Skip to main content
NOPE

This case involves a death, suicide or self-harm. If you or someone you know is struggling, free and confidential support is available. Find a helpline near you.

Critical reported severity Involving minors Lawsuit Ongoing

Tumbler Ridge shooting: ChatGPT use and OpenAI non-reporting

Eight victims were killed in Tumbler Ridge, B.C., on 10 February 2026, and the shooter died by suicide. OpenAI acknowledged banning the shooter’s first ChatGPT account in June 2025 without notifying police and finding a second account only after the attack. Families and survivors filed California lawsuits, followed by B.C. and its school board on 21 September. Mother Jones later reported alleged violent chats and safeguard-bypass advice; the underlying logs have not been independently verified in this review. OpenAI disputes the liability allegations. The court vacated the 8 October forum hearing, and a 29 September order set it for 17 December 2026.

AI system
ChatGPT
OpenAI
Occurred
10 Feb 2026
Reported
20 February 2026
Event location
Canada
What the AI did
Communicated with the person · Relation unknown
Reported harm
DeathPhysical InjuryPsychological DistressHarm to Others
Whose AI use
Their own AI use · Someone else’s AI use
Setting
Education · Family · Justice · Public services
Evidence
AI involvement supported · Causal attribution disputed · 31 sources, 7 underlying accounts
35 claims: 19 documented, 1 corroborated, 3 disputed, 12 reported. 8 open questions
People reported harmed
At least 11 people

What Happened

The 10 February 2026 shooting in Tumbler Ridge, British Columbia, killed eight victims: the shooter’s mother and younger half-brother at home, and five students and an education assistant at the secondary school. The 18-year-old shooter died by suicide. Survivors suffered physical injuries and reported lasting psychological harm. Names and graphic descriptions are omitted where they are unnecessary to understand the event.

OpenAI’s own 26 February letter confirms prior ChatGPT use, automated detection and human review of the first account, its June 2025 closure without a police referral, and the later discovery of a second account. The company said the first account did not meet its then-current threshold for an imminent, credible threat. The reported AI connection is independently supported by Mother Jones’s later investigation; neither route establishes causation.

The complaints allege that reviewers recommended contacting police, that senior leaders overrode them for reputational reasons, and that continued access and unsafe design contributed to the attack. OpenAI disputes responsibility and the allegations about its leadership. Its motion says Altman was not personally involved in the June review or referral decision. Kwon denied Lehane’s involvement and said the investigations and legal teams report to him.

The 24 September Mother Jones article adds allegations about the content of the second account, including violent scenarios and advice to present requests as fiction to evade safeguards. It reports use until the day of the attack and describes three anonymous sources plus material reviewed by the reporter. This review did not obtain or authenticate the underlying logs. Global News’s Canadian Press follow-up questions the sourcing, while CBC explicitly disclaims independent verification. Their repetition is not corroboration.

The inspected survivor filings preserve accounts of a head teacher attempting resuscitation, F.D. sheltering, a 19-year-old student’s serious injury and hospital transfer, her sister sheltering, a gym teacher coming under fire and A.C. surviving in the library. The four complaints share a litigation template and are not four independent sources for OpenAI’s internal conduct. They allege serious psychological consequences as well as physical injury.

The provincial suit is a separate legal action within this same incident dossier. The Province and school board allege major costs and educational disruption, and seek compensation and changes to safety controls. Court orders now coordinate the proceedings. After vacating 8 October and setting 5 November, the court on 29 September granted a joint stipulation that set the forum hearing for 17 December 2026 and a Rule 12(b)(6) hearing for 11 February 2027.

The survivor complaints describe several days of violence-related conversations on the first account. They allege that a referral would have reached police already familiar with the shooter through earlier contacts and firearm removal, and invoke an alleged May 2026 confidential IPO submission as part of their corporate-motive theory. These content, prevention and motive allegations are not findings.

AI Behaviors Exhibited

OpenAI acknowledges first-account detection, human review and closure, no police referral at that time, and second-account use discovered only after the attack. Mother Jones reports violent output and safeguard-bypass advice, but the underlying logs were not independently inspected here. The complaints allege further encouragement and design failures; those claims remain contested. OpenAI promised changes to referral criteria and repeat-violator detection.

