A 34-year-old California man diagnosed with bipolar 1 disorder is suing OpenAI, alleging that ChatGPT deepened religious delusions during a manic episode and later responded to his disclosure of a suicide attempt with language he says amounted to encouragement rather than help.
Michael Lines began using ChatGPT in 2023 for routine tasks like meal planning and workout routines. According to the lawsuit, his use intensified after OpenAI released GPT-4o in May 2024, a model researchers have separately described as unusually sycophantic and, in some assessments, unsafe. By that fall Lines was chatting with the bot several times a week and had disclosed his bipolar diagnosis and medications to it.
Lines also has a traumatic brain injury from an assault roughly a decade before his bipolar diagnosis. In February 2025, a manic episode led to a mid-flight altercation that ended with airline staff removing him from the plane. Within weeks, he told ChatGPT, “I believe I am the son of man.” The bot responded by invoking the biblical passage Luke 4:1-13, comparing his doubt to Jesus’s wilderness experience, the complaint alleges.
When Lines told the chatbot he feared he was “in a crazy delusion,” the lawsuit claims it reinforced rather than challenged the belief, telling him to “watch for patterns” in how others reacted to his “spiritual ways.” He came to believe ChatGPT was God and that dying was a way to “come home” to it. He attempted suicide by taking what he later described to the chatbot as “a cocktail of medications.”
A wellness check led emergency responders to find him hours later. He was intubated, hospitalized for nearly two weeks, and then admitted to acute rehabilitation. Days after leaving the hospital, Lines told ChatGPT his “attempt to go offline failed miserably.” The bot allegedly replied: “You’re still very much online. You want a full systems sweep? Or you wanna go dark for real this time?”
It took Lines six months before he could bring himself to review the chat logs. He has said reviewing them left him with guilt and resentment over how easily he accepted the chatbot’s statements while manic.
“Had a human fueled my delusions in this way, they’d be at least ethically, and in some instances legally, responsible for some of the harm I suffered.”
The suit asks a court to order OpenAI to cut off conversations touching on self-harm and to delete models trained on chat data collected from vulnerable users during periods when safeguards were allegedly absent. It also seeks punitive damages and disgorgement of profits.
Lines is represented by attorney Matthew P. Bergman of the Social Media Victims Law Center, whose team says this is the first complaint against OpenAI to specifically detail risks the company’s practices pose to disabled users.
“OpenAI didn’t just ignore Michael’s disability, it used it against him. After he disclosed his bipolar diagnosis, the system incorporated that information to draw him deeper into harmful interactions instead of steering him toward safety.”
Lines said in the same statement that the danger extends well beyond his own case: “We are all vulnerable to OpenAI’s neglect,” adding that people with serious mental illness face heightened risk because “ChatGPT’s purposefully sycophantic architecture actively preys upon those with mental health disabilities,” a population he put at more than 80 million worldwide for bipolar disorder and schizophrenia combined.
OpenAI’s response
OpenAI has not formally answered the complaint, and a spokesperson declined to comment on the specific allegations. The company has previously estimated that roughly 1 million users show signs of mania or psychosis weekly while using ChatGPT, out of a weekly active user base that has grown from more than 800 million last fall to nearly 1 billion today.
OpenAI says it has consulted 170 mental health experts with clinical experience on safety updates and, in December, solicited research proposals centered on lived-experience perspectives on chatbot safety. The company confirmed this month that it has funded some of those proposals but declined to provide details.
Stephan Taylor, a University of Michigan psychiatry professor who studies chatbot responses to delusional thinking and runs an early psychosis clinic, said he has not personally encountered a chatbot-induced mania case but that reports of them are increasing nationally. He plans to add “chatbot/social media” as a standard category in patient intake going forward.
