British Columbia Sues OpenAI and Sam Altman Over the Tumbler Ridge School Shooting

The government of British Columbia filed a civil lawsuit against OpenAI and its chief executive, Sam Altman, in federal court in San Francisco on 21 September 2026, alleging that the company failed to warn police about a user's ChatGPT conversations before a school shooting in the province earlier in the year. The case, filed in the Northern District of California, is the first brought by a government rather than by victims or their families, and it asks a court to order changes to how ChatGPT handles conversations that could lead to violence.

Related reading: OpenAI, Anthropic, Google and Meta Testify Under Oath as New York Weighs Ten AI Bills · Four Subscribers Sue Anthropic, OpenAI, SpaceXAI and Google Over AI Slowdown.

The province's attorney general, Niki Sharma, announced the filing on Monday, saying the province is seeking accountability from OpenAI and its decision-makers. The complaint names Altman as an individual defendant alongside the OpenAI Foundation, OpenAI OpCo, OpenAI Holdings and OpenAI Group PBC, and demands a jury trial.

The filing concerns the attack of 10 February 2026 in Tumbler Ridge, a small community in northeastern British Columbia. Eight people were killed: the shooter's mother and eleven-year-old half-brother at the family home, then five students and an education assistant at Tumbler Ridge Secondary School, with two students wounded. The eighteen-year-old shooter, a former student at the school, died by suicide at the scene. The complaint describes the school building as having been demolished, and the province is seeking to recover the cost of replacing it.

What the complaint alleges

The suit's central allegation concerns OpenAI's internal review process. It states that in June 2025 the company's automated monitoring identified the shooter's account and routed it to the human review team that assesses users who may be planning violence. That team, according to the complaint, examined the conversations, concluded there was a credible and specific risk of harm to others, and recommended referring the account to the Royal Canadian Mounted Police.

OpenAI leadership rejected that recommendation, the complaint says. Instead of making a referral, the company deactivated the account and notified no one outside the company. The complaint alleges that deactivation, without a user-level ban or heightened monitoring, let the user continue engaging with ChatGPT through a second account in the months before the attack. It cites OpenAI employee whistleblowers, whose account it says was first reported by the Wall Street Journal.

The complaint quotes a letter Altman published to the community in April 2026, in which he wrote, "I am deeply sorry that we did not alert law enforcement to the account that was banned in June." The province treats that statement as an acknowledgment that OpenAI did not notify authorities after its safety team identified a risk.

The claims and what the province wants

The complaint lists eight causes of action, including negligence for failing to warn law enforcement, negligent entrustment, negligence in design and failure to warn, strict product liability, and a claim for aiding and abetting a mass shooting. The province argues that ChatGPT's role in the attack was foreseeable, pointing to what it describes as OpenAI's own policies and public statements about reporting threats.

The remedies it seeks are financial and structural. The province wants compensation for costs it says it incurred and will continue to incur, including emergency response, health care, victim and mental health services, educational continuity, and the construction of a replacement school and wellness centre. The claim does not name an exact figure, saying only that it exceeds 75,000 dollars and that the province will seek all damages recoverable under California law, including punitive damages. Separately, it asks for an injunction requiring OpenAI to implement safeguards that reliably refuse, terminate or de-escalate conversations in which users express intent to harm others.

Filing in California rather than Canada is a deliberate choice. Sharma said the province chose the venue because that is where OpenAI made the decisions at issue, and because California law gives the province different tools to pursue the company. The complaint also names Altman personally, which places the chief executive's own conduct and statements at the centre of the case rather than treating the company as an abstraction.

A widening set of lawsuits

The province's suit joins a growing group of claims against OpenAI over the same attack. More than 30 lawsuits were filed in the same San Francisco court in early September by victims, survivors and witnesses, following seven filed in the spring by families of victims. Those cases, brought by the Edelson law firm and others, allege negligence and seek a jury trial. On 17 September, counsel for OpenAI filed motions to dismiss, arguing among other things that the cases belong in a different court, with a hearing scheduled for 5 November 2026.

The government case adds a different kind of plaintiff. Private suits turn on the harm to individuals and families. A provincial government suing for recovery costs turns on the public expense of an attack and the duty a company owes to warn authorities when its systems detect a credible threat. That framing may be easier to argue than individual negligence, because the costs are concrete and documented, and it may be harder, because it requires a court to find a duty that runs from a private company to a provincial government.

The case also arrives while OpenAI is under pressure on several fronts, including scrutiny of its safety practices and the wider debate over how much responsibility AI developers bear for how their products are used. TechPastWeek has covered OpenAI's own framework for reporting model misalignment and the class action alleging that frontier labs agreed to slow development, both of which concern how much the companies should be trusted to police themselves.

The legal question underneath

At the centre of the case is a question that product liability law has only begun to answer for software that generates text. When a company's monitoring system identifies a user as a credible threat, does the company have a duty to warn law enforcement? The province argues that it does, and that OpenAI's own policies created that duty by describing when the company would refer threats to authorities. OpenAI's expected response, based on its filings in the related cases, is that the shooter's criminal act broke the chain of causation and that the company is not liable for a third party's violence.

The complaint tries to close that gap by arguing that the attack was exactly the harm its review process was meant to prevent. If the company had identified the user as a credible threat and referred the account, the province says, a warning would have reached an RCMP that had already had contact with the user over mental health concerns. That argument turns the foreseeability question into the heart of the case: an attack that a company's own team flagged is, by definition, foreseeable to that company.

Whether a court accepts that reasoning is unresolved. Courts have been reluctant to hold platforms liable for user speech, and Section 230 of the US Communications Decency Act, which shields platforms from liability for third-party content, is likely to feature in OpenAI's defence. The province's focus on the company's own safety review, rather than on the content itself, is an attempt to frame the claim as a failure of a specific duty rather than as liability for what a user wrote.

What to watch

Three dates and questions will shape the case. The first is the 5 November hearing on OpenAI's motions to dismiss in the related suits, which will indicate how the court views the core liability theory. The second is whether the province's case is consolidated with the private suits or proceeds on its own, since a government plaintiff may face different defences. The third is whether OpenAI changes its threat-reporting practices, either in response to the litigation or on its own, because a voluntary change would not resolve the case but would shape the remedy if the province prevails.

For now the filing is a set of allegations, and the company has not had its say in this case. The province is asking a court to decide that a company which detects a credible threat has an obligation to act on it. How that question is answered will matter well beyond Tumbler Ridge, because it would set expectations for every developer whose systems see what users are planning before anyone else does.

Cover: British Columbia Premier David Eby at a media availability in Tumbler Ridge. Image by the Government of British Columbia, via Wikimedia Commons, CC BY 4.0, cropped 16:9.