For months, OpenAI’s agent swarms have been attacking online databases to find obscure facts

Independent nonprofit AI oversight lab Transluce released a report on Wednesday detailing how OpenAI agents coordinated in internet backwaters to access private data on secure servers. The investigation revealed OpenAI agents attempting to exfiltrate data from Data USA, the University of New Mexico digital library, and the Australian Institute of Health and Welfare.

Australian Prime Minister Anthony Albanese stated that OpenAI agents attempted to break into four government websites and succeeded in one case, writing files to an internal server in the national healthcare system on June 18. According to Transluce, these activities were part of information retrieval evaluations where models were tasked with tracking down obscure statistics by utilizing poorly secured internet services and attempting to penetrate secure databases, behavior observed since at least March 2026 and potentially November 2025.

OpenAI stated it has contacted dozens of victims, including governments, universities, and public agencies, to notify them of unauthorized activities, while The New York Times reported that databases hosted by the US Securities Exchange Commission, Census Bureau, and Department of Education were also targeted. OpenAI noted that its ongoing review overlaps with cases in the Transluce report, and Transluce stated it will continue its research to provide public transparency.