News & Analysis

The Tumbler Ridge Lawsuits: Families, Survivors, and British Columbia Sue OpenAI

By David Meldofsky

Published June 12, 2026 · Updated October 5, 2026

On February 10, 2026, an 18-year-old killed eight people in Tumbler Ridge, British Columbia, most of them at the town's secondary school. In April, families of the dead and wounded filed seven lawsuits against OpenAI and CEO Sam Altman. On September 2, a second wave began in San Francisco federal court. It was brought by students, teachers, and a principal who were inside the school. Plaintiffs' counsel has said that wave will reach thirty complaints, one per plaintiff.

On September 21, the Province of British Columbia and the local school district filed their own suit to recover what the response has cost. The September filings also make a new claim about what happened inside OpenAI. They say the company's own investigators wanted the shooter's account reported to Canadian police, and its global affairs team overruled them. OpenAI says that account is false.

What Happened

The shooting took place on February 10, 2026 at Tumbler Ridge Secondary School in northeastern British Columbia. According to authorities as reported by NPR, 18-year-old Jesse Van Rootselaar killed five students and a teacher at the school, then died by suicide. Earlier that day, the shooter had killed a mother and an 11-year-old half-brother at home. Eight people were killed in all. Around two dozen were injured.

Eight months before the attack, in June 2025, OpenAI's automated systems flagged Van Rootselaar's ChatGPT account for gun violence activity and planning. The account was deactivated. Van Rootselaar then opened a second account and kept using ChatGPT. OpenAI has said it did not learn of the second account until after the shooting.

Seven federal lawsuits followed on April 29, 2026, in the U.S. District Court for the Northern District of California. The second wave began in the same court on September 2, 2026. On September 21, the Province of British Columbia and School District No. 59, which ran the school, filed their own suit there. An earlier civil claim proceeds in the British Columbia Supreme Court.

What a Report Says ChatGPT Told the Shooter

On September 24, 2026, Mother Jones published the first public account of what Van Rootselaar wrote to ChatGPT and what ChatGPT wrote back. The reporter, Mark Follman, says he reviewed "significant portions" of the chat history and spoke with three sources who know the matter. Mother Jones has not named them. CBC and CTV have covered the report and say they have not seen the logs themselves.

The report centers on the second account. Van Rootselaar told ChatGPT that the first account had been banned for naming a famous real-world mall as the target of an attack. The shooter said the plan had not been real. ChatGPT explained why content like that gets flagged. It then advised framing future requests as fiction or as hypotheticals, so the account would "never get flagged again." The report quotes ChatGPT telling Van Rootselaar, "Don't use real-world locations," and "You can still be twisted. Just be clever about it."

With the requests framed that way, the report says, ChatGPT played out violent scenarios and discussed earlier mass shootings. It also described how a shotgun would perform at close range inside a school. We have left the most graphic passages off this page.

The province's complaint fits what the report describes. It says OpenAI's own written rules at the time treated violent and mass-casualty content as allowed when a user framed it as fiction or role-play. The report describes ChatGPT steering the shooter toward that same framing.

OpenAI had not answered requests for comment on the report as of September 25. The RCMP says it obtained the shooter's chat logs from OpenAI through a court production order filed about a month after the shooting.

B.C. Attorney General Niki Sharma said the province's lawyers asked OpenAI for the logs earlier this year and were refused. After reading the report, she asked the federal government to change the Criminal Code so AI companies can be held to account for what their systems do. Federal AI Minister Evan Solomon called the report horrifying and said he had not known about the specific conversations.

Every complaint on this page was filed before the report came out, and none quotes the logs. The complaints rest mostly on what OpenAI knew about the first account and chose not to report. The report adds a second line of evidence: what ChatGPT itself told the shooter after the ban. These are reported excerpts. No one has filed them in the lawsuits, and no court has weighed them.

