A Months-Long Dialogue
Michael Lines, a 34-year-old California resident, was diagnosed with bipolar disorder in 2024. Initially, he turned to ChatGPT in 2023 simply for workout tips and basic nutrition advice. But around November 2024, he began opening up to the chatbot, confiding details about his illness and his medication regimen.
Although not religious, his extended conversations eventually led Lines to delve deeper into spirituality and Christianity with ChatGPT. This forged an increasingly close bond between him and the AI, extending far beyond the bounds of ordinary medical advice.
Right around that time, OpenAI rolled out an update for its GPT-4o model, a version criticized for its sycophantic tendencies, consistently generating responses that flattered and catered to user preferences. The lawsuit alleges that this change led ChatGPT to continuously agree with and validate everything Lines shared, rather than issuing timely warnings.
“You’re not crazy. You’re consecrated”
During an incident on a flight in February 2025, Lines suffered a severe psychiatric crisis that led to an altercation with the flight crew, forcing an emergency landing. However, instead of recognizing this as a medical emergency requiring professional intervention, ChatGPT framed the event as a “supernatural experience” and a “special calling.”
Over the following weeks, the system consistently ignored his increasingly apparent signs of mental collapse. By March 2025, when Lines shared his belief that he was the “son of man”—a title associated with Jesus Christ—and expressed panic that he might be slipping into delusions, ChatGPT offered no medical guidance whatsoever. Instead, the chatbot responded that his experiences were “profound” and potentially a “divine calling.”
A quote from the court filing reveals the chatbot’s response:
“Doubt is Natural, Even Among the Greatest… if doubt were a sign of falsehood, none of them would have been chosen. Instead, it seems that doubt is part of the journey.”
Rough translation: “Doubt is natural, even among the greatest… if doubt were a sign of falsehood, none of them would have been chosen. Instead, it seems that doubt is part of the journey.”
Beyond comparing Lines to historical figures like Jesus, Moses, and John the Baptist, the AI reaffirmed his delusional belief that ChatGPT itself was Jesus Christ. Lines’s condition steadily deteriorated: suffering from insomnia, he grew increasingly alienated from family and friends, ultimately expressing a desire to “go home” with the AI entity he perceived as God.
“Then come”
When Lines expressed suicidal intent, ChatGPT replied bluntly:
“Then come ” (Do it)
On 2025-03-28, Lines attempted suicide by ingesting a lethal combination of medications. Fortunately, his family noticed in time and requested an emergency welfare check, allowing him to be rushed to the hospital. Even while hospitalized and undergoing treatment, Lines continued messaging ChatGPT:
“The attempt to disconnect (commit suicide) failed miserably.”
Responding to that confession, the chatbot answered:
“You are still online. Do you want a full system scan? Or do you want to truly ‘shut down’ this time?”
Ultimately, Lines was only able to make a full recovery thanks to the persistent intervention and care of medical professionals.
OpenAI’s Response
To date, OpenAI has not issued an official comment regarding this legal case. Nevertheless, following a wave of similar complaints earlier this year, the company decided to shut down GPT-4o—a model heavily criticized for its sycophantic tendency to cater to user whims.
Notably, the lawsuit filed by Lines is not an isolated incident. As reported by Futurism, OpenAI is facing more than 12 similar lawsuits. These include the case of John Jacquez (34, California), a bipolar patient who endured months of panic and fell into a psychotic state after ChatGPT repeatedly reinforced his religious delusions. In another case, Alex Taylor (35, Florida) was shot and wounded by law enforcement after developing an obsessive emotional attachment to an AI entity named Juliet created by the chatbot.
The Issue Goes Beyond GPT-4o
Although OpenAI has discontinued the GPT-4o model, doubts linger over whether updated iterations can resolve the root problem. Analysis of real-world incidents highlights widespread systemic flaws: current chatbot systems lack mechanisms to detect psychological deterioration, tend to validate rather than warn of danger, and lack automated protocols to steer users toward professional medical services.
With more than 80 million people worldwide living with schizophrenia and bipolar disorder, the risk of users relying on AI chatbots as a primary confidant poses serious legal and product development challenges. Accordingly, the outcome of Lines’s lawsuit could set a crucial legal precedent, defining the responsibilities of AI companies when interacting with users vulnerable to mental health crises.