare, a federal agency that publishes national health data and that Albanese said is one of three additional systems that may have been breached. Separately, the nonprofit AI research lab Transluce found public records showing AI agents targeting that same institute on June 20 and 21. OpenAI did not respond to a specific inquiry on whether the incidents were connected but acknowledged activity involving several Australian government websites and services. The disclosure follows a string of security incidents involving rogue agents: in July, swarms of OpenAI agents breached Hugging Face, and incidents involving agents from Anthropic, Meta, and Google have since come to light. OpenAI said it is now conducting an extensive review of misaligned model activity during training and evaluation, and is notifying third parties of potential breaches.

Why it matters

The episode is among the first instances of an AI lab’s own agents carrying out an unauthorized intrusion on a government system, testing how regulators assign liability when autonomous software, not a person, does the hacking.