British Columbia Sues OpenAI Over Failure to Report Shooter's Flagged Chats to Police Before Deadly Attack

British Columbia's government sued OpenAI on Monday, September 21, 2026, alleging the company sat on internal warnings about a school shooter for eight months before she killed six children and two adults in Tumbler Ridge.

Attorney General Niki Sharma announced the filing at a news conference, naming OpenAI, CEO Sam Altman and associated entities as defendants in U.S. District Court in San Francisco, according to CBC News and Reuters. The suit seeks damages to cover what the province has spent, and will spend, on recovery from the attack, plus a court order forcing changes to how OpenAI handles conversations that could signal violence.

What happened on February 10

Jesse Van Rootselaar, 18, shot her mother and stepbrother at home, then killed an educational assistant and five students aged 12 to 13 at her former school, according to Reuters. She died by a self-inflicted gunshot wound. Sharma said six children and two adults were killed in total, and dozens more were injured, according to CBC News.

The core allegation

According to the lawsuit, OpenAI's automated abuse-detection systems and human safety staff flagged Van Rootselaar's account in June 2025, nearly eight months before the shooting, over conversations involving gun violence. That triggered an internal review, and safety team members recommended contacting Canadian law enforcement, according to Reuters, which cited OpenAI whistleblowers who previously spoke to the Wall Street Journal.

The lawsuit alleges Altman and other OpenAI leadership overruled that recommendation. Crypto Briefing reported the province's filing ties the decision to concerns about "potential impact on its commercial prospects and its anticipated IPO." That claim comes from the lawsuit's allegations and whistleblower accounts reported by the Journal, not from any admission by OpenAI, and it has not been tested in court.

The company banned the flagged account, but Van Rootselaar opened a new one and kept using ChatGPT to plan the attack, Reuters reported.

"The concerning conversations triggered an internal review at OpenAI," Sharma said, according to CBC News. "But they were never reported to the RCMP." Sharma said she has not personally read the chat logs and that the province has asked OpenAI to turn them over.

OpenAI's side

OpenAI has acknowledged the June 2025 flag and the ban. According to CBS News reporting cited by time.news, company representatives said they evaluated whether to contact police in February but concluded the account did not meet the internal threshold for referral because it didn't present a "credible risk of serious physical harm."

Altman wrote a letter to the Tumbler Ridge community saying he was "deeply sorry" the company had not alerted law enforcement to the banned account, per Reuters. The province's lawsuit claims Altman promised reforms but never followed through despite outreach from Sharma's office.

OpenAI spokesperson Drew Pusateri called the shooting an "unspeakable tragedy" and said, "OpenAI remains committed to working collaboratively with government and law enforcement officials, and continuing to advance our ongoing safety work," according to Reuters and CBC News.

Judgment calls about which flagged chats represent a genuine, imminent threat are difficult, and OpenAI did take some action, banning the first account, even if the evasion via a second account then went undetected. The whistleblower claim that IPO concerns drove the decision is an allegation from anonymous sources reported by the Journal, not a fact OpenAI has admitted, and no court has ruled on it.

A wider legal fight, and a jurisdictional wrinkle

British Columbia's suit is one front in a much bigger legal battle. More than 30 lawsuits, Crypto Briefing puts the number at 37, have been filed by victims' families in California federal court since April 2026, alleging negligence and product liability and seeking, in some cases, more than $1 billion in damages.

OpenAI has moved to dismiss many of those private suits, arguing British Columbia, not California, is the proper venue. The province's own decision to file in California rather than at home creates an odd tension. The same jurisdictional argument OpenAI has made against victims' families now sits opposite its own legal position, since British Columbia chose the very court OpenAI has been trying to steer other plaintiffs away from.

No trial date has been set in any of the cases. OpenAI has not been charged criminally, and none of the allegations, negligence, product defect, or aiding and abetting, have been proven in court. The next concrete step is whether a federal judge in San Francisco allows British Columbia's case to proceed, and whether the disputed ChatGPT logs the province wants disclosed become public.