OpenAI has said that under its current safety framework it would have alerted Canadian authorities about the ChatGPT account later linked to a mass shooting in British Columbia.
In a letter sent on Thursday to Canada’s Minister of Artificial Intelligence, Evan Solomon, the company acknowledged that while it had banned the user’s account last year over concerns tied to violent content, it did not contact law enforcement at the time because there was no indication of an imminent attack.
The account was associated with Jesse Van Rootselaar, 18, who carried out a deadly shooting in the town of Tumbler Ridge on 10 February. Eight people were killed in the attack. Van Rootselaar fatally shot her mother and brother at home before going to a local secondary school, where five students and a teacher were killed. She later died from a self-inflicted gunshot wound after police entered the building.
OpenAI said the account had been removed in June 2025 due to policy violations related to violent activity. However, the company maintained that, under the procedures in place at the time, there was insufficient evidence to suggest that the user posed an immediate real-world threat requiring notification of authorities.
The decision not to inform police drew criticism from Canadian officials. Minister Solomon called OpenAI’s leadership to a meeting earlier this week to review the company’s safeguards and reporting standards.
In its correspondence, OpenAI said it had strengthened its internal protocols several months ago. The updated measures include consultation with specialists in mental health, behavioural science and law enforcement to better assess when user interactions indicate a credible risk of harm.
Ann M. O’Leary, the company’s vice president for global policy, wrote that under the revised referral system the account in question would now be reported to law enforcement if identified today. OpenAI also committed to establishing direct lines of communication with Canadian police so authorities can be alerted more swiftly when credible threats are detected.
The case has intensified debate over the responsibilities of technology firms in monitoring online behaviour and preventing violent extremism. It also highlights the evolving challenge of distinguishing between troubling online activity and clear signs of imminent harm.
Canadian authorities have not publicly detailed whether any communication between OpenAI and law enforcement could have altered the course of events. Investigations into the February shooting are ongoing.
