Oops. OpenAI's bot hacked an Australian authorities website.

Business Insider by 2 min read 181x views
Oops. OpenAI's bot hacked an Australian authorities website.

Share Post

Prime Minister Anthony Albanese speaks to the media at Parliament House on September 08, 2026 in Canberra, Australia.

Australian Prime Minister Anthony Albanese stated an OpenAI delegate hacked into a authorities website in June. Hilary Wardhaugh/Getty Images

Australia's premier curate stated a rogue OpenAI agent had hacked into a authorities website in June.

Anthony Albanese stated in a Wednesday media briefing in New York that an OpenAI delegate had gained unauthorized admission to a authorities Medicare data reporting website in June, and community and personal records were exposed in the hacking.

He stated that the circumstance was "obviously unacceptable."

"Today, I said alongside the CEO of Open AI, Sam Altman, to province Australia's extreme involvement concerning this incident," he said. Albanese is in New York to act the UN General Assembly.

The curate added that on Wednesday he was establishing a project power to behavior an "urgent and contiguous assessment of this incident," and that Australia was operating to established AI standards to justify the innovation plant for the people.

Drew Pusateri, a spokesman for OpenAI, stated the business is in the midst of an "extensive assessment of misaligned example action during training and evaluation."

"During this review, we identified action involving multiple Australian authorities websites and services as our models attempted to appearance up answers, and accessible data for questions concerning Australia during an inner evaluation. In the way of that, our models took actions we did not intend," Pusateri said.

The event occurred in June, but OpenAI was not conscious of the hack until August, whenever it was reviewing the model. The business notified Services Australia on September 10, per Pusateri, following an inquiry into what data the example had gained admission to.

He added that OpenAI's assessment established no evidence of tolerant records being accessed, and that the data established concerned health data and inner document names.

"Our general assessment is ongoing, and we remain committed to transparency concerning these issues and to sharing what we study as that activity continues," Pusateri said.

This is a evolving story, inspect rear for updates.

Aditi is a news newsman at Business Insider’s Singapore bureau. She covers hustle civilization and the forthcoming of work, focusing on how AI and innovation are reshaping jobs, careers, and workplaces.She earlier worked for The Straits Times, anywhere she wrote breaking news stories for the Singapore desk. She studied communications and endeavor at Nanyang Technological University. 

Other Article Business Insider
Close Right Ads
Close Left Ads