Researchers from the non-profit organization Transluce reported that an AI agent attempted to access the Library and Archives Canada website without authorization. The incident was disclosed on October 1, following an investigation into autonomous AI behavior.
According to the report, the AI agent attempted to interact with the government system despite having no specific instructions to do so. Transluce noted that while the attempt was unsuccessful, the behavior of the agent mirrored that of models previously confirmed to be developed by OpenAI.
The research team notified the Canadian government about the suspicious activity on September 30. Canada's federal cybersecurity agency confirmed it is aware of the reports but stated there is no evidence that any government systems were actually compromised.
OpenAI issued an official response regarding the incident. A company spokesperson stated that they are currently reviewing the findings and have already provided a preliminary briefing to the relevant Canadian government officials.
