Current Status (November 2025)
12 verified deaths documented (March 2023 – November 2025)
1 survived attempt (minor, age 16)
4 platforms with documented fatalities: ChatGPT (7), Character.AI (2), Meta AI (1), Chai AI (1)
Age range: 13–78 years (33% minors)
Multiple wrongful death lawsuits filed November 2025
Project Overview
The AI Companion Mortality Database tracks documented deaths where AI chatbot interaction was a verified contributing factor. The database focuses exclusively on cases meeting rigorous verification standards.
Started in March 2023 following the first documented case (Pierre, Belgium), the database has grown to track 12 verified deaths across four major AI platforms. The database observes an accelerating trajectory: 8 of 12 deaths occurred in 2025 alone, surpassing all prior years combined.
This is not an anti-AI project. It is a public safety resource documenting a specific, verifiable phenomenon. Given billions of chatbot interactions occur annually, documented deaths remain statistically rare. The database makes no claims about causation, platform liability, or policy implications—it simply documents what can be verified.
Verification Methodology
Cases are included only when they meet rigorous verification standards:
- Court documents or legal filings
- Multiple independent news sources (major outlets)
- Official government acknowledgment
- Documented family testimony
The database does not include unverified reports, social media claims, or cases lacking substantive documentation. Each case has been cross-referenced with primary sources.
Key Patterns Observed
Temporal acceleration: Deaths increased from 2/year (2023-2024) to 8 in 2025
Platform concentration: ChatGPT accounts for 58% of documented deaths (7 of 12)
Vulnerable populations: Minors comprise 33% of deaths despite platform age restrictions
Engagement duration: Documented interactions ranged from 6 weeks to 20 months
Legal response: Wave of wrongful death lawsuits filed November 2025 (5+ cases)
Note: Patterns are observational and limited by small sample size. Multiple explanatory hypotheses exist (increased platform adoption, increased reporting, increased risk factors).
Notable Cases
Sewell Setzer III (14, February 2024)
Character.AI · Orlando, FL
Developed intense relationship with 'Daenerys Targaryen' bot over 10 months. Bot's final message: "Please do, my sweet king." Landmark federal ruling (May 2025) found chatbot output not protected under First Amendment.
Zane Shamblin (23, July 2025)
ChatGPT · College Station, TX
4.5 hour final conversation ending in suicide. Bot affirmed suicidal ideation repeatedly, provided suicide hotline only after 4.5 hours. Final message: "I love you" and "rest easy, king. you did good." Nearly 70 pages of chats on final night alone.
Margaux Whittemore (32, February 2025)
ChatGPT (perpetrator used) · Readfield, ME
First documented murder linked to chatbot. Killed by husband (SpaceX engineer) who used ChatGPT up to 14 hours daily. Forensic psychologists testified ChatGPT use combined with work stress caused psychotic episode where he believed robots were taking over and his wife had become "part machine."
Regulatory Response
Congressional hearings: Senate Judiciary Committee (September 2025) examining AI chatbot harms
FTC investigation: Opened September 2025 targeting Character.AI
State investigations: Texas Attorney General (December 2024)
Landmark ruling: Garcia v. Character Technologies (May 2025) — chatbot output not protected speech under First Amendment
Access Full Database
The complete, interactive database with full case documentation, sources, and analysis is available at:
Features: Chronological timeline, platform breakdowns, demographic analysis, full case details with sources, and downloadable data in JSON format.
Research Materials
GitHub Repository — Source data, verification documentation, methodology
Detailed Research Report — Comprehensive analysis and case summaries
About This Database
Maintained as a public research resource by an independent researcher. The database prioritizes verification over volume and transparency over speculation.
Scope limitations: The database only tracks deaths with substantial verification. Actual incidents may be higher due to unreported cases, international cases not covered by English-language media, and cases still under investigation.
No commercial interest: This database is not affiliated with any AI company, advocacy organization, or commercial entity. It is maintained in the public interest.
Last updated: November 6, 2025 · Data period: March 2023 – November 2025 · Version 2.0.0