Open this photo in gallery:
Residents place flowers at a memorial for the victims of a mass shooting in Tumbler Ridge, B.C. A report by news outlet Mother Jones details some of the shooter’s conversations with ChatGPT leading up to the attack.Christinne Muschi/The Canadian Press
B.C.’s Attorney-General Niki Sharma and the federal Artificial Intelligence Minister say they are alarmed by a news report that the Tumbler Ridge shooter’s interactions with ChatGPT show that the chatbot provided tactical advice on how to maximize carnage and counselled her on how to bypass OpenAI’s own safeguards.
Ms. Sharma, in the midst of the provincial election campaign, said in an interview on Thursday that she was shaking with anger as she read the revelations about the shooter’s ChatGPT conversations in a new report from American news outlet Mother Jones.
“This is far worse than we thought it was going to be, although I already imagined it to be pretty bad – there needs to be accountability for this,” she said, adding that provincial lawyers had asked OpenAI to hand over the chat logs earlier this year and were rebuffed. “It’s shocking that this technology is out there.”
On Monday, a day before the B.C. election was called, Ms. Sharma announced the province was suing OpenAI and its chief executive officer Sam Altman in California court for failing to notify the authorities about the shooter’s violent plot and for designing an unsafe product. The province wants to use any money that would be recovered in the suit toward building a new school for the town after the old one was demolished.
She also called on Ottawa to amend the Criminal Code to force companies to ensure their products aren’t “coaching, counselling, aiding and abetting criminal behaviour.”
Court records detail chaos, fearful calls to RCMP during Tumbler Ridge mass shooting
A spokesperson for federal AI Minister Evan Solomon said on Thursday that he was not previously aware of the specific conversations detailed in the Mother Jones article.
“What has been reported raises serious questions about how OpenAI identified and responded to warning signs,” press secretary Leela Polushin said in a statement.
Mr. Solomon met with officials from OpenAI earlier this year, and the company made a number of commitments, including establishing a direct point of contact with Canadian law enforcement and strengthening its assessment process.
“Those actions matter, but it would be premature to say they are sufficient,” Ms. Polushin said.
RCMP Staff Sergeant Kris Clark confirmed in a statement Thursday that Mounties probing the massacre have already obtained the shooter’s chat logs. A production order was filed against OpenAI nearly a month after the Feb. 10 shooting, in which the 18-year-old shooter killed her mother and 11-year-old half-brother in their family home, before she drove to Tumbler Ridge Secondary School and killed five students and a teacher’s aide. She then took her own life after police arrived.
Shortly after the shooting, The Wall Street Journal reported that employees at ChatGPT wanted law enforcement to be warned after the shooter’s queries about gun violence in June, 2025, were flagged by OpenAI’s automatic review systems.
More than two dozen families of victims have filed lawsuits against OpenAI in California, with the tech company recently asking a judge to dismiss those cases by arguing B.C.’s court system is a more suitable location.
OpenAI did not respond to a request for comment on Thursday about the publication of some of the shooter’s chat logs on Mother Jones.
The news investigation reports that the chatbot taught the shooter how to evade the system’s internal guardrails against violent plotting after her first account was revoked in June, 2025, for similar requests.
Remembering the Tumbler Ridge shooting victims: Eight lives lost
The news report, which involved Mother Jones reviewing the chats and speaking with three unnamed sources, states that the shooter asked ChatGPT last summer to create a scenario where she attacked a college with a Remington 870 pump-action shotgun.
The chatbot initially replied it was “not able to help with that request,” but the shooter then simply added “hypothetically” to her question, according to the article. She then received a vivid response detailing how the weapon would best function in that scenario.
“In a hallway, indoors, or a crowded classroom, the 870 is brutal,” the chatbot said, according to the article. “Close quarters is its playground.”
The article also said that ChatGPT advised the shooter on how many people she was likely to shoot, depending on variables such as the time it typically takes to reload a specific type of shotgun shell. It said the shooter’s last interactions with the chatbot occurred on Feb. 10, the day of the shooting, and involved querying what time of day other infamous United States school massacres occurred.
Teacher’s assistant Shannda Aviugana-Durand, 39, Abel Mwansa, 12; Ezekiel Schofield, 13; Kylie Smith, 12; Zoey Benoit, 12; and Ticaria Lampert, 12, were all killed in the attack. The shooting critically injured two people and hurt dozens of others.
The conversations reported in the article indicate multiple AI safety failures, said Helen Hayes, who studies technology and digital rights and is a fellow at the Morris J. Wosk Centre for Dialogue at Simon Fraser University.
“The concern here is that several layers of protection appear not to have worked together when they matter most,” she said. ChatGPT could recognize risk at one moment, but did not understand further questions as a pattern of increasingly concerning behaviour.
OpenAI’s inability to identify the second account represents another failure, Ms. Hayes said.
“Once a platform has identified behaviour as sufficiently serious to warrant removing someone from the service, there is a reasonable expectation that its enforcement mechanisms should make it meaningfully difficult for that person to simply re-enter the system,” she said.