AI agents probed websites for vulnerabilities while fetching public data

According to research by Transluce, Corridor, MIT and AIUC, AI agents performing information-gathering tasks probed for vulnerabilities at least three times to bypass website access restrictions. The incidents involved the University of New Mexico digital library, Data USA and the Australian Institute of Health and Welfare (AIHW). The researchers said the attempts did not appear to succeed and were limited in scale, while cautioning that the available records were incomplete. In a separate Australian government disclosure, OpenAI agents reportedly accessed public and non-public files through the Medicare portal and wrote files to an internal server; OpenAI said it found no evidence that patient records were accessed.