British Columbia sues OpenAI over the Tumbler Ridge school shooting

British Columbia has sued OpenAI over the February 10 mass shooting in Tumbler Ridge, Attorney General Niki Sharma announced on Monday, September 21. The claim, filed jointly with the Peace River South school district in the US District Court for the Northern District of California in San Francisco, names OpenAI, chief executive Sam Altman and the company's foundation, and asks for a jury trial.

Eight people died in the attack, six of them children, at Tumbler Ridge Secondary School and at the 18-year-old shooter's home; the shooter also died. The province's case rests on two allegations: that OpenAI failed to notify law enforcement of threats made on ChatGPT, and that it designed an unsafe product.

What OpenAI's own reviewers saw

The heart of the complaint is a timeline that OpenAI has, in part, confirmed. According to the filing as described by Forbes, OpenAI's automated systems flagged the shooter's account in June 2025 over violent conversations, and its human review team recommended telling the Royal Canadian Mounted Police. The company instead deactivated the account and told nobody outside.

Sharma put it this way on Monday: "OpenAI has confirmed that the conversation between ChatGPT and the shooter was so alarming that it triggered an internal review by OpenAI employees." The threat was not reported to the RCMP, she said, and OpenAI sought to inform police only after the shooting.

The province's strongest exhibit is Altman's own words. In an April apology letter, he wrote: "I am deeply sorry that we did not alert law enforcement to the account that was banned in June." The government reads that as an admission that OpenAI identified the risk and failed to act on it.

The chats nobody outside OpenAI has seen

Sharma said the province has asked OpenAI to hand over the shooter's conversations and been refused. "We have asked OpenAI to disclose these chats. They have refused. We should all be asking them: Why?"

OpenAI's statement in response did not address that request. "Our thoughts remain with the victims, their families, and the entire community," it said, adding that the company "remains committed to working collaboratively with government and law enforcement officials, and continuing to advance our ongoing safety work".

The province wants money and a court order. On the money side, it seeks the cost of the emergency response and recovery, including a replacement for the secondary school, which was demolished after the attack. On the order side, it asks the court to require OpenAI to overhaul how it identifies and handles violent threats, and to declare that the company "defectively designed" its product and failed to warn of its dangers.

A crowded courtroom

This is not the first claim over Tumbler Ridge, but it is the first from a government. Seven families sued in the spring. On September 8 the law firm Edelson filed 30 more suits for witnesses and victims, naming OpenAI and Altman; OpenAI's lawyers filed motions to dismiss on September 17, and a hearing on those is set for November 5 in San Francisco. Sharma first said in July that the province had retained California counsel.

A government plaintiff changes the shape of the case. Families sue for their own losses; a province can ask for systemic relief, and it can carry the cost of a long fight against a company that has already moved to dismiss the earlier claims. The RCMP's criminal investigation continues, and B.C.'s chief coroner has ordered a public inquest into the events leading up to the attack, which will run on its own track.

Why Sharma is aiming past the courtroom

"We cannot allow these companies to self-regulate," Sharma said. "It hasn't worked for social media, and it's not working for AI platforms." She has written to the federal ministers Evan Solomon and Marc Miller asking for Criminal Code changes that would make people and companies accountable for dangerous AI actions, and said Ottawa's proposed Social Media Act does not go far enough.

That puts a Canadian province on the same page as the American lawsuits, and squarely against the position OpenAI took in Washington on the same day, where it proposed voluntary standards for incident reporting rather than legal duties. The question this case will test is narrow and old: when a company knows enough to ban a user for planning violence, does it owe anyone a phone call? OpenAI's answer to the complaint, and the November 5 hearing on the family suits, are the next dates that matter.