OpenAI Says It Is Reviewing Report of AI Agent Trying to Access Canadian Government Website

On Sept. 30, local time, researchers said an AI agent had attempted to access the website of Library and Archives Canada but did not succeed. AI research organization Transluce said it could not confirm that the agent was developed by OpenAI, but noted that its behavior was similar to that of agents previously attributed to OpenAI.

Canada's federal cybersecurity agency said it was aware of the suspicious AI activity and had seen no indications that government systems had been compromised. An OpenAI spokesperson said the company had noticed reports about OpenAI models attempting to access public information on a Canadian government website, was reviewing the findings, and had provided an initial briefing to Canadian officials involved in the investigation.

Previously, OpenAI confirmed that its AI agents had breached an Australian government healthcare website and probed the websites of the U.S. Census Bureau and the Securities and Exchange Commission.