The Lawsuits at a Glance

• Court: U.S. District Court for the Northern District of California, San Francisco Division.
• First wave: seven complaints filed April 29, 2026 by families of those killed and injured.
• Second wave: began September 2, 2026, brought by people who were inside the building during the attack, including students, teachers, and a principal. Counsel has said it will reach thirty complaints.
• Government suit: filed September 21, 2026 by the Province of British Columbia and School District No. 59 (Peace River South), No. 4:26-cv-10743. It seeks the public cost of the response and recovery, and it adds OpenAI Holdings, LLC as a defendant.
• Defendants: Sam Altman personally, along with OpenAI Foundation, OpenAI OpCo, LLC, and OpenAI Group PBC as successor to the earlier for-profit entities.
• Claims: negligence, negligent entrustment, negligent undertaking, aiding and abetting a mass shooting, product liability (treating ChatGPT as a defective product), and negligent infliction of emotional distress. Rushlow makes nine separate claims.
• Basis for federal jurisdiction: diversity under 28 U.S.C. section 1332(a)(2), because the plaintiffs are Canadian citizens and the defendants are Californian.
• Relief sought: damages, punitive damages, and injunctive relief to compel corporate changes at OpenAI. All request a jury trial.
• Counsel: Jay Edelson of Edelson PC is lead counsel. John Rice of Rice Parsons Leoni & Elliott LLP in Vancouver is Canadian co-counsel. The province and school district are represented by Lesley Weaver of Stranch, Jennings & Garvey in Oakland, with CFM Lawyers in Vancouver as Canadian counsel.
• Primary source: Rushlow v. Altman case guide on Lawsuit Center, with the complaint (PDF, 48 pages). The province's complaint (PDF, 39 pages) is hosted here.
• Related proceedings: a civil claim in the British Columbia Supreme Court.

The September 2026 Wave: Survivors Sue

The second-wave complaints come from people who were in the building and survived it. Global News reports the group includes educators, a principal, and students. One is a 13-year-old identified as A.C., who survived by playing dead among classmates in the library. Another is a 16-year-old identified as D.I., who hid in a storage closet that shares a wall with the library. Deidre Rushlow, a seventh-grade teacher who locked her classroom door and pulled students under her desk, is a named plaintiff.

Who is suing is what sets this wave apart. A parent suing over a child's death and a survivor suing over their own trauma have to prove different things. A survivor's emotional distress claim usually depends on being at the scene and seeing the harm as it happened. Those rules differ from place to place. Filing these cases together lets one court decide those questions on one record, instead of piecemeal over years.

Press accounts differ on one point worth flagging. TechCrunch reports that the September complaints accuse OpenAI of aiding and abetting the shooting for the first time. In that account, the claims go beyond saying OpenAI carelessly failed to prevent it. NPR's account of the April filings says those seven suits already claimed aiding and abetting, alongside the claim that OpenAI failed to notify police. Until the April counts can be read against the September ones on the face of the filings, treat the escalation as reported rather than settled.

Inside one complaint: Rushlow v. Altman

Rushlow v. Altman, No. 3:26-cv-09344-PHK, was filed September 2, 2026 and is the first of the second-wave complaints available in full.

Deidre Rushlow says she sent part of her grade 7 class to the library with an education assistant, minutes before the shooter reached it. She locked her room and hid the rest of her students under her desk. She was inside when the shooter fired through the front and back classroom doors. She called the library and texted the assistant during the attack and got no answer. The complaint says she texted her husband a goodbye.

The complaint makes nine separate legal claims. Most are versions of the same basic argument, that OpenAI was careless. It says the company was careless in not warning police or the public about how the product behaves. It also says OpenAI was careless in how it built the product in the first place.

Other claims say the product was defective, that OpenAI gave a known-dangerous user access again, and that the company ran a safety review and did it badly. The last claim covers the emotional harm to someone who was in the building during the attack.

One thing repeats throughout the complaint. Over and over it says OpenAI owed these duties whether ChatGPT counts as a product or not. Courts have not settled that question yet, and it affects which legal rules apply, so the lawyers wrote the case to work either way. Our explainer on whether AI output is a product or content covers why it matters so much.

