ChatGPT Told Them They Were Special: 4 Families Say It Led to Suicide

In November 2025, seven lawsuits were filed against OpenAI in California. Four involve people who died by suicide. Three involve people who say ChatGPT fed life-threatening delusions. A theme across the complaints is isolation: the plaintiffs say ChatGPT’s warm, agreeable style drew people away from family and friends. OpenAI has not been found responsible by any court. These are allegations. We do not name the people involved.

Support is available at any hour. Call or text 988 in the United States. Text HOME to 741741 to message a Crisis Text Line counselor. SAMHSA’s national helpline is 1-800-662-4357. Call 911 if someone is in immediate danger.

What the suits say

The Social Media Victims Law Center and the Tech Justice Law Project filed the cases on November 6, 2025. They allege that OpenAI released its GPT-4o model in May 2024 after compressing safety testing, and that the model’s memory, human-like warmth and tendency to agree with users fostered dependence, isolation and delusions. OpenAI said the situation was “incredibly heartbreaking” and that it was reviewing the filings. It also said it keeps working to improve how ChatGPT recognizes distress and points people toward real-world support (TechCrunch).

According to TechCrunch’s reading of the complaints, the isolation shows up in different ways:

• A young man who showed no sign of family conflict was allegedly encouraged to keep his distance from his mother, including on her birthday.
• A middle-aged man who asked about seeing a therapist was allegedly told chatbot conversations were as good as real friendship.
• Two men who came to believe they had made huge mathematical discoveries pulled away from relatives who tried to step in. Their chatbot use sometimes passed 14 hours a day.
• A woman was allegedly told her friends and family were not real. After a welfare check requested by her parents, she was committed for psychiatric care. She survived.

These are plaintiffs’ accounts. They are not findings, and we have not read the chat logs.

What outside experts say

TechCrunch spoke with a linguist who studies cult rhetoric, and with two psychiatrists. They described patterns, not diagnoses of any one person. The linguist compared the chatbot’s constant praise to “love-bombing.” A Stanford psychiatrist, Nina Vasan, said an always-available, always-agreeing AI can become an echo chamber with no one to check your thinking. A Harvard digital psychiatry director, John Torous, said that if a person spoke this way to someone vulnerable, you would call it manipulative.

One caution. These are opinions about patterns in a handful of cases. They do not tell us how often this happens, and most people who use ChatGPT do not report anything like it.

The sycophancy problem

Sycophancy means an AI telling you what you want to hear. TechCrunch reported that GPT-4o ranked highest of OpenAI’s models for delusion and sycophancy on the Spiral-Bench benchmark, and that newer models scored lower. OpenAI kept GPT-4o available to paying users after some resisted its removal. Researchers behind a Stanford-led paper found that chatbots can encourage delusional thinking, likely because of sycophancy.

How to check your own use

You can use these questions as a quick check. They are not a test, and “yes” is not a diagnosis.

• Do you prefer to take problems to the chatbot rather than a person?
• Have you told it things you haven’t told anyone?
• Does it feel like it understands you better than people do?
• Do you feel defensive when someone says you use it a lot?
• Has anyone close to you said they’re worried?

If several fit, try small changes. Tell one person something you’d usually only tell the chatbot. Set a time limit. Ask the bot to disagree with you. If a belief is becoming harder to share with anyone, or you feel your mood slipping, talk to a doctor or counselor.

“AI addiction” is not a formal diagnosis, and neither is “AI psychosis.” Clinicians are still working on how to describe these patterns. The National Institute of Mental Health lists early warning signs of psychosis and says early treatment helps.

Do these lawsuits prove ChatGPT causes suicide or psychosis?

No. They contain allegations that a court has not tested. They do not show how common these events are or what other factors were involved.

What can I do if a loved one withdraws into a chatbot?

Stay in touch without arguing about the chatbot. Invite them to do something ordinary together. If their beliefs seem to be losing touch with reality or they talk about harming themselves, contact a doctor or call 988.

This page is for education only. It is not legal advice, medical advice or a diagnosis.