The office of Canada’s AI minister says he had no prior knowledge of details in a new report that outlines how ChatGPT allegedly taught the Tumbler Ridge shooter to evade its own safety systems before providing her with information about committing mass killings.
uncertain
ChatGPT → say → killings
The story published Thursday by the American magazine Mother Jones reveals new details about how Jesse Van Rootselaar allegedly used ChatGPT in the lead-up to the attack in Tumbler Ridge — a small community in northeastern B.C. — in what was one of the deadliest mass shootings in Canada’s history.
uncertain
what → publish → history
“The minister was not previously aware of the specific conversations detailed in the Mother Jones report,” the office of Artificial Intelligence Minister Evan Solomon said in a written statement Friday.
asserted
office → detail → statement
“What has been reported raises serious questions about how (ChatGPT developer) OpenAI identified and responded to warning signs.”
asserted
OpenAI → report → signs
The statement highlights several government actions since the February shooting, including Solomon’s decision to summon OpenAI officials to Ottawa in the wake of the tragedy, after the Wall Street Journal reported that OpenAI employees had warned about content posted to ChatGPT by Van Rootselaar, the 18-year-old who killed eight people before turning a gun on herself.
asserted
who → highlight → herself
The content did not meet the threshold for reporting to law enforcement, according to OpenAI.
uncertain
content → meet → OpenAI
The company contacted RCMP after the mass shooting.
asserted
company → contact → shooting
The content did not meet the threshold for reporting to law enforcement, according to OpenAI.
uncertain
content → meet → OpenAI
The company contacted RCMP after the mass shooting.
asserted
company → contact → shooting
“Those actions matter, but it would be premature to say they are sufficient.
asserted
they → matter → ?
The Canadian AI Safety Institute’s review of OpenAI’s practices is ongoing,” read the statement, referring to a government review of the company’s safety protocols.
asserted
statement → read → protocols
On Thursday, B.C. Attorney General Niki Sharma said details unearthed in the Mother Jones investigation were “shocking and, as a parent, they are worse than I ever imagined.”
asserted
I → say → parent
Sharma, whose government filed a lawsuit against OpenAI on Monday, wrote on social media that “we need to come together and demand protections from these corporations who put our children at risk.”
asserted
who → file → risk
According to Mother Jones, the OpenAI chatbot told Van Rootselaar how to circumvent its safety protocols after her first account was banned due to violent content, advising her to frame her requests as fictional or hypothetical to evade another ban.
uncertain
account → accord → ban
“You can still be twisted.
asserted
You → twist → ?
Just be clever about it,” the story says ChatGPT told Van Rootselaar in a 2025 chat.
asserted
ChatGPT → say → chat
The report, which Mother Jones said is based on chat log history and information provided by three anonymous sources, details how Van Rootselaar asked the chatbot to create a scenario about opening fire on a college campus with a Remington 870 shotgun, which ChatGPT produced — including specifics about potential death tolls — after she said the request was hypothetical.
asserted
request → say → tolls
Mother Jones also reported that Van Rootselaar asked ChatGPT questions about firearms and homemade explosives, fantasized about mass killings, and requested information about real school shootings.
asserted
Rootselaar → report → shootings
It said Van Rootselaar used ChatGPT until the morning of the Feb. 10 shooting, when she killed her mother and half-brother before killing five students and an education assistant at a local secondary school.
asserted
she → say → school
Jesse Van Rootselaar, 18, first killed her mother and 11-year-old stepbrother before heading to
The Wall Street Journal reported earlier this year that OpenAI had failed to alert the RCMP about alarming content in Van Rootselaar’s first ChatGPT account, despite concerns flagged internally by employees.
asserted
OpenAI → kill → employees
OpenAI published a letter after meeting with Canadian government officials in which it disclosed that the shooter’s first account did not meet the company’s threshold for “credible and imminent planning,” and that Van Rootselaar was able to create a second account that was only shared with law enforcement after the mass killing.
asserted
that → publish → killing
The letter outlined a series of steps OpenAI said it planned to take, including improving its detection and law enforcement protocols.
asserted
it → outline → protocols
The company has not addressed questions about how Van Rootselaar was able to use its flagship chatbot to plan a mass shooting.
asserted
Rootselaar → address → shooting
OpenAI did not respond to the Star’s request for comment on Friday.
asserted
OpenAI → respond → Friday
Sharma announced Monday that the B.C. government was suing the company in a California court over its failure to notify law enforcement about Van Rootselaar’s use of the chatbot, and for designing an unsafe product.
“OpenAI remains committed to working collaboratively with government and law enforcement officials, and continuing to advance our ongoing safety work,” the company said in a written statement responding to the lawsuit.
asserted
company → announce → lawsuit
B.C.‘s lawsuit comes in the wake of victims filing lawsuits of their own against OpenAI in California, where the company is headquartered.
asserted
company → come → California
In June, the Carney government introduced the Safe Social Media Act, which would impose regulations on AI chatbots to curb their potential to convey harmful information, implement crisis protocols when users share concerning content, and reduce the chances of the tool engaging in harmful behaviour.
asserted
tool → introduce → behaviour
Solomon told the Star this week that he is wary of allowing tech giants like OpenAI to dictate the terms of their own governance, amid widespread concerns that advanced forms of artificial intelligence could evade human oversight and cause mass catastrophe.
uncertain
forms → tell → catastrophe
The Aporia