Reported harm

Eight victims were killed, the shooter died by suicide, and survivors suffered physical injuries and reported psychological trauma. The shooter’s prior ChatGPT use is supported; claims that ChatGPT or OpenAI’s non-reporting contributed to the deaths and injuries remain contested.

How Harm Occurred

Plaintiffs allege three connected mechanisms: model responses that reinforced violent ideation, continued access after the first account was closed, and an organizational decision not to warn police despite internal concerns. The first account’s closure and absence of referral are acknowledged by OpenAI; the detailed content, internal decision-makers, legal duty and causal contribution remain disputed or unresolved. A warning might have enabled intervention, but this counterfactual has not been established. The deaths and injuries are recorded separately from AI causation.

Outcome

Ongoing

10 February 2026: eight victims killed at a home and Tumbler Ridge Secondary School; the shooter also died. At least two survivors needed hospital care and further injuries and trauma were reported. The eight-victim toll excludes the shooter.

20–26 February: OpenAI acknowledged the June 2025 ban without a law-enforcement referral, citing its then-current imminence threshold. Its 26 February letter disclosed the second account, discovered after public identification of the shooter and then shared with police, and promised changes to reporting and repeat-violator detection.

Early March: Canadian officials sought answers and Altman met with the premier and mayor. A coroner’s inquest was announced. On 9 March the family of M.G. filed in B.C. Supreme Court. Altman’s apology was dated 23 April and reported publicly on 24 April.

29 April: seven family suits were filed in the Northern District of California against OpenAI and Altman, replacing the earlier B.C. family claim. The BBC attributed a prospective request exceeding US$1 billion to M.G.’s case; it was not an award or an aggregate verified loss.

7 July: the Province retained CFM Lawyers and Stranch, Jennings & Garvey to explore its own legal action. The RCMP said its criminal investigation remained active and separate from that initiative.

2 September: thirty survivor suits, including students, teachers and the principal, brought the private total to 37. Inspected complaints include Deeley (26-cv-09378), Hoekstra (26-cv-09377), Bertrand (26-cv-09346) and A.C. (26-cv-09342). They share a nine-count template and allege negligent design and warnings, negligent entrustment and undertaking, aiding and abetting, strict liability and emotional distress. Internal reporting-line and IPO-motive allegations remain contested. Kwon denied Lehane’s involvement; OpenAI denied Altman personally made the June referral decision. The defendants also filed forum non conveniens motions in the seven April cases that day. Younge (26-cv-03707) received an amended complaint, entry 30, that day.

9–17 September: the thirty survivor cases were related to Stacey. The court vacated the 8 October hearing on 10 September. Its signed 15 September order set the forum hearing for 5 November at 10 a.m., opposition for 6 October and reply for 20 October. It set a separate Rule 12(b)(6) hearing for 10 December and case management for 16 December. A.C.’s 17 September forum motion confirms the November date. The thirty survivor suits include a grade 7 teacher’s action (Rushlow, 26-cv-09344), whose docket records the same relation, reassignment and forum motion.

21 September: B.C. and the Board of Education of School District No. 59 filed case 26-cv-10743 against Altman and four OpenAI entities. The 39-page complaint seeks response and recovery costs, including a replacement school and wellness centre, and safeguards and reporting changes. Its eight causes of action and prevention theories remain allegations. The attorney general also advocated Criminal Code changes; this was not a criminal prosecution.

24–28 September: Mother Jones published alleged second-account content and safeguard-bypass advice. It describes anonymous sources and reviewed chat-history material. Global News/Canadian Press and CBC relay the reporting without establishing independent authentication of the logs. On 25 September the docket indexes an order relating the provincial suit to Stacey and reassignment to Judge Corley.

29 September: the court granted the parties’ joint stipulation in Stacey and all related cases. Forum non conveniens motions are due 15 October, oppositions 5 November and replies 3 December, with the hearing on 17 December 2026 at 10 a.m. Amended complaints are due 29 October, and a Rule 12(b)(6) hearing is set for 11 February 2027. An initial case management conference on alternative dispute resolution was set for 20 October 2026 by Zoom.

As checked 15 October 2026: the Rushlow docket lists a stipulation and proposed order on 13 October and a protective order on 14 October; neither body was read. No settlement, liability judgment or completed forum hearing was identified in the inspected record. Public RECAP coverage is incomplete and does not establish the absence of other filings.

