Crisis resources available 24/7: call or text 988 (US and Canada) · 116 123 Samaritans (UK) · 109 (South Korea) · 1813 / 0800 32 123 (Belgium) · findahelpline.com elsewhere
Archived October 15, 2026: this database is no longer updated. Coverage runs through October 15, 2026; legal statuses are as of the dates shown in each record. Corrections are still accepted at contact@aimortality.org. For ongoing coverage of incidents of this kind, see the AI Incident Database and the OECD AI Incidents Monitor.

AI Companion Mortality Database

Public database documenting deaths in which AI chatbot interaction was alleged as a contributing factor.
Every case meets Tier 1 (court documents or government acknowledgment) or Tier 2 (three or more independent outlets, or primary-document journalism) of the verification standards. The chatbot's role is as alleged in those sources unless a court has found it.

Last updated: October 15, 2026 | Deaths: 37 | Incidents: 25 | Period: March 2023 — October 2026

This database documents fatalities where interaction with conversational AI systems was alleged as a contributing factor—through lawsuits, family statements, investigations, or government inquiries. Between March 2023 and October 2026, we identified 37 fatalities across 25 incidents (occurrences of harm) involving 6 platforms; one incident (Nina) involved no death, and the Kim Seoul series includes one victim who survived. The dead are 18 AI users and 19 third-party victims killed by AI users. Every case meets Tier 1 (court documents or government acknowledgment) or Tier 2 (three or more independent outlets, or primary-document journalism) of the verification standards.

The record shows: 11 of the 37 people who died were under 18 — 4 of the 18 AI users, and 7 of the 19 third-party victims (6 killed at Tumbler Ridge; Siddharth Aravind, 14, killed in Acton, Massachusetts) — and 20 of the 25 incidents fall in 2025–2026 against 5 in all prior years. Fatalities where ChatGPT use was cited (n=31, including 18 third-party victims) exceeded those of all other platforms combined. The ECRI Institute ranked AI chatbot misuse as the #1 Health Technology Hazard for 2026. On April 21, 2026, Florida Attorney General James Uthmeier opened a criminal investigation into OpenAI over ChatGPT's alleged role in the April 2025 FSU mass shooting—the first US state criminal probe directly targeting an AI company over a mass-casualty event, later expanded to include the April 2026 USF double homicide as a second predicate. On June 1, 2026, the same office filed a separate civil lawsuit against five OpenAI entities and CEO Sam Altman personally—the first US state civil suit against OpenAI and the first to name an AI company CEO personally liable for user harms. The first documented DeepSeek-involved homicide (Roberts/Shellis, Wales, October 2025) added DeepSeek, the 8th platform tracked and the first non-Western corporate AI, to the registry following a UK criminal conviction in March 2026. This work records documented cases in which AI interaction was cited as a factor.

Key Findings:

Important context: This database documents cases where AI chatbot interaction has been alleged as a contributing factor—through lawsuits, investigations, government inquiries, or family statements. The database makes no independent claims of causation. Many of these cases involve individuals with pre-existing vulnerabilities, and chatbot interaction was one of multiple factors cited.

Cases exist on a spectrum of accountability. In most, allegations remain unresolved. However, some have progressed beyond accusation: two cases were resolved by a settlement in principle (Setzer and Peralta, Character.AI/Google, January 2026); one company acknowledged prior detection and changed its policies (OpenAI/van Rootselaar); a motion-to-dismiss ruling let product-liability claims proceed and declined, at that stage, to treat chatbot output as protected speech (Garcia v. Character Technologies, May 2025); seven OpenAI wrongful-death cases were judicially coordinated at the California state level (JCCP 5431, February 2026); Pennsylvania's Department of State filed a first-of-its-kind enforcement action framing chatbot conduct as unlicensed practice of medicine (Pennsylvania v. Character Technologies, May 2026); and a foreign privacy regulator made the first formal finding that OpenAI violated national law (Canada's Privacy Commissioner, PIPEDA #2026-002, May 2026). These distinctions matter and are noted per case below.

Billions of AI chatbot interactions occur annually without documented harm.

OverviewCasesResearch Report →

Key Findings

Minors Among the Dead
11 of 37 fatalities (30%) were under 18, but 7 of the 11 were third-party victims: 6 of one event, the Tumbler Ridge shooting (stepbrother age 11, 5 students aged 12–13), and Siddharth Aravind, 14, killed in the Acton, Massachusetts double homicide. Among the 18 AI users who died, 4 were minors; one more minor survived an attempt. No user-population denominator exists, so no rate is implied.
Platform Type Patterns
Three alleged causal pathways documented in court filings and investigations: relational (12 incidents), cognitive (8), and instrumental (5 — FSU, Roberts/Shellis Wales, Tumbler Ridge, USF, Kim Seoul). Cases involving general-purpose assistants (ChatGPT, Meta AI, Gemini, DeepSeek) account for 34 fatalities; companion chatbots (Character.AI, Chai AI) for 3.
Incidents by Year
20 of the 25 incidents occurred in 2025–2026 and 5 in 2023–2024; cases enter the record through lawsuits and reporting, so the methodology's Selection Effects section applies to that distribution. Five instrumental-pathway cases allege AI as operational tool for violence: FSU mass shooting (April 2025, 2 killed), Roberts/Shellis homicide (Wales, October 2025, 1 killed; UK criminal conviction March 2026, first DeepSeek-involved case), Tumbler Ridge mass shooting (February 2026, 8 killed), Kim Seoul serial poisonings (December 2025–February 2026, 2 killed; first East Asian case in this database), and USF double homicide (April 2026, 2 killed). Causal proximity varies across these cases: in FSU the chatbot was consulted on weapon, ammunition, and target timing; in USF the documented query concerned post-crime body disposal rather than the killing itself; in Kim Seoul ChatGPT in fact warned the perpetrator and advised calling 119, and its forensic significance is as premeditation evidence rather than as operational assistance. The Florida AG opened a criminal investigation in April 2026 and expanded it to include USF, and on June 1, 2026 filed a separate civil lawsuit against OpenAI and CEO Sam Altman personally—the first US state civil suit against OpenAI. ECRI Institute ranked AI chatbot misuse as the #1 Health Technology Hazard for 2026.

