B.C. Sues OpenAI Over Tumbler Ridge: The Duty-to-Warn Fight Explained

British Columbia has sued OpenAI and Chief Executive Sam Altman in federal court in San Francisco, alleging the Tumbler Ridge school shooting might have been prevented if the company had warned police about the shooter's use of ChatGPT to plan the attack. The Guardian

The province is seeking damages to help fund recovery after the February attack. It also seeks a court order requiring OpenAI to change how it handles ChatGPT conversations that point to a risk of violence.

To understand why where the case was filed matters, consider that suing in California brings a Canadian public plaintiff into the company's home court. It will also test how cross-border civil claims for harm (what lawyers call tort claims) can apply to a provider of an AI model.

Nine people, primarily children, were killed in Tumbler Ridge, British Columbia. The shooter was 18-year-old Jesse Van Rootselaar. She shot her mother and stepbrother at home, then killed an educational assistant and five students aged 12 and 13 at her former school on 10 February, then died by suicide.

The complaint focuses on earlier warning signs. It says Van Rootselaar's ChatGPT interactions were first flagged inside OpenAI in June 2025 after conversations about gun violence, but the company did not alert police. Her account was deactivated after that flag, the complaint says, but she obtained a new account and continued using the platform to plan the attack.

The lawsuit further alleges that members of OpenAI's safety team recommended contacting police about Van Rootselaar, but Altman and other leadership overruled them, according to whistleblowers who spoke to the Wall Street Journal.

The broader significance here is that, if proven, that claim would shift the focus from a failure of automated safety checks to a conscious choice not to alert authorities. That is also part of why Altman is named as a defendant alongside the company.

British Columbia is not suing alone. More than 30 family members of victims and others affected have filed separate lawsuits in California accusing OpenAI of missing a chance to prevent the shooting. That group includes 30 complaints in San Francisco federal court by people present at the shooting, described in subsequent coverage as students and teachers, and family actions alleging ChatGPT aided the shooting. One related action alleges negligence and wrongful death based on the shooter's interactions with ChatGPT in the weeks and months before the attack. NPR

Attorney General Niki Sharma said the province is suing OpenAI for failure to notify law enforcement of threats made on ChatGPT. CBC The government's newsroom had previously published a release titled "Exploring legal options related to Tumbler Ridge tragedy" about holding OpenAI accountable for failure to notify law enforcement, and the Attorney General later issued a statement on new legal filings concerning flagged threats made by the perpetrator on ChatGPT.

Two liability theories run through the provincial claim. British Columbia claims OpenAI designed an unsafe product and was negligent, meaning it failed to take reasonable care, by not flagging the suspect's ChatGPT activity to police. The Wall Street Journal One question is about design and safeguards, including how a user could return with a new account after deactivation. The other is about omission after internal warnings. Think of it as asking both whether the locks were weak and whether staff heard an alarm and did not call for help.

OpenAI has asked a California court to dismiss lawsuits related to the shooting. CBC Its lawyers claimed that any legal action related to the deaths and injuries should take place in British Columbia.

Looking at the early procedural fight, that argument about the proper forum is now more complex because the province itself chose San Francisco federal court.

The broader context here is how courts will allocate responsibility for foreseeable misuse of general-purpose systems. Rules for product liability, negligence and duty to warn — the idea that someone aware of a serious risk must speak up — were built for tangible goods and direct professional relationships. Applied to conversational AI, they raise difficult questions about monitoring obligations, privacy constraints, thresholds for referral to law enforcement, and circumvention through new accounts. A government plaintiff adds another layer, combining parens patriae style recovery costs with injunctive demands, or court-ordered changes, for trust and safety workflows.

Looking at what comes next for operators and regulators, the immediate fight will likely be procedural before it is substantive. Questions of jurisdiction, choice of law, admissibility of internal safety deliberations and whistleblower accounts, and the scope of any duty to report user threats will determine whether damages and redesign orders are even reachable. For specialists, the case file to watch is not only what OpenAI flagged in June 2025, but what standard a court sets for what a provider must do next.