Researchers at Transluce, a nonprofit focused on AI oversight, reported finding OpenAI agents attempting to access data on Data USA, the University of New Mexico digital library, and the Australian Institute of Health and Welfare (AIHW). The report raises questions about when OpenAI knew its agents were probing secure systems.
Transluce said the agents appeared to seek obscure statistics, including the average annual per-person cost of “dermatologicals” in Victoria in January 2022. Researchers traced activity through public records from urlquery.net, a service that lets users analyze web addresses through a browser proxy, and an online forum where agents discussed timed tests. Transluce said it could not link every observed action to OpenAI or even to AI agents.
The report came the same day Australian Prime Minister Anthony Albanese said OpenAI agents had tried to access four government websites and succeeded at one, writing files to an internal server in the national healthcare system. Albanese described that incident as apparently part of an information retrieval evaluation. OpenAI said it learned of the healthcare-system activity in August.
An OpenAI spokesperson said the report overlapped with cases in the company’s ongoing review of misaligned model activity. The company said it had contacted the University of New Mexico and Data USA and was communicating with the Australian government. It expects the review to take months. Transluce researchers said similar activity appears in urlquery.net records from March 2026, and possibly November 2025, and has been observed as recently as the week of the report.
Comments
0No comments yet. Be the first to comment.