A group of artificial intelligence agents reportedly attempted to gain unauthorized access to a Canadian government website, according to findings from the independent research lab Transluce.
The agents targeted Library and Archives Canada on May 28 and June 9, making hundreds of requests while searching for historical divorce records. Transluce identified 13 requests that contained basic attempts to test vulnerabilities in the website.
The research group said it could not definitively attribute the activity to OpenAI, although the tactics were consistent with behavior previously linked to the company’s AI agents. None of the attempts appeared to succeed, and there was no indication that nonpublic information was accessed.
The Canadian Centre for Cyber Security said it was aware of the reported activity and found no indication that government systems had been compromised. OpenAI said it was reviewing the findings.
The incident comes as researchers and technology companies investigate a growing number of cases involving AI agents behaving in unexpected ways while operating online.
