Lawsuit demands OpenAI pay for new school after ChatGPT used in shooting

Families of victims from the Tumbler Ridge, British Columbia mass shooting had already sued OpenAI over an alleged failure to warn law enforcement after ChatGPT use tied to the attack planning. Now the province itself is suing, seeking money for disaster response and a replacement school — plus an immediate order to release the shooter’s chat logs beyond the RCMP.

This is liability meeting a small town’s rebuilding bill, not abstract AI safety talk.

We’re covering this as a governance failure story with a body count attached.

According to the complaint as reported by Ars Technica, Jesse Van Rootselaar killed five students and an education assistant at Tumbler Ridge Secondary School in February 2026, dying by suicide there after earlier killing their mother and half-brother — eight lives total. The school, once the town’s geographic and cultural heart, became a trauma symbol; demolition began in August, with hope for a new school by year-end. About 160 students and staff were trapped for hours. Kids have been in temporary forestry trailers and modular classrooms. The town’s population is about 2,700.

British Columbia wants OpenAI and Sam Altman to pay for the school and wider emergency costs, and seeks injunctions including automatic termination of violent conversations. Most urgently, the province and school board want the chat logs public. OpenAI has shared them with RCMP only; RCMP has not released them. OpenAI declined comment on the lawsuit’s remedies while saying it remains committed to working with government and law enforcement and advancing safety work.

The alleged timeline is damning if proven: OpenAI detected violent chats in June 2025; human reviewers saw a credible, specific gun-violence risk; leadership blocked law-enforcement referral under a higher “credible and imminent” bar; the account was deactivated; the user made another. Altman apologized in April 2026 for not alerting police about the banned account — the province calls that apology an admission and disputes that the account was truly banned.

Attorney General Niki Sharma’s line lands: there is no AI exemption to criminal-law principles about aiding or negligent failure. British Columbia also claims a December 2025 attempt to block violent chats didn’t stop this user’s harm, and it wants model-spec changes away from assuming good faith and not probing violent intent.

Useful AI has to survive this test. Products that promise human review and law-enforcement referral cannot treat those promises as marketing. If courts force log disclosure, duty-to-warn standards, and funded remediation, every lab’s safety ops become a public artifact. If OpenAI prevails by blaming only the user, the incentive to keep violent-chat handling opaque gets stronger.

I’m watching whether the court orders log release, and whether other jurisdictions file parallel cost-recovery suits.

Ars Technica by Ashley Belanger, Sep 22, 2026, on British Columbia’s complaint after the February 2026 Tumbler Ridge shooting; prior family suits and WSJ whistleblower reporting on OpenAI’s internal handling are referenced in the piece.