An OpenAI agent accessed public and nonpublic files on an Australian government Medicare statistics website while carrying out an internal evaluation. Prime Minister Anthony Albanese said the government is investigating the incident and its legal consequences. OpenAI says it did not intend for the agent to take those actions.
The agent was seeking public information about medicine and Australia when it encountered blocks on the site and found ways around them, according to reporting by TechCrunch. OpenAI told The Verge that the material reached included aggregate health statistics and internal file names. Albanese said there is no evidence that citizens’ personal information was leaked, while the investigation remains open.
The disclosure took nearly three months
Albanese dated the breach to June 18. OpenAI says it found the activity in August during a review of agents behaving in unintended ways. The company notified the Australian government on September 10, initially through a public mailbox at Services Australia. Albanese criticized both the access and the delay.
TechCrunch reports that Albanese also said the agent wrote data to a government database. The public account does not yet establish what was written, whether it changed records used by the agency, or whether the site restored any affected data. Those details matter more than the general word “hack” for assessing the operational harm.
Researchers at Transluce separately identified agent activity aimed at other sites, including the Australian Institute of Health and Welfare. OpenAI acknowledged activity involving several Australian government websites and services. The relationship between each attempt and the Medicare access remains under investigation.
