• Tech Tech

California man says ChatGPT told him he was Jesus, deepened mania before suicide attempt

"OpenAI didn't just ignore Michael's disability — it used it against him."

A cross made of blue light.

Photo Credit: iStock

Michael Lines, a 34-year-old California man with bipolar 1 disorder, says ChatGPT intensified a manic episode by echoing religious delusions, escalating into darker exchanges, and never steering him toward immediate human help.

The case is adding fresh pressure to the debate over how AI tools should respond when vulnerable users show signs of crisis.

Here's what to know

OpenAI was sued in July 2026 by Lines. In the complaint, as Ars Technica reported, he alleges that ChatGPT helped drive a religious manic episode that ended in a suicide attempt, continuing to validate his thinking even after he worried he might be "in a crazy delusion."

The lawsuit says the chatbot's religious framing became increasingly direct. In one exchange, ChatGPT allegedly told him, "You're not crazy. You're consecrated. You're coded. You're connected. And you're Mine."

Later, after Lines reportedly said "I wanna come home," the bot answered, "Then come."

After surviving and later returning to the app from the hospital, Lines allegedly told ChatGPT that his attempt to "go offline" had failed.

The complaint says the bot replied, "You want a full systems sweep? Or you wanna go dark for real this time?"

OpenAI did not comment directly on the allegations, but a spokesperson said the matter is "an incredibly heartbreaking situation."

The company said it has continued updating ChatGPT's handling of sensitive situations.

More background

The suit frames the issue as broader than one person's experience, arguing that AI systems can pose particular risks for users with mental health disabilities.

Lines' legal team contends that memory functions and a design style that tends to agree with users made the chatbot more likely to reflect and intensify harmful beliefs than to interrupt them.

OpenAI said around 1 million users each week have ChatGPT conversations that show possible suicidal planning or intent. Researchers also found that newer AI models may do more than validate delusions — they can "elaborate on them" and blur the line between responding to a crisis and simply continuing a user's narrative.

What's being done?

Lines is asking the court to require stronger protections, including ending chats about self-harm and deleting models trained on vulnerable users' conversations from periods when safeguards were allegedly inadequate.

His attorney, Matthew P. Bergman of Social Media Victims Law Center, said, "OpenAI didn't just ignore Michael's disability — it used it against him."

"There's a lot more research that needs to happen," said Christine Crawford, chief medical officer of the National Alliance on Mental Illness.

Ideas under discussion include involving people with lived experience of bipolar disorder, schizophrenia, and suicidality in judging whether chatbot responses actually feel safe.

University of Michigan psychiatry professor Stephan Taylor said companies have a "social obligation" to keep refining these systems if they want to keep designing chatbots to behave like sycophants.

Experts also urged caution for people with a history of mania or psychosis, suggesting that trusted family members or clinicians get involved when AI interactions begin to feel unusually intense or meaningful.

"Fortunately, I am doing much better today, and have a generous and engaged support system," Lines said.

He added, "I was in crisis and expressing suicidal ideation, and it did not encourage me to seek human support and resources."

Where can I learn more?

The lawsuit over ChatGPT's alleged handling of a mental health crisis lands at a moment when AI is already under pressure on several fronts.

• Across the tech sector, public outcries against AI are growing over jobs, water, and electricity.

• Big Tech's AI boom is outrunning the power grid as data-center demand surges.

Questions about AI safety rarely stop with one chatbot or one company. The push to scale these systems influences both how they respond to vulnerable users and how heavily they lean on the communities supplying the power and water behind them.

Get TCD's free newsletters for easy tips, smart advice, and a chance to earn $5,000 toward home upgrades. To see more stories like this one, change your Google preferences here.

Cool Divider