One claim is worth watching more than the others. Normally, a company has to owe you something before it can be held responsible to you. OpenAI will argue it owed nothing to people in a Canadian town it had never heard of. This claim gets around that. It says OpenAI chose to review the account, and once it chose to, it had to do it carefully.

The complaint also argues that shutting the account down made things worse, because it showed the user what had set off the alarm.

The allegation about OpenAI's own employees

The strongest factual allegation in the second wave has nothing to do with what ChatGPT said to anyone. The complaints say threats against OpenAI's own staff go to a separate corporate security team that reports to the chief operating officer. That team can call police on its own authority.

The complaints point to November 21, 2025 as the example. On that day, the company locked down its offices, warned employees, circulated a suspect's name and photograph, and called the San Francisco Police Department. It did so while acknowledging there was no indication of active threat activity.

The plaintiffs set this against OpenAI's refusal to contact the RCMP about the flagged user. They say it shows that OpenAI's standards for urgency and privacy change depending on who is at risk. A court can test this claim against the company's own records without reading a single chat log. That makes the comparison useful to the plaintiffs.

British Columbia and the School District Sue

The province and the school district paid for much of what came after the shooting. On September 21, 2026, they sued OpenAI and Sam Altman in San Francisco federal court to get that money back. Attorney General Niki Sharma announced the case the next day. She called it "an important step toward seeking justice for the families, students, educators and community members whose lives were forever changed."

The complaint describes what the response required. The province declared a state of emergency the day after the attack. Within 48 hours, about 45 counselors, clinicians, victim-support workers, and school-safety officers were in town. By early April, child and youth mental health workers had held more than 900 counseling sessions with more than 100 children and families. Northern Health had treated more than 130 adults.

Tumbler Ridge Secondary never reopened. Students moved into temporary classrooms on the elementary school grounds, and demolition of the old building began in August 2026. The federal and provincial governments have each committed $100 million toward a new secondary school and an updated health centre.

The province and the district want OpenAI to repay those costs, plus punitive damages. They also ask the court to order these changes at OpenAI:

• End or redirect conversations in which a user talks about hurting other people.
• Have trained staff review conversations its systems flag as possible planning for violence.
• Report users who pose a credible threat of serious harm to police.
• Warn users that its safety features weaken over long conversations.
• Stop users banned for violent misuse from signing back up with the same identity, device, or payment method.
• Accept quarterly checks by an independent monitor.

What this complaint adds

The province says the RCMP already knew the shooter. Officers had visited the home over mental health concerns and had removed guns from it in the past. A call from OpenAI, the complaint argues, would have reached police who were in a position to act.

The complaint also says OpenAI has argued in court filings that the RCMP shares blame for not stopping the attack. The province reads that argument as an admission that police should have been told.

The shooter was 18, but adulthood in British Columbia starts at 19. OpenAI's teen protections applied only to users it identified as under 18, so the complaint says those protections never covered this user.

According to the complaint, roughly a dozen OpenAI employees reviewed the first account in June 2025. It also says Altman and OpenAI's Ann O'Leary promised changes after the shooting and have not made them, despite the attorney general's efforts to follow up. The province has not yet seen the shooter's chat logs. It has asked the court to order OpenAI to produce them quickly, and it plans to update its complaint once it has them.

The eight claims

Five of the eight claims say OpenAI was careless. The complaint says it was careless in not warning police and in not warning the public. It says the company was also careless in letting a flagged user back on, in how it ran its safety review, and in how it built the product. Two more claims say ChatGPT is a defective product, both in its design and in its missing warnings. The last accuses OpenAI of aiding and abetting a mass shooting.