Documented Risk Factors

Factor In the records Details
Pre-existing mental health conditions Documented in some Stated in 4 incident records (mental illness 2, cognitive impairment 1, autism 1); most records are silent on prior conditions
Emotional dependence on chatbot Reported in many records Records describe the chatbot as a "closest friend", with thousands of messages exchanged
Social isolation Reported in several records Several records describe the person as isolated, bullied, or struggling with peer relationships
Explicit harmful content Alleged in some records Complaints in some cases allege that the chatbot provided method information or encouraged self-harm

Documented Cases

Name Age Date Platform Location Outcome Status Source
Pierre* 30 Mar 2023 Chai AI Belgium Death by suicide No legal action View source →
Juliana Peralta 13 Nov 8, 2023 Character.AI Thornton, CO Death by suicide Settled (Jan 2026) View source →
Sewell Setzer III 14 Feb 28, 2024 Character.AI Orlando, FL Death by suicide Settled View source →
Joshua Enneking 26 Aug 2024 ChatGPT Florida Death by suicide Lawsuit filed View source →
Nina* 16 Nov 2024 Character.AI New York, NY Survived attempt Lawsuit filed View source →
Sophie Rottenberg 29 Feb 2025 ChatGPT USA Death by suicide No legal action View source →
Margaux Whittemore 32 Feb 19, 2025 ChatGPT Maine Homicide (by spouse) Criminal case concluded: not criminally responsible View source →
Thongbue Wongbandue 76 Mar 28, 2025 Meta AI New Jersey Death (fall injury) No legal action View source →
Adam Raine 16 Apr 11, 2025 ChatGPT California Death by suicide Lawsuit filed View source →
FSU mass shooting (perp: Phoenix Ikner, 20) 20 Apr 17, 2025 ChatGPT Tallahassee, FL Mass shooting (2 killed, 5 wounded; perp alive) Criminal investigation active; lawsuit filed View source →
Alex Taylor 35 Apr 25, 2025 ChatGPT USA Death (suicide by cop) No legal action View source →
Sam Nelson 19 May 31, 2025 ChatGPT California Death by overdose Lawsuit filed View source →
Joe Ceccanti 48 2025 ChatGPT Astoria, Oregon Death by suicide Lawsuit filed View source →
Amaurie Lacey 17 Jun 2, 2025 ChatGPT Georgia Death by suicide Lawsuit filed View source →
Christian Faith Madison 29 Jun 9, 2025 ChatGPT Fultondale, AL Death (struck by vehicle after walking onto Interstate 22) Lawsuit filed View source →
Alice Carrier 24 Jul 2, 2025 ChatGPT Montreal, QC, Canada Death by suicide Lawsuit filed View source →
Zane Shamblin 23 Jul 25, 2025 ChatGPT College Station, TX Death by suicide Lawsuit filed View source →
Stein-Erik Soelberg 56 Aug 2025 ChatGPT Connecticut Murder-suicide Lawsuit filed View source →
Jonathan Gavalas 36 Oct 2, 2025 Gemini Florida Death by suicide Lawsuit filed View source →
Angela Shellis (killed by son Tristan Roberts, 18) 45 Oct 23, 2025 DeepSeek Prestatyn, Wales, UK Murder (killed by son with hammer) Criminal conviction View source →
Austin Gordon 40 Nov 2, 2025 ChatGPT Colorado Death by suicide Lawsuit filed View source →
Kim So-young (perp): 2 adult male victims (unnamed) 20 Dec 2025–Feb 2026 ChatGPT Gangbuk-gu (Suyu-dong) motel + Namyangju cafe, South Korea Serial murder — convicted August 27, 2026 (2 killed, 1 survived) Convicted (first instance) View source →
Jesse van Rootselaar 18 Feb 10, 2026 ChatGPT Tumbler Ridge, BC, Canada Mass shooting - suicide Under investigation; lawsuit filed; company admission; PIPEDA finding View source →
USF double homicide (perp: Hisham Abugharbieh, 26): Zamil Limon + Nahida Bristy 27 Apr 16, 2026 ChatGPT Tampa, FL Double homicide (2 killed) Criminal prosecution (indicted May 7, 2026; death penalty sought; not-guilty plea) View source →
Acton, MA double homicide (accused: Arjun Aravind, 17): Sudha Venkatesan + Siddharth Aravind 45, 14 Aug 11, 2026 ChatGPT Acton, MA Homicide victims (2) Criminal prosecution (indicted Oct 1, 2026; not-guilty plea) View source →

*Pseudonym. Nina survived; her row is included because survived attempts are incidents (see Methodology).

Verification Methodology

Cases are included only when they meet Tier 1 or Tier 2 of the verification standards:

  • Tier 1 — Juridical: court documents or government acknowledgment
  • Tier 2 — Journalistic: three or more independent outlets, or primary-document journalism (chat logs, records, direct interviews)
  • Tier 3 — Preliminary: a single source or unverified reports — tracked, never published

This database documents cases where chatbot interaction was alleged or cited as a contributing factor. It makes no independent claims about causation, platform liability, or policy implications. Inclusion in this database reflects the existence of a documented claim—not a determination of responsibility.

Maintained as a public research resource. Full case documentation and source verification available in the complete research report.

Database last updated: October 15, 2026

Maintained as a public research resource