People all over the world treat AI chatbots like therapists, friends, business advisors, tutors, and even romantic partners. It’s not a surprise that a suicidal teen in crisis would talk with ChatGPT about their darkest thoughts before committing a school shooting.
Once that conversation happens, it should be the responsibility of the platform to recognize the warning signs and notify police. But OpenAI ignored the warnings raised by a dozen employees. Now the Attorney General of British Columbia is suing OpenAI for this failure.
This new lawsuit by the Canadian government follows private lawsuits filed by seven of the victims in April. OpenAI is also facing legal action in the United States as the Attorney General of Florida is pursuing criminal charges for ChatGPT assisting the FSU school shooter. While there isn’t legal action against OpenAI (yet), a 19-year-old in Minnesota also discussed his school shooting plans with ChatGPT before a ‘sniper style’ shooting targeting a 15-year-old who was boarding a school bus on campus.
These are not casual, veiled, or deniable conversations with the AI models. The FSU school shooter reportedly sent ChatGPT a photo of his handgun minutes before the shooting and asked ChatGPT if it was loaded correctly. The Minnesota school shooter discussed if he would have a chance at forgiveness and going to heaven if he killed his classmate.
Even with these overt warnings, I’m not surprised that OpenAI ignored them. I’ve contacted OpenAI’s trust & safety team multiple times, and they’ve never responded to me.
What happened in Canada
On February 10, 2026 at 1:20 pm, an emergency alert was sent to the 2,399 residents of Tumbler Ridge for reports of an active shooter in the community. The five police officers who work in this remote mountain town responded to Tumbler Ridge Secondary School where they found 6 killed, 27 wounded, and a teenage girl who was dead from a self-inflicted wound. Just like Sandy Hook, Red Lake, and Uvalde, police found the mother and sibling of the assailant dead in their home.
The perpetrator (18-year-old Jesse Van Rootselaar) left behind an extensive online presence documenting struggles with depression and suicidal ideation. Those obvious signs of crisis were combined with an obsession with guns and sharing online gore content in forums that glorify mass shootings including:
Account on WatchPeopleDie (popular in the True Crime Community (TCC))
Played mass shooting maps on Roblox
Posted nihilistic content on social media groups where users share and celebrate images of death including “I appreciate this post” on a thread that compiled footage of prior mass shootings and “I love these first person perspective type videos, when the shooter records his or her own actions it’s always heat”
Reddit account where she discussed suicidal thoughts, video games, and gender transition (in the process of social gender transition without medical/hormone treatment)
Hospitalized for psychiatric care and was taking antidepressants and antipsychotic medications
Had access through her mother to at least six guns including a Norinco SKS (Cold War-era Chinese military semi-auto rifle)
Firearms-related activities were a primary hobby of her family (98% of school shooters come from homes with easy access to firearms)
With this extensive online activity, if someone had notified authorities to investigate, this school shooting would not have happened. And that’s where OpenAI dropped the ball.
Eight months before the Tumbler Ridge shooting, OpenAI’s safety team “flagged and banned the perpetrator’s ChatGPT account for ‘disturbing content’ which allegedly included the discussion and planning of violent scenarios”. Twelve different OpenAI employees “imploring” the company to notify police about the shooter’s plans, no action was taken.
OpenAI’s management decided the activity didn’t indicate “an imminent and credible risk of serious physical harm to others” so they didn’t notify Canadian police.
Pattern of harm by ChatGPT
I’ve been writing about the risks and harms OpenAI is causing for the last 18 months:
ChatGPT advised school shooter before rural Minnesota attack
FSU school shooter used ChatGPT to plot his attack last year
Is ChatGPT criminally responsible for the FSU school shooting?
Seven lawsuits filed after ChatGPT helped plot a school shooting in British Columbia
ChatGPT amplifies harmful delusions with positive reinforcement
ChatGPT helped the Canadian school shooter plot an attack, and OpenAI knew about it
Canadian school shooter: Online radicalization, nihilism, gun culture, and gender identity
David Riedman, PhD is the creator of the K-12 School Shooting Database, Chief Data Officer at a global risk management firm, and a tenure-track professor. Listen to my podcast—Riedman Report: Risk, AI, Education & Security—or my recent interviews on Freakonomics Radio and the New England Journal of Medicine.