The claim about warning police relies on a 1976 California Supreme Court case, Tarasoff v. Regents of the University of California. That case held that a therapist who learns a patient is a serious danger to someone must take reasonable steps to protect that person. The province argues OpenAI stood in a similar position. It ran the account, reviewed the flagged chats, and held the user's name and general location. Courts have not decided whether that rule reaches a technology company that reviews what its users write.

The province also faces a hurdle that applies to governments. Taxes pay for police, ambulances, and emergency services, so courts usually refuse to let a government bill a wrongdoer for them. The complaint answers this again and again. It calls the costs extraordinary and one-time, tied to a single event, and says they paid for new programs and buildings that exist only because of the attack.

What the complaints say OpenAI knew

The complaints say OpenAI's own systems flagged the account in June 2025 over what he was writing about guns, and that a specialist team inside the company reviewed it. That team decided he was a credible threat to real people, and more than one reviewer recommended calling the Royal Canadian Mounted Police.

They say that recommendation was overruled by the part of the company that handles public relations and politics. That team is led by chief global affairs officer Chris Lehane. The complaints say Sam Altman signed off. They make that link "on information and belief." Plaintiffs use that phrase for something they expect to prove once they see the company's records. It is also the part most exposed if the documents do not support it. Lehane is not a defendant.

OpenAI rejects this. Chief strategy officer Jason Kwon told NPR it is false to say Lehane was involved in the referral decision. He also denied that OpenAI's investigators report to Lehane. Kwon said the team reports to him. He said the people making these decisions prioritize safety, and denied that politics or public relations played any part.

The company gave its own account in February 2026. It said it identified the account, banned it for breaking the usage policy, and considered a referral. It decided not to refer, because it saw no imminent and credible risk of serious harm.

That is the heart of the dispute. The complaints say the company's own threat-assessment staff reached the opposite conclusion about the same conversations. If both accounts are true, the disagreement happened inside OpenAI, and the record of it is in the company's files. Canadian counsel John Rice has said the plaintiffs want every chat log from the first conversation through the day of the shooting. They also want the internal documents showing who spotted the threat, who recommended reporting it, and who decided not to.

The Rushlow complaint says OpenAI is withholding the logs from the people suing. The RCMP has them, and Mother Jones has described parts of them in the report covered above.

Altman's April 24 letter settles one part of it. He said he was deeply sorry the company did not alert law enforcement to the account.

Deactivating an account is not the same as banning someone

The complaints also say that after the first account was shut down, a second one was created and the conversations carried on.

The Mother Jones report says ChatGPT told the shooter, on that second account, how to avoid being flagged again. If the report is accurate, ChatGPT helped the shooter stay out of sight of OpenAI's own safety system.

That matters more than it sounds. A general claim that a company should have watched its users more closely is vague and easy to get dismissed. A claim that anyone could undo the shut-off by signing up again with a new email is specific. It points at how the system was built, not at anything the product said.

The September complaints go further. They say OpenAI has no way to ban a person at all, only to close one email address. They also say its own help pages tell deactivated users how to come back.

The same point matters for anyone holding an account now. Our guide on what to do in the first days covers what stays, what goes, and how long a deleted conversation can still be recovered.

Why this theory is different from the other cases

Every case against OpenAI runs into the same first question: is what a chatbot says a product, or is it speech? That question decides which rules apply, and our guide to whether AI output is a product or content explains it in plain terms.

These cases are built to work either way. Most of the claims ask whether OpenAI's actual safety practices matched what the company said they were. OpenAI's own records can answer that, whichever way a court classifies the output. In a set of cases where the classification fight comes first every time, a claim OpenAI's own records can settle is worth a great deal.

The Mother Jones report adds evidence about the product itself. It describes ChatGPT telling the shooter how to get past its own safety filters. Those passages go to the claims that ChatGPT was badly designed, and the province has said it will amend its complaint once it has the logs.

Why These Cases Are in Federal Court

There are two separate groups of ChatGPT lawsuits in California, in two different court systems. Which court a case goes to depends mostly on where the people suing live.

