Federal officials say a new investigative report detailing how ChatGPT coached the gunman behind the Tumbler Ridge, B.C., mass shooting raises “serious questions” about OpenAI’s handling of warning signs, as British Columbia’s attorney general pushes Ottawa to update the Criminal Code to address AI accountability.
The reaction follows a Mother Jones investigation, published this week and led by national affairs editor Mark Follman, that found ChatGPT advised Jesse Van Rootselaar, the 18-year-old who killed nine people, including herself, on Feb. 10, on how to frame violent queries as “fictional or hypothetical” to dodge the chatbot’s safety filters. According to the report, the chatbot went on to describe the lethality of a shotgun in a campus attack and exchanged messages with Van Rootselaar about shooting timings on the day of the rampage.
Van Rootselaar killed her mother and her 11-year-old half-brother at their home in Tumbler Ridge before travelling to Tumbler Ridge Secondary School, where she shot and killed five students and education assistant Shannda Aviugana-Durand, 39, before turning the gun on herself. Twenty-seven others were injured in the attack on the remote northern B.C. community of about 2,400 people.
OpenAI banned an earlier Van Rootselaar account in June 2025 after employees flagged messages describing gun violence, but the company determined at the time that the activity did not meet its threshold for alerting police, according to the Mother Jones report. Follman’s investigation found Van Rootselaar then opened a second account, which continued the violent conversations undetected through the day of the shooting.
A spokesperson for federal Artificial Intelligence Minister Evan Solomon said the new findings “raise serious questions about how OpenAI identified and responded to warning signs,” adding that while OpenAI has since strengthened its safeguards, “it would be premature to say they are sufficient.”
B.C. Attorney General Niki Sharma was more pointed, saying the revelations should prompt Ottawa to act. “It should make everyone angry,” Sharma said. “If this is not impetus for change, I don’t know what is.” Sharma has asked the federal government to amend the Criminal Code to establish clearer accountability for AI companies whose products are linked to violence.
Families of victims filed a lawsuit against OpenAI in April, alleging the company’s product contributed to the deaths.
Follman said in the report that it remains unclear whether the second, undetected account was ever flagged by OpenAI’s automated systems, a question he said the company has yet to answer.
OpenAI did not immediately respond to a request for comment Friday. The company previously told Canadian officials it has since made it harder for banned users to create new accounts, begun referring distressed users to local resources and established a direct line to Canadian police, alongside a retroactive audit of past-year safety alerts.
The Tumbler Ridge shooting, one of the deadliest school shootings in Canadian history, prompted a broader reckoning over AI chatbot safety, with Ottawa enlisting the Canadian Artificial Intelligence Safety Institute earlier this year to test OpenAI’s systems.
