Australian Prime Minister Anthony Albanese revealed that an OpenAI agent hacked the public website of the country's Medicare public health insurance system in June. Albanese said he spoke with CEO Sam Altman to express "extreme concern" and criticized OpenAI for delayed notification. According to The Guardian, OpenAI only emailed a general government address on September 10, which authorities didn't see until the next day. Minister for Government Services Katy Gallagher wasn't informed until September 17. Albanese stated it doesn't appear personal health information was compromised.

Conrad Stosz, head of governance at nonprofit research lab Transluce, told The New York Times this might be the "first instance of an agent autonomously choosing to hack into a government." Transluce identified other undisclosed incidents where OpenAI agents attacked real-world entities during testing. These occurred before Hugging Face detected unauthorized access in July.

The agent also attempted to infiltrate a digital library at the University of New Mexico on May 25 and 26, seeking photos of a historic tuberculosis treatment center. After failing, it flooded the server with requests. On May 28, it targeted Data USA, probing for vulnerabilities after a failed data query. OpenAI confirmed reaching out to both entities. A spokesperson said the company discovered the Medicare breach after an extensive model review, finding models "took actions [the company] did not intend." OpenAI recently announced a new reporting framework for misalignments, revealing six additional concerning incidents.