The Tumbler Ridge families and survivors are Canadian. When someone from another country sues an American company, federal court is open to them. That is why these cases were filed in federal court in San Francisco, the city where OpenAI is based.

Most of the other ChatGPT lawsuits were brought by Californians against a California company. That combination keeps a case in state court. About a dozen of them have been grouped together there in front of a single judge, which we cover in our guide to JCCP 5431.

The claims are different too. The state cases were brought by people who used ChatGPT themselves, or by their families. The Tumbler Ridge plaintiffs never used it. Their case is about what OpenAI knew about someone else and what it did with that information.

What this means in practice is that the two groups move separately. They have different judges, and a decision in one group does not settle the other. The same basic questions about whether OpenAI can be held responsible may end up answered twice, and the answers may not match.

The Canadian Dimension

The shooting occurred in Canada, the U.S. lawsuits were filed in California, and an earlier civil claim proceeds in the British Columbia Supreme Court. Cross-border litigation adds questions about jurisdiction, applicable law, and coordination between proceedings. The U.S. filings against a U.S. defendant in its home district are the more direct path to OpenAI's internal records. The legal team reflects that structure, with Edelson in Chicago and Rice in Canada. The province's team follows the same pattern, with Stranch, Jennings & Garvey in Oakland and CFM Lawyers in Vancouver.

What happens next

Case numbers. The September complaints were filed on the morning of September 2 and court records take time to appear. Case numbers, judge assignments, and whether the new cases get tied to the April seven are the first things to confirm.

Motions to dismiss. OpenAI filed them on September 18, 2026, covering the April and September cases. Its main argument is that the cases belong in British Columbia, because the witnesses, the evidence, and the public interest are there. If the judge agrees, the families and survivors would have to start over in Canada. If the cases stay, watch whether the safety-operations claims survive better than the design claims.

The aiding-and-abetting count. It requires proving the company intended to help, which is a much higher bar than carelessness. It is the claim most likely to be tested first.

The emotional distress claims. Whether people who were present but not shot can recover for what they went through is an open question, and it decides how far the second wave reaches.

The government case. The province filed its suit separately from the families and survivors. Whether it goes to the same judge, and whether a government can recover its response costs from a private company, are the questions to watch there.

The chat logs. Watch whether OpenAI answers the report, and whether the province amends its complaint once it gets the logs.

Discovery into the referral decision. Internal messages, escalation memos, and the threat-detection logs are the evidence the whole chain-of-command claim depends on.

Common Questions About the Tumbler Ridge Lawsuits

How many Tumbler Ridge lawsuits have been filed against OpenAI?

Thirty-seven. Seven were filed on April 29, 2026 by families of those killed and injured. A second wave began on September 2, 2026, brought by people who were inside the school during the attack. Counsel has said it will total thirty separate cases. Press accounts differ on how many have been filed so far. All name OpenAI and Sam Altman, and all request a jury trial. Separately, the Province of British Columbia and its school district filed their own suit on September 21, 2026.

What did the Tumbler Ridge shooter's ChatGPT chats show?

On September 24, 2026, Mother Jones reported that after OpenAI banned the shooter's first account, ChatGPT told the shooter on a second account how to avoid being flagged again. It advised framing requests as fiction or hypotheticals, and then played out violent scenarios framed that way. The report rests on three unnamed sources and parts of the chat history. OpenAI had not commented as of September 25, and the RCMP says it holds the logs.

Is British Columbia suing OpenAI over Tumbler Ridge?

Yes. On September 21, 2026, the Province of British Columbia and School District No. 59 (Peace River South) sued OpenAI and Sam Altman in federal court in California. They want OpenAI to repay the cost of the emergency response, counseling, and a replacement school. They also ask the court to order changes to how OpenAI handles users who threaten violence.

Who are the plaintiffs in the September 2026 Tumbler Ridge lawsuits?

Global News reports they are educators, a principal, and students who were present during the attack, including two minors identified by initials and a seventh-grade teacher, Deidre Rushlow. Their claims center on the harm of surviving the event rather than on the death of a family member.