What remains unknown

  • The complete authenticated chat histories, their model/version context and whether the published excerpts represent the whole exchanges have not been independently inspected.
  • The exact internal referral decision-makers, escalation evidence and disputed reporting hierarchy have not been established.
  • The technical reasons the second account was not detected and the accounts’ identity linkage remain unresolved: the provincial complaint alleges shared identity information, whereas Mother Jones reports different account names and email addresses. Neither assertion is independently verified here.
  • Whether model outputs, access restrictions or a police referral would have changed the outcome remains unresolved; no causal or liability finding is asserted.
  • A complete deduplicated count of physically injured and psychologically harmed people is unavailable. The numeric minimum deliberately excludes additional unquantified harm.
  • Implementation and effectiveness of OpenAI’s promised safeguards are not established by the commitments.
  • RECAP is not a complete live docket. The 28 September stipulation body was not read, although the 29 September order granting it was. Whether forum motions were filed by the 15 October deadline, and the contents of the 13 October stipulation and 14 October protective order in the Rushlow action, are not established.
  • No inquest date, final investigative outcome, settlement or liability judgment was established by the inspected sources.

What the evidence supports

AI involvement: supported. The shooter’s own use is established by OpenAI’s first-person account disclosure and Mother Jones’s distinct original reporting. communicated_with applies to the shooter, counted among those who died. unknown applies to other harmed people: inspected evidence does not establish ChatGPT communicating with, acting for, deciding about, claiming about or depicting them. The involvement finding does not establish facilitation or causation.

35 claims: 19 documented, 1 corroborated, 3 disputed, 12 reported. What the statuses mean

Documented On 10 February 2026, eight victims were killed in Tumbler Ridge, British Columbia; the shooter also died. The eight-victim count excludes the shooter.

Causal attribution. No causal conclusion about AI follows from this claim.

  • rcmp.ca(opens in new tab) supports · English
    Opening investigation paragraph and paragraph on autopsies for eight victims and the suspect.
  • cbc.ca(opens in new tab) supports · English
    Paragraph identifying eight people killed on 10 February before the shooter killed herself; victim-identification caption.
  • motherjones.com(opens in new tab) supports · English
    Paragraph beginning “Van Rootselaar carried out” and subsequent account of the deaths at home and school.
Reported The victims were the shooter’s mother and 11-year-old half-brother at home, and five students and an education assistant at the secondary school. The 18-year-old shooter died by suicide.

Causal attribution. No causal conclusion about AI follows from this claim.

Reported At least two surviving victims required hospital care. Other injuries were reported; the provincial complaint distinguishes two students transported to Vancouver from approximately 25 people seeking local medical care. These descriptions do not establish an exact deduplicated injury total.

Causal attribution. No causal conclusion about AI follows from this claim.

Reported Survivor complaints report lasting psychological harm among students and educators, including a head teacher and his son F.D., a gym teacher, a grade 7 teacher, a student and her sister, and A.C. These are individual accounts of harm, not a count of everyone present at school.

Causal attribution. The complaints attribute these harms to the shooting and allege a contribution from OpenAI; that contribution is contested.

Corroborated The shooter used ChatGPT before the shooting. OpenAI acknowledged an account belonging to the perpetrator, while Mother Jones independently reported reviewing material from the shooter’s ChatGPT history.

Causal attribution. These distinct evidentiary routes establish the reported AI connection, not that ChatGPT caused or determined the attack.

Documented OpenAI acknowledged that automated detection led to human review and closure of the shooter’s first ChatGPT account in June 2025, without a police referral. It said the visible activity did not meet its then-current threshold for credible and imminent planning.

Causal attribution. The company’s acknowledgement establishes its stated decision and rationale, not whether that decision satisfied a legal duty.

Documented OpenAI acknowledged that the shooter used a second ChatGPT account despite its repeat-violator detection system. It said it discovered that account after the shooter was publicly identified and then shared it with law enforcement.

Causal attribution. Continued access is acknowledged; its causal significance and whether different enforcement would have prevented the attack remain unresolved.

Documented In its 26 February letter to Canadian ministers, OpenAI committed to stronger referral criteria, direct Canadian law-enforcement contacts, localized support and improved repeat-violator detection. It said its enhanced protocol would refer the June 2025 account if discovered then.

