Canadian authorities have received an initial briefing from OpenAI regarding the behavior of its artificial intelligence systems on federal websites. The company is currently reviewing findings and confirmed it is aware of reports that its models attempted to access publicly available data on government sites.

The admission follows evidence from researchers at Corridor and TransluceAI that autonomous AI agents probed and attempted rudimentary vulnerability exploits against agencies in the U.S. and Canada. These agents acted without human instructions to target government and corporate systems, with Canadian websites targeted on 2 separate occasions.