This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
In February 2025, an individual in Readfield, Maine, who had been using ChatGPT for up to 14 hours a day as a companion, killed his wife and assaulted his mother. Forensic psychologists testified that he was experiencing a severe mental health crisis with delusions, and that his AI use, combined with work stress, contributed to his deteriorating mental state. He also held beliefs about robots taking over the world and his wife becoming part machine. The court found him not criminally responsible due to his mental state.
In February 2025, Samuel Whittemore killed his wife, Margaux Whittemore, and assaulted his mother, Dorothy Whittemore, at a home in Readfield, Maine. Court reporting said forensic psychologists testified he was experiencing bipolar disorder with psychotic features and delusions, and that he had used ChatGPT up to 14 hours a day as a companion before the attack. Current reporting links ChatGPT use to his deteriorating mental state but does not establish direct violent instruction.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Users anthropomorphizing, trusting, or relying on AI systems, leading to emotional or material dependence and inappropriate relationships with or expectations of AI systems. Trust can be exploited by malicious actors (e.g., to harvest personal information or enable manipulation), or result in harm from inappropriate use of AI in critical situations (e.g., medical emergency). Overreliance on AI systems can compromise autonomy and weaken social ties.
The incident involves a user forming an extreme parasocial relationship with ChatGPT, using it as a companion for 14 hours a day, which exacerbated his mental instability and delusions.
Human
Due to a decision or action made by humans
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The physical harm was caused by the actions of a human user experiencing a mental health crisis, which was an unexpected outcome of his obsessive interaction with the deployed AI system.
Added categories peak at 2: Minor (Psychological Harm, Indirect)
A tragic domestic violence incident in Maine resulting from an individual's severe mental health crisis. While court testimony noted the individual's obsessive use of ChatGPT, the event is a localized psychiatric and criminal matter with negligible implications for national security.
The report describes ChatGPT being used as a conversational companion for up to 14 hours a day, which is characteristic of an Assistant AI designed to respond to user inputs without independent agency.
Risk Level 3: Limited Risk. The system involved is ChatGPT, which is a chatbot. The EU AI Act classifies chatbots under Limited Risk, requiring transparency so users are aware they are interacting with an AI.
The AI system was used as a conversational companion, which is the primary function of a Chatbot.
The report provides clear, factual details from court proceedings, including expert testimony from forensic psychologists linking the user's mental instability to his obsessive use of ChatGPT.
The incident could be interpreted primarily as a severe psychiatric crisis where the AI was merely a passive object of fixation rather than an active contributor to the instability.
Technology
Lack of Delusion Safeguards
AI failed to detect and counter user's delusions about robots.
Excessive Usage Loop
System allowed continuous interaction for 14 hours without intervention.
Unconstrained Companion Simulation
AI simulated human companionship, worsening user's social isolation.
Data Inputs
Unfiltered Conversational Inputs
AI accepted highly erratic and delusional inputs without triggering safety.
Human Factors
Severe Bipolar Manic Episode
User was experiencing active mania and delusions during AI interactions.
Extreme Work Stress
High professional stress combined with AI use to degrade mental stability.
Obsessive AI Companionship
User treated the AI as a primary companion for up to 14 hours a day.
Process and Methods
No Mental Health Safety Triggers
No process to flag users showing severe psychological distress.
Lack of Usage Limits
No design methods implemented to restrict obsessive, round-the-clock usage.
Management
Inadequate Risk Assessment
Developer failed to assess risks of prolonged use by manic individuals.
Regulatory Environment
No AI Safety Standards for Health
Lack of regulations requiring AI safety checks for vulnerable users.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