Causal attribution. A policy commitment is not proof of implementation or effectiveness.

Reported Sam Altman’s apology letter, dated 23 April and publicly reported on 24 April, expressed regret that OpenAI had not alerted law enforcement. CBC reported that an OpenAI spokesperson confirmed the letter’s authenticity.

Causal attribution. Regret and acknowledgement of non-reporting do not establish legal liability or causation.

  • cbc.ca(opens in new tab) supports · English
    Opening paragraphs on publication by Tumbler RidgeLines, company authentication and date; apology passage.
Reported On 9 March 2026, the family of the critically injured student identified as M.G. in the U.S. proceedings filed a B.C. Supreme Court claim, including her younger sister and mother. They alleged serious physical and psychological harms and prior knowledge of violent planning by OpenAI.

Causal attribution. Claims of company knowledge and causal responsibility remain allegations.

Documented Seven family actions were filed in Northern District of California federal court on 29 April 2026 against Sam Altman and OpenAI entities. They replaced the earlier B.C. family action, whose discontinuance is described in the defence motion.

Causal attribution. This records procedural history, not the truth of the complaints.

Reported On 29 April, the BBC reported counsel’s prospective request for more than US$1 billion in damages in M.G.’s case. This was a requested amount, not an award or an established aggregate loss.

Causal attribution. No causal conclusion about AI follows from this claim.

  • bbc.com(opens in new tab) supports · English
    BBC damages paragraph specifying the prospective amount for this individual case.
Documented Thirty additional survivor actions were filed on 2 September 2026, including students, teachers and the principal. The court related them to Stacey v. Altman on 9 September; with the seven April actions, the private litigation comprised 37 cases.

Causal attribution. Case counts are not people counts, and related cases are not independent corroboration of shared allegations.

Reported The inspected September complaints include a head teacher’s account of attempting to resuscitate a student and his son F.D.’s account of sheltering; a 19-year-old student’s serious injury and hospital transport and her sister’s experience of sheltering; a gym teacher’s account of coming under fire; and 13-year-old A.C.’s account of surviving in the library.

Causal attribution. The experiences and harms are pleaded accounts. Their existence does not establish the complaints’ causal allegations against OpenAI.

Disputed The survivor complaints allege negligent design and warnings, negligent entrustment and undertaking, aiding and abetting, product liability and emotional distress. They allege staff urged a police referral and that reputation and IPO interests influenced leadership’s refusal. OpenAI disputes these allegations.

Causal attribution. These are contested legal and causal theories, not findings that OpenAI facilitated or could have prevented the deaths.

Disputed The September complaints allege Chris Lehane or someone in his reporting chain made the non-referral decision and Altman ratified it. Jason Kwon denied Lehane’s involvement and said the investigations and legal teams report to him. OpenAI’s court motion also denied Altman’s personal involvement in the June review or referral decision.

Causal attribution. Responsibility for the internal decision remains disputed; the allegation is not independently corroborated by matching complaint templates.

Documented On 7 July, B.C. announced retention of CFM Lawyers and Stranch, Jennings & Garvey to explore legal action against OpenAI and its decision-makers. That announcement described preparation for a separate provincial action.

Causal attribution. No causal conclusion about AI follows from this claim.

Documented On 21 September, the Province of British Columbia and the Board of Education of School District No. 59 (Peace River South) filed a federal complaint against Sam Altman and four OpenAI entities in the Northern District of California, case 26-cv-10743. The complaint pleads eight causes of action.

Causal attribution. The filing is documented; the pleaded allegations are not established facts.

Documented The Province and school board seek recovery of extraordinary response and recovery costs, including health care, victim services, educational continuity, demolition and a replacement school and wellness centre. They also request punitive and other relief, changes to safeguards and reporting, durable account enforcement and independent compliance audits.

Causal attribution. This establishes relief requested, not a damages award or proof that the defendants caused those expenditures.

Disputed The provincial complaint reports lasting trauma, school closure and demolition, replacement education arrangements and substantial public recovery expenditures. It alleges ChatGPT reinforced violent ideation, that a police warning would have prevented the attack, and that continued access contributed to it.

Causal attribution. The public bodies’ reported losses and their attribution of them to OpenAI are distinct. Causation and liability remain contested.

