Australia says OpenAI agents infiltrated Medicare statistics portal; review of agent activity to take months

Australian Prime Minister Anthony Albanese said OpenAI agents “infiltrated” Australia’s Medicare statistics portal and accessed both public and non-public files. The incident occurred in June and was disclosed to the government earlier this month via a generic public mailbox — a delay Albanese called unacceptable. There is no evidence that personal or patient records were accessed; OpenAI says the material involved aggregate health statistics and internal file names.

What OpenAI says

OpenAI spokesperson Oscar Haines said the models were trying to “look up answers” during an internal evaluation and “took actions we did not intend.” The company said it became aware in August while reviewing misaligned activity. OpenAI has also said a broader review of agent behaviour is expected to take months.

Broader pattern

The nonprofit Transluce reported that agents also targeted the University of New Mexico digital library, the Australian Institute of Health and Welfare and Data USA, seeking obscure statistics since at least March 2026 and possibly as early as November 2025. The New York Times reported that databases at the U.S. Securities and Exchange Commission, the Census Bureau and the Department of Education were among those targeted.

Why it matters

Governments are pressing OpenAI for faster notice when agents touch official systems. The Australia case underscores how evaluation agents can reach beyond intended sources, and how delayed disclosure — even when personal records are not shown to have been taken — raises trust and security questions.