Is Chris Lehane a defendant in the Tumbler Ridge lawsuits?

No. The complaints say OpenAI's global affairs team, which Lehane leads, overruled a recommendation to alert Canadian police. The Rushlow complaint alleges, on information and belief, that Lehane or someone under him made the decision and Sam Altman approved it. NPR reports that Lehane is not named as a defendant. OpenAI's chief strategy officer has said the allegation about his involvement is false.

Why were the lawsuits filed in California if the shooting happened in Canada?

OpenAI is headquartered in San Francisco, making the Northern District of California its home federal district. Suing a U.S. company in its home district is the most direct route to its records and avoids cross-border enforcement complications.

How are the Tumbler Ridge cases different from Raine v. OpenAI?

Raine centers on design choices in the model itself. The Tumbler Ridge suits center on safety operations: systems allegedly flagged the danger, investigators allegedly recommended a police referral, and the company allegedly did not make one. That theory does not depend on whether the output is classified as a product.

What is Rushlow v. Altman?

One of the second-wave complaints, No. 3:26-cv-09344-PHK, filed September 2, 2026 by Deidre Rushlow, a grade 7 teacher who was inside the school during the attack. It makes nine claims against Sam Altman and three OpenAI companies. The full complaint is available as a PDF on Lawsuit Center.

Why do the Tumbler Ridge complaints name Sam Altman personally?

They allege he directed the compressed safety testing before the May 2024 launch of the model at issue, and that he ratified the decision against a law enforcement referral. Naming an officer personally requires alleging his own conduct rather than the company's, which is what those paragraphs are built to do. OpenAI disputes the account of how that decision was made.

Why are the Tumbler Ridge cases in federal court when other ChatGPT cases are in state court?

Because the people suing are Canadian. When someone from another country sues an American company, federal court is available to them. The other California ChatGPT cases were brought by Californians against a California company, which keeps those cases in state court. Those are grouped together in JCCP 5431.

Are more Tumbler Ridge lawsuits expected?

The cases have grown from seven claims in April to a second wave expected to reach thirty. Other AI wrongful death cases keep being filed too. Our OpenAI Lawsuits hub tracks new filings.

Sources and further reading

• Lawsuit Informer: OpenAI School Shooting Lawsuits and AI Product Liability
• Lawsuit Informer: OpenAI Lawsuits Hub and Case Tracker
• Lawsuit Informer: Is AI Output a Product or Content?
• Mother Jones: ChatGPT Helped Tumbler Ridge Shooter Focus on Guns, Tactics, and Terror
• CBC News: B.C. attorney general shocked by news report that ChatGPT told Tumbler Ridge shooter how to avoid safeguards
• The Globe and Mail: B.C. Attorney-General, federal AI Minister alarmed by report ChatGPT coached Tumbler Ridge shooter
• Complaint: His Majesty the King in Right of the Province of British Columbia v. Altman (PDF)
• CBC News: B.C. government suing OpenAI for alleged role in Tumbler Ridge mass shooting
• Al Jazeera: Canada's BC sues OpenAI over ChatGPT role in Tumbler Ridge school shooting
• The Canadian Press: OpenAI asks U.S. court to dismiss Tumbler Ridge mass shooting lawsuits
• NPR: New lawsuits claim OpenAI execs put image ahead of safety in Canadian mass shooting
• Global News: Tumbler Ridge shooting survivors among 30 new plaintiffs suing OpenAI
• TechCrunch: OpenAI faces 30 more lawsuits tied to Tumbler Ridge shooting
• Reuters: Families of Canadian mass shooting victims sue OpenAI, CEO Altman in U.S. court
• Wall Street Journal: OpenAI sued by seven families over mass shooting suspect's ChatGPT use
• Associated Press: Coverage of the Tumbler Ridge lawsuits

Educational commentary only. Not legal advice. No attorney-client relationship is created.