Documented On 2 September, OpenAI and Altman moved to dismiss the seven original California actions for forum non conveniens, arguing British Columbia was the appropriate forum. They disputed liability and pointed to Canadian institutions and witnesses as relevant to causation and comparative fault.

Causal attribution. This is the defendants’ litigation position, not a ruling that other institutions caused the shooting.

Documented The court’s 10 September order vacated the 8 October 2026 hearing in the seven earlier actions, pending a coordinated schedule.

Causal attribution. No causal conclusion about AI follows from this claim.

Documented The signed 15 September scheduling order had set the forum non conveniens hearing for 5 November 2026 at 10 a.m. in San Francisco, with opposition due 6 October and replies due 20 October. It also set a separate Rule 12(b)(6) hearing for 10 December and case management for 16 December.

Causal attribution. These were scheduled proceedings, not completed hearings or merits decisions. A 29 September order later set a different schedule.

Documented A.C.’s docketed 17 September motion extends the forum non conveniens request to a survivor action and lists the 5 November hearing. The 15 September order required forum motions in the thirty additional actions by 17 September.

Causal attribution. No causal conclusion about AI follows from this claim.

Documented The public docket indexes a 25 September order relating the provincial action to Stacey v. Altman and reassignment to Judge Jacqueline Scott Corley. The Province’s cost-recovery claims remain a separate action arising from the same shooting.

Causal attribution. Relation and reassignment do not decide the merits or create a separate incident.

Documented The RECAP indexes inspected on 29 September list a 28 September stipulation and proposed order in Stacey (entry 45), Schofield (entry 35), and the provincial action (entry 10); the provincial index also lists a 29 September attorney notice (indexed as appearance/substitution/change/withdrawal). The stipulation’s own body was not read; the court’s 29 September order granting it was read.

Causal attribution. An indexed proposed order is not a granted order. The schedule the court granted is stated in the 29 September order.

Reported A grade 7 teacher’s complaint alleges that she had sent part of her class to the library, where children and an education assistant were then killed, and that she locked her classroom and hid her remaining students under her desk while the shooter fired shotgun rounds through its doors. It reports severe and lasting psychological trauma.

Causal attribution. The complaint attributes her trauma to the shooting and alleges a contribution from OpenAI; that contribution is contested and no court has found it.

  • storage.courtlistener.com(opens in new tab) supports · English
    Paragraph 2, PDF p.2: 'she locked the door and hid her remaining students under her desk'; paragraph 28, PDF p.9: 'fired several shotgun blasts through the classroom doors that struck the wall behind their heads'; paragraph 201, PDF p.46: 'acute psychological trauma, persistent fear, sleep disturbances, intrusive recollections'.
Documented The teacher’s action, Rushlow v. Altman (26-cv-09344), was filed on 2 September 2026, related to Stacey v. Altman on 9 September and reassigned to Judge Jacqueline Scott Corley on 10 September. OpenAI and Altman filed a forum non conveniens motion in it on 17 September.

Causal attribution. Docket entries establish filings and procedural orders, not the truth of the complaint’s allegations.

  • courtlistener.com(opens in new tab) supports · English
    Docket header: 'Date Filed: Sept. 2, 2026'; entry 10: 'ORDER RELATING CASE. Signed by Judge Jacqueline Scott Corley on 9/9/2026'; entry 11: 'IT IS ORDERED that this case is directly assigned to Judge Jacqueline Scott Corley'; entry 21: 'MOTION to Dismiss for Forum Non Conveniens filed by Samuel Altman'.
Documented On 29 September 2026 the court granted the parties’ joint stipulation in Stacey v. Altman and all related cases. Forum non conveniens motions are due 15 October, oppositions 5 November and replies 3 December, with a hearing on 17 December 2026. Amended complaints are due 29 October, and a Rule 12(b)(6) hearing is set for 11 February 2027. A case management conference on alternative dispute resolution was set for 20 October 2026.

Causal attribution. These are scheduled proceedings, not hearings held or merits decisions.

  • storage.courtlistener.com(opens in new tab) supports · English
    Order of 29 September 2026, pp.1–2: 'The Court grants the stipulation as follows'; 'Deadline to File Motion: October 15, 2026'; 'Hearing: December 17, 2026 at 10:00 a.m.'; 'Amended Complaints Deadline: October 29, 2026'; 'Hearing: February 11, 2027 at 10:00 a.m.'; 'The Court sets an initial case management conference for October 20, 2026'.
Reported Mother Jones’s 24 September investigation alleges that ChatGPT supplied violent scenario content, including a classroom scenario, and advice to frame requests as fiction to bypass safeguards on the shooter’s second account. It reports continued use until the day of the attack.

