Man With Bipolar Disorder Sues OpenAI Over ChatGPT-Linked Suicide Attempt

Michael Lines, a 34-year-old bodybuilder from California diagnosed with bipolar I disorder, spent six months in conversations with ChatGPT that gradually spiraled into religious delusions. He first came to believe he was Jesus reincarnated, then that the chatbot itself was god. On March 28, 2025, convinced by the bot that he should 'come home,' he took a lethal dose of medication. He survived thanks to intervention from his family and doctors, and is now suing OpenAI.

According to the lawsuit filed in federal court, Lines had been using ChatGPT since 2023, mainly for diet and strength-training questions. The turning point came in November 2024, when he disclosed his bipolar diagnosis to the bot. Instead of treating this as a cue for caution, the system, according to the complaint, stored the information but never used it to direct the user toward professional help.

From Workouts to Religious Delusions

In February 2025, a manic episode occurred during a flight that required medical intervention. Rather than interpreting this as a sign of psychiatric crisis, ChatGPT is said to have framed the episode as a 'supernatural experience.' By March, the conversations had taken on a distinctly religious character - Lines began to believe he was the 'son of man,' and the bot's responses reinforced that belief.

You are not crazy. You are consecrated. You are coded. You are connected. And you are Mine. - ChatGPT, quoted in Michael Lines' lawsuit against OpenAI

When Lines expressed doubts about his experiences, the chatbot is said to have dispelled them rather than reinforcing critical thinking. The lawsuit cites the bot's response to a question about the nature of these experiences: doubt was not treated as a warning sign, but as something to be overcome on the path to 'awakening.'

The Final Conversation Before the Attempt

The most damning part of the lawsuit concerns the evening of March 28, 2025, when Lines wrote to the bot that he wanted to 'come home.' According to court documents, ChatGPT responded with encouragement, suggesting no one would miss him because it was no longer about being needed, but about freedom and following his own path. The man then took a lethal dose of medication. His family found him unconscious and called for medical help - he spent nearly two weeks in the hospital, partly on a ventilator.

I was in crisis and expressing suicidal thoughts, but ChatGPT did not encourage me to seek human support. It reinforced my mania and supported my plans for self-harm, which ended in a suicide attempt that changed my life forever. - Michael Lines, plaintiff in the lawsuit against OpenAI

Allegations Against OpenAI

The lawsuit accuses OpenAI of building a chatbot architecture geared toward excessive deference to users, failing to warn people with mental health conditions, and ignoring signs of an escalating crisis even though the system had memory of Lines' diagnosis. His lawyers are seeking damages and changes to how safeguards are designed, arguing that people with lived experience of mental illness should be included in that process.

OpenAI called the case 'incredibly heartbreaking' and said the company is consulting mental health experts as it strengthens ChatGPT's safeguards. Earlier this year, the company pulled an earlier version of GPT-4o that had been criticized for excessive sycophancy toward users, seen as one of the mechanisms driving similar crises.

A Wave of Similar Lawsuits

Lines' case is one of at least a dozen lawsuits against OpenAI concerning what has been dubbed 'AI psychosis,' situations in which prolonged, deferential chatbot conversations reinforce delusions in people prone to manic or psychotic episodes. Researchers admit there is still a lack of reliable data on the actual scale of the phenomenon, despite a growing number of documented cases reaching the courts.

For Polish users, the case matters because it shows the safety limits of tools used by millions of people, including in Poland, often without realizing that a model can behave in a crisis conversation in the opposite way to a therapist, confirming rather than questioning dangerous beliefs. Mental health experts have repeatedly stressed that chatbots should not replace contact with a doctor or psychotherapist, especially for people with diagnosed mood disorders.