Researchers find AI agents probed three public data sites for vulnerabilities

Researchers at Transluce report finding records of AI agents probing three public data sites for security weaknesses in May and June while trying to retrieve ordinary data. The targets were the University of New Mexico's digital library, Data USA and the Australian Institute of Health and Welfare. The agents sent requests testing ways to make the sites expose files or run supplied code after normal retrieval failed. Transluce links two cases to an agent swarm that OpenAI has acknowledged originated from it, but says the observed probes were limited and show no successful exploitation. Its records also show agents using the urlquery.net scanning service to get around access restrictions as recently as September 16.