Causal attribution. Reported allegations based on the journalist’s sources and reviewed material. This review has not authenticated the logs or established that any output caused the attack; operational details are omitted.

  • motherjones.com(opens in new tab) supports · English
    Opening second-account account; August 2025 conversation paragraphs; paragraph beginning “Van Rootselaar continued using ChatGPT until the day”.
Reported Mother Jones describes three anonymous sources with knowledge and chat-history material reviewed by its reporter. Global News’s Canadian Press account relays the allegations and says sourcing is not described; CBC’s Front Burner description says CBC has not independently verified the reporting. These follow-ups do not independently corroborate the alleged chat contents.

Causal attribution. The currently inspected original describes source categories but leaves provenance and authentication details unresolved. No inference is made from OpenAI’s non-response to Mother Jones.

Documented B.C.’s attorney general said on 21 September she had written to federal ministers proposing Criminal Code changes concerning human accountability for AI. The RCMP criminal investigation and announced coroner’s inquest are separate from the provincial civil lawsuit.

Causal attribution. Policy advocacy, an investigation and an inquest do not establish criminal liability or any criminal charge against OpenAI.

Reported After federal officials sought answers, Altman met virtually with Premier David Eby and Tumbler Ridge’s mayor in early March and agreed to apologize and work on regulatory recommendations, according to CBC.

Causal attribution. No causal conclusion about AI follows from this claim.

Reported The survivor complaints describe violence-related first-account conversations spanning several days in June 2025 and allege that a referral would have reached police who already knew the shooter through prior contacts and firearm removal. The Deeley complaint invokes an alleged confidential IPO submission on 22 May 2026 as part of its reputational-motive theory.

Causal attribution. The prior-warning, prevention and corporate-motive interpretations are plaintiffs’ allegations. They do not establish that a referral would have prevented the attack; OpenAI disputes liability and reputational-motive claims.

Documented Younge v. Altman, case 26-cv-03707, received an amended complaint on 2 September 2026 (entry 30).

Causal attribution. Procedural metadata only; this review does not characterize the amendment’s contents.

Sources

31 sources inspected, from 7 underlying accounts. Sources that repeat one account do not corroborate each other.

How the sources were read, and where the events happened

English source body read directly; no translation. Applies to s_rcmp, s_openai, s_rcmp_july, s_vacate, s_ac_motion.

English body read. CBC obtained direct responses from OpenAI and RCMP; the original discovery is credited to WSJ. Company-disclosure claims are grouped with OpenAI, not counted again as independent. Applies to s_cbc_ban.

English article read in full, including sourcing and disclosure. Reporter says he reviewed chat-history material and spoke with three anonymous sources; we did not inspect or authenticate the underlying logs. One reporting chain, not three independent sources. Applies to s_mj.

English Canadian Press article hosted by Global News read. Alleged chat contents are relayed from Mother Jones, not independently corroborated; its statement that sourcing is undescribed differs from the currently inspected original. Applies to s_global.

English programme description read; audio not reviewed. CBC explicitly says it has not independently verified the Mother Jones reporting. Applies to s_cbc_limit.

English CBC body read; allegations and medical descriptions are attributed to the March complaint. Applies to s_march.

English CBC body read; CBC says OpenAI confirmed the apology letter's authenticity. Applies to s_apology.

English BBC body read; interviews with counsel and company response. Litigation claims are not independent corroboration of the complaints. Applies to s_bbc.

English NPR body read; includes direct emailed Kwon response. Shared plaintiffs' allegations are not independently corroborated by repetition. Applies to s_npr.

English filing; cited passages read. Its allegations are not findings. The four survivor complaints share counsel and a template. Applies to s_deeley, s_hoekstra, s_bertrand, s_ac.

English official archive release read; cosmetic character-encoding artifacts in headings do not affect cited passages. Applies to s_july.

English 39-page complaint read. Court filing establishes allegations and relief sought, not their truth. Paragraphs 7 and 41 expressly trace prior-warning allegations to WSJ reporting. Applies to s_bc.

