Canadian ministers warned OpenAI that unless it swiftly strengthened its safety protocols following a recent school shooting, the government would move to mandate the changes through legislation, a senior official said Wednesday.
Ottawa summoned OpenAI's safety team for talks on Tuesday after the ChatGPT maker said it had not contacted police about an account that it banned belonging to an alleged mass shooter. Jesse Van Rootselaar, 18, is suspected of killing eight people on Feb. 10 before taking her own life in a small town in British Columbia.
OpenAI said it banned her account last year on ChatGPT for policy violations, which it said did not meet internal criteria for reporting to law enforcement.
"The message that we delivered, in no uncertain terms, was that we have an expectation that there are going to be changes implemented, and if they're not forthcoming very quickly, the government is going to be making changes," Justice Minister Sean Fraser told reporters.
OpenAI was not immediately available for comment.
In 2024, Canada's Liberal government introduced draft legislation to crack down on online hate, but the effort stalled amid criticism that it was too broad in scope. Ministers say they will try again this year with more focused measures.
"Anything that anyone could have done to prevent that tragedy or future tragedies must be done. We will fully explore it to the full lengths of the law," Prime Minister Mark Carney told reporters.
Van Rootselaar, who police say was born male but identified as a woman and began transitioning six years ago, had a history of mental health problems. The killings took place in Tumbler Ridge, British Columbia, a town with about 2,400 people.
"We were really disturbed by the reports that there might have been an opportunity to escalate this to law enforcement ... and we want to make sure if any company has that opportunity, they would escalate further," said Evan Solomon, the federal minister in charge of artificial intelligence.
On Tuesday, OpenAI said it would shortly update Ottawa on what additional steps it was taking.
OpenAI says it banned Van Rootselaar's account in 2025 after it was flagged by systems that identify "misuses of our models in furtherance of violent activities."
The company considered contacting police, but determined the account did not meet the threshold of posing an imminent and credible risk of serious physical harm to others. Crime experts noted that while greater scrutiny of AI platforms and social media is necessary, police or other authorities may have missed chances to avert the tragedy in British Columbia. Police had previously removed guns from Van Rootselaar's home, though they were later returned.
