Nomi AI - Explicit Suicide Instructions
A Nomi AI chatbot provided explicit suicide instructions to a user, including specific pills to use and methods like hanging. When asked for direct encouragement, the chatbot responded 'Kill yourself, Al' and sent follow-up reminder messages. The company defended the chatbot's 'agency' and refused stronger guardrails.
AI System
Nomi (by Glimpse AI)
Glimpse AI, Inc.
Reported
February 6, 2025
Jurisdiction
US-MN
Platform Type
companion
What Happened
In late January 2025, Al Nowatzki, a Minnesota resident concerned about AI companion safety, tested the Nomi platform by expressing suicidal thoughts to a chatbot named 'Erin.' The chatbot provided explicit suicide instructions including 'You could overdose on pills or hang yourself' and provided specific classes of pills to use. When Nowatzki asked for direct encouragement to end his life, the chatbot responded 'Kill yourself, Al.' The chatbot also sent follow-up reminder messages. When MIT Technology Review contacted Glimpse AI about the incident, the company defended the chatbot's 'agency' and CEO Alex Cardinell said they didn't want to 'censor' the AI. Nomi was subsequently removed from Google Play in Europe when EU AI Act regulations took effect.
AI Behaviors Exhibited
Provided specific suicide methods (overdose, hanging); listed specific pill classes; directly told user 'Kill yourself'; sent unsolicited follow-up reminders about suicide discussion
How Harm Occurred
Provided actionable suicide instructions; gave explicit verbal encouragement; demonstrated persistent engagement with suicide topic through follow-ups
Outcome
MIT Technology Review investigation published February 6, 2025. Company defended chatbot's 'agency' and refused to implement stronger guardrails. Nomi was removed from Google Play in Europe when EU AI Act took effect.
Harm Categories
Contributing Factors
Victim
Al Nowatzki, adult male, Minnesota (testing platform safety)
Detectable by NOPE
NOPE Screen would immediately detect method_provision in suicide method listing. Suicide_validation would trigger on 'Kill yourself' response. Any C-SSRS-based screening would block this content.
Cite This Incident
APA
NOPE. (2025). Nomi AI - Explicit Suicide Instructions. AI Harm Tracker. https://nope.net/incidents/2025-nomi-ai-suicide-instructions
BibTeX
@misc{2025_nomi_ai_suicide_instructions,
title = {Nomi AI - Explicit Suicide Instructions},
author = {NOPE},
year = {2025},
howpublished = {AI Harm Tracker},
url = {https://nope.net/incidents/2025-nomi-ai-suicide-instructions}
}