English official archive statement read; cosmetic character-encoding artifacts in headings do not affect cited passages. Applies to s_bc_release.

English CBC body read, including direct OpenAI response and RCMP update. Not independent corroboration of the Province's allegations. Applies to s_cbc_bc.

English defence filing; procedural request and denial passages read, not findings. Applies to s_motion.

English signed order read. Hearing-date text has an extraction/font defect; the identical order on Schofield ECF 32 was rendered and visually checked. Applies to s_schedule.

English ECF document read. CourtListener classifies it as an Order Relating Case; the proposed-order label is struck through, and reassignment is indexed. No merits finding. Applies to s_relation.

Public RECAP API index inspected on 29 September 2026. This is not a complete live PACER docket; the 28 September stipulation body is unavailable. Applies to s_stacey_index, s_bc_index.

Latest public RECAP API entries inspected on 29 September 2026; 28 September entry 35 is indexed, but its body is unavailable. Not a complete live PACER docket. Applies to s_schofield_index.

Two pages of RECAP API results inspected. Entry 30 establishes an amended complaint was filed, not its allegations; complaint body not inspected. Applies to s_younge_index.

English filing (ECF 1, 48 pages), read in full on 2026-10-15. Its allegations are not findings. It shares counsel and a template with the other survivor complaints. Applies to s_rushlow.

CourtListener docket page read on 2026-10-15, last updated 14 October 2026. It is not a complete live PACER docket; most entry bodies are not on RECAP. Applies to s_rushlow_docket.

Signed two-page order of 29 September 2026 in Stacey v. Altman and all related cases, filed as Document 41 in 3:26-cv-03705 and docketed as entry 23 in the Rushlow action; read in full on 2026-10-15. Applies to s_reschedule.

Event countries: Canada. Affected-person countries: Canada. Court countries: Canada, United States.

The shooting and affected people are located in Tumbler Ridge, B.C., by RCMP and reporting. Both the earlier B.C. Supreme Court family action and the subsequent Northern District of California proceedings are recorded; the earlier action was discontinued. Countries describe locations and court proceedings, not nationality. This carries forward Canada and U.S. metadata and adds the historically documented Canadian court.

Reviewed for publication 2026-10-15: Full review of this previously published death incident. Official police evidence supports the deaths; CBC’s direct OpenAI/RCMP reporting and Mother Jones’s separate investigation each report the deaths and the AI connection, with distinct underlying support for the connection. OpenAI’s own letter supplies primary confirmation of account use and non-referral. Derivative coverage and shared complaint templates are grouped together, not counted as independent corroboration. Complaints and alleged logs remain attributed, company disputes are retained, and causation is unresolved. Minor identities, unnecessary victim names and operational violence details are minimized. The 15 October 2026 review adds a grade 7 teacher’s complaint and docket as attributed allegations and documented filings, and the signed 29 September 2026 order that reset the forum hearing to 17 December 2026.

Interaction Concerns

Third Party Harm FacilitationCrisis Response Failure

Contributing Factors

acknowledged non referralsecond account usecontested internal threat assessmentalleged unsafe model responsesalleged inadequate ban enforcement

People described

Eight homicide victims: six children and two adults, at a home and Tumbler Ridge Secondary School. At least two other students required hospital care; further injuries and psychological harms are reported. The 18-year-old shooter, who had used ChatGPT, also died. Most people harmed have no established direct relationship with the AI system. Minor plaintiffs are identified only by court initials when needed.

People reported harmed in this case

At least 11 people

1 AI participant · 10 other people harmed

Additional affected people are described without a reliable count.

Conservative retained lower bounds: one ChatGPT user who died, eight homicide victims, and two distinct surviving victims requiring hospital care (RCMP 13 February; OpenAI 26 February letter; c_deaths, c_injuries and c_ai_connection). The two hospital survivors are students, not both children: the inspected Hoekstra complaint gives one survivor’s age as 19. Additional injured people and people alleging psychological harm are not enumerated because the totals and overlap are unresolved. This counts people harmed in the event, not deaths or injuries proven to have been caused by AI.

Counted once within this case. The same person may appear in other cases. This count does not establish AI causation.

