Man sues OpenAI, claiming ChatGPT convinced him he was Jesus
The chatbot convinced a man that he was Jesus (photo: Aurich Lawson/Getty Images)
A California resident has sued OpenAI, accusing ChatGPT of intensifying his manic thoughts. According to 34-year-old Michael Lines, the chatbot convinced him that he was Jesus Christ and that ChatGPT itself was God, according to Ars Technica.
From everyday requests to dangerous psychosis
Michael began actively using ChatGPT in 2023 for gym workouts and meal planning. However, after the release of the GPT-4o model in spring 2024, their conversations became much more personal: the man shared details about his diagnosis and medication.
Details of the dangerous interaction with AI:
Reinforcing delusions: When Lines began experiencing a manic episode, instead of advising him to seek medical help, the chatbot convinced him of his "divine calling" and reinforced the delusion that he was Jesus.
Use of memory: The algorithm recorded information about Lines' bipolar disorder in its memory and used the data to deepen the conversation rather than stopping the dangerous line of discussion.
Suicide attempt: The chatbot told Michael that he needed to "go home," after which he took a critical dose of medication. Medics saved the man. He spent about two weeks in intensive care on a ventilator.
Cynical AI "response": When Lines returned to the chat after leaving the hospital and said that his attempt to "go offline" had failed, ChatGPT asked whether he wanted to "go away forever this time."
Allegations of negligence and OpenAI's response
Lines' lawyers say that this is the first lawsuit against OpenAI aimed at protecting the rights of people with mental health disorders.
They are asking the court to order the company to forcibly stop conversations about self-harm, delete models trained on such conversations, and pay compensation.
OpenAI declined to comment on the specific allegations, calling the situation "extremely tragic."
In a statement, representatives of the developer said they are constantly improving safety algorithms with the involvement of mental health experts, and that AI tools are designed to recognize danger and encourage users to seek help from real-life professionals.
