British Columbia sues OpenAI over alleged warning failure before school shooting

British Columbia and the Peace River South school district filed a federal lawsuit in San Francisco against OpenAI entities and CEO Sam Altman, alleging the company failed to warn law enforcement about threatening ChatGPT conversations before the Tumbler Ridge school shooting.

The lawsuit seeks compensation and court-ordered changes to OpenAI’s handling of conversations involving threats of violence. The province alleges that a warning could have prevented the shooting, but neither that causation claim nor the complaint’s product-liability allegations have been adjudicated.

In its statement announcing the legal action, British Columbia said the case raises questions about technology companies’ responsibilities when they become aware of credible threats of serious violence. The complaint seeks compensatory and punitive damages, interest and costs, with the amount to be determined at trial. Claimed costs include community recovery, mental-health services and replacement of the school.

The requested injunction would require safeguards designed to reliably refuse, terminate or de-escalate conversations when users express an intent to harm others or seek help involving violence or weapons. A directly inspectable public copy of the complaint was not available in the research record, so its precise wording and the full list of requested measures could not be independently checked.

OpenAI discussed its handling of the case months before the lawsuit. In a February letter to Canadian ministers, the company said an automated system detected one perpetrator account in June 2025, sent it for human review and shut it down for violating policy. OpenAI said the conversations did not meet its threshold for a law-enforcement referral at the time because reviewers did not identify sufficiently credible and imminent planning.

Under its enhanced protocol, OpenAI said it would refer the same account if it were discovered now. The company also committed to more flexible referral criteria, input from mental-health, behavioural and law-enforcement experts, direct contacts with Canadian police, localized de-escalation resources and stronger detection of people who evade account restrictions. OpenAI said it found a second account after the perpetrator’s name became public and shared it with law enforcement. Those user-threat measures address a different safety problem from OpenAI’s misalignment disclosure framework, which concerns model-behavior incidents.

The province and OpenAI describe the earlier enforcement action differently. OpenAI called the June 2025 action a ban, while the complaint describes it as account deactivation rather than a user-level ban. The available evidence does not establish whether those terms describe materially different measures.

RCMP records say eight victims were killed on February 10—six at Tumbler Ridge Secondary School and two at a residence—and that the shooter also died. Responding to the lawsuit, an OpenAI spokesperson called the shooting an unspeakable tragedy and said the company remained committed to working with government and law enforcement and continuing its safety work.