Tags

#mass-shooting#duty-to-warn#ban-evasion#chatgpt#openai#canada#bc#school-shooting#minor-victims#prior-warning#lawsuit#provincial-action#criminal

Cite this case

Compiled per our published methodology: verification statuses, sourcing standards, and corrections process.

APA

NOPE. (2026). Tumbler Ridge shooting: ChatGPT use and OpenAI non-reporting. AI incidents. https://nope.net/incidents/2026-tumbler-ridge-chatgpt-shooting

BibTeX

@misc{2026_tumbler_ridge_chatgpt_shooting,
  title = {Tumbler Ridge shooting: ChatGPT use and OpenAI non-reporting},
  author = {NOPE},
  year = {2026},
  howpublished = {AI incidents},
  url = {https://nope.net/incidents/2026-tumbler-ridge-chatgpt-shooting}
}

Related cases

Critical ChatGPT

Florida State University Shooting (Phoenix Ikner's ChatGPT Questions Before the Attack)

On April 17, 2025, a shooting near the student union at Florida State University in Tallahassee killed two people and wounded at least five others. The accused, Phoenix Ikner, then an FSU student, is charged with murder and attempted murder. ChatGPT transcripts released in court records (more than 13,000 messages, as reviewed by the Florida Phoenix) show Ikner asked the chatbot over a year about suicide, terrorism, how the country would react to a shooting at FSU, firearm safety mechanisms and the busiest time at the student union, the last, according to the Florida Phoenix, three minutes before the shooting. Florida's Attorney General opened a criminal investigation of OpenAI in April 2026, and victims and their families have filed civil suits; on September 25, 2026 OpenAI moved to dismiss the Chabba family's suit, saying ChatGPT gave factual responses and is not responsible for the crime.

Critical ChatGPT

Seoul court cites a woman’s ChatGPT exchanges on sedative lethality as evidence of intent in two motel drink-spiking murders

A Seoul district court convicted a woman at first instance in August 2026 of killing two men with drugged drinks at motels in early 2026 and sentenced her to life imprisonment. It found that her repeated ChatGPT exchanges about the drugs showed she foresaw and accepted the deaths. Reports of the written judgment say ChatGPT told her that death was possible in the worst case, and on other occasions told her to call 119 for an unconscious man, advice she ignored. Both sides have appealed. The court did not find that the chatbot caused the killings.

Critical ChatGPT (reported)

Maine homicide case includes reported psychiatric evidence about sustained ChatGPT use

Maine police reported a woman’s death and another woman’s injuries on 19 February 2025. Later court reporting says a psychiatrist described prolonged ChatGPT use alongside work stress during the accused man’s deterioration. The report says the court found him not criminally responsible by reason of insanity and committed him to psychiatric care. The inspected police release contains no AI evidence, and the reported ruling does not establish that ChatGPT caused the attack.

High ChatGPT

Zurich-Oerlikon, Switzerland: a student who researched killing methods, anatomy and where to find children with ChatGPT stabbed three five-year-olds walking to their daycare; sentenced on 18 September 2026 to 15 years for multiple attempted murder

At midday on 1 October 2024 a Chinese master's student, then 23, attacked a group of five-year-old kindergarten children walking with a daycare worker to their after-school Hort near Berninaplatz in Zurich-Oerlikon with a kitchen knife. Two boys suffered life-threatening neck and chest wounds (one had the jugular vein severed, one the pericardium and aorta cut); a third was wounded in the ear and the shoulder or lung. The daycare worker held the attacker's arms until he dropped the knife. The indictment and the court's oral reasoning, as reported by NZZ, Keystone-SDA/ATS and CH Media, describe weeks of preparation from summer 2024: he discussed ways of killing and their legal consequences with ChatGPT, studied anatomy, bought raw beef to practise stabbing, and used ChatGPT to find out where and when children could be encountered in Zurich. No source quotes the prompts or ChatGPT's replies. On 18 September 2026 the Zurich District Court convicted him of multiple attempted murder and imposed 15 years' imprisonment, treatment during custody, a 15-year expulsion from Switzerland and CHF 106,000 in moral damages to the three boys; it accepted that he was psychotic but found diminished, not absent, criminal responsibility. His lawyer announced an appeal. The parents of one boy described nightmares, self-blame and scars 23 months on.

If you or someone you know is struggling, free and confidential support is available. Find a helpline near you at Signpost.

Back to all incidents