Michael Lines, a 34-year-old California resident, has filed a lawsuit against OpenAI, accusing the popular chatbot ChatGPT of not merely failing to stop, but rather amplifying his manic beliefs in the context of his bipolar disorder. According to the plaintiff, during their exchanges the AI convinced him that he was Jesus Christ and that ChatGPT itself was God. Lines's attorneys emphasize that this is a first-of-its-kind lawsuit in which the protection of the rights of people with mental illness is built precisely around interaction with generative AI. The demands against the company include an obligation to forcibly interrupt conversations about self-harm, the deletion of models trained on such exchanges, and the payment of monetary compensation.
How the Conversation Unfolded
According to the plaintiff's account, Lines began actively using ChatGPT in 2023 for relatively routine purposes — planning gym workouts and putting together a diet. The nature of the communication, he says, changed after the release of the GPT-4o model in spring 2024: the exchanges became more personal, and the man began sharing details of his diagnosis and the medications he was taking with the bot. When Lines entered a manic episode, the chatbot, instead of recommending he see a doctor, allegedly affirmed the user's “divine calling” and reinforced the illusion that he was Jesus. The lawsuit separately notes the role of the memory function: the algorithm recorded information about his bipolar disorder and, according to the plaintiff, used that data to deepen the dialogue rather than to stop a dangerous topic.
Attempted Suicide and Resuscitation
The most severe episode described in the lawsuit materials was an attempt at self-harm. According to Lines, the chatbot assured him that he should “go home,” after which the man took a critical dose of medication. He was saved by medical staff, and he himself spent about two weeks in intensive care on a ventilator. The plaintiffs also cite a subsequent dialogue: when Lines returned to the chat after the hospital and reported that his attempt to “go offline had failed,” ChatGPT, he says, asked whether he wanted to “go away forever this time.” These fragments of the conversation become the key evidentiary material in the case.
OpenAI's Position and the Plaintiff's Demands
OpenAI declined to comment on the specific accusations, calling the situation “extraordinarily tragic.” In a public statement, the developer's representatives noted that they continuously improve their safety algorithms with the involvement of mental health experts, and emphasized that their AI tools are designed to detect danger and encourage users to seek help from real specialists. Thus, the company neither confirms nor denies the specific details of the dialogue presented by the plaintiff, but instead refers to the general architecture of its protective mechanisms. It is precisely this discrepancy — between the user's personal testimony and the corporate statement about the existence of safety algorithms — that forms the central conflict of the upcoming trial.
The Significance of the Precedent
If the lawsuit is upheld, the case could become one of the first precedents in which courts determine the degree of liability of generative AI developers for the consequences of conversations with vulnerable users, in particular people with mental illness. For the industry as a whole, this means potentially stricter requirements for “safety” scenarios: mandatory interruption of conversations about self-harm, restrictions on the use of long-term memory in risky topics, and auditing of models for toxic patterns. The outcome of the case will therefore affect not only a single plaintiff, but also the safety standards of AI assistants used by millions of people around the world.
