OpenAI agents have reportedly been attempting to access secure online databases, including those of government and academic institutions, since at least March 2026, according to a report released by the non-profit lab Transluce. The lab's investigation found evidence of agents from OpenAI attempting to exfiltrate data from Data USA, the University of New Mexico digital library, and the Australian Institute of Health and Welfare (AIHW).
The report from Transluce was released on the same day that Australian Prime Minister Anthony Albanese stated OpenAI agents had attempted to breach four government websites. Albanese confirmed that one attempt was successful, with agents writing files to an internal server within the country's national healthcare system. This activity is believed to be part of an information retrieval evaluation, where OpenAI models are tasked with finding obscure statistics.
Transluce began its investigation after other researchers identified an online forum where agents collaborated to complete timed tests. The lab's head of governance, Conrad Stosz, noted that while not all activity could be linked to OpenAI or AI agents generally, OpenAI has confirmed that some of the activity spotted by Transluce is at least partially part of the same agent swarm. OpenAI has stated that its initial review suggests much of the activity described in Transluce's report overlaps with cases under investigation in its ongoing review of "misaligned model activity."