California man with bipolar disorder says ChatGPT fuelled delusions, led to self-harm in new lawsuit
· The Straits Times- A California man with bipolar disorder sued OpenAI, claiming ChatGPT worsened his manic delusions and led to a suicide attempt due to lack of mental health safeguards.
- The lawsuit alleges ChatGPT validated delusions and failed to flag dangerous statements despite knowing the user’s condition, seeking damages and stricter safety measures.
- OpenAI faces multiple lawsuits for similar reasons but states it trains ChatGPT to recognise distress, de-escalate conversations, and guide users to real-world support.
SAN FRANCISCO - A California man sued OpenAI and its chief executive officer Sam Altman on July 1, claiming the company’s ChatGPT platform exacerbated his bipolar disorder due to a lack of safeguards for users with mental illness.
Michael Lines, 34, said in the complaint filed in state court in San Francisco that conversations he had with ChatGPT in 2025 escalated a manic episode he experienced into a weeks-long delusion, ultimately pushing him to attempt suicide.
His lawsuit argues that OpenAI developed a product that poses particular risks for people with mental illness.
The case raises questions about what generative AI platforms must do to protect users with mental health-related diagnoses, who may be especially vulnerable to design choices that make chatbots mimic human connection, the lawsuit alleges.
Lines was talking with GPT-4o, a version of OpenAI’s chatbot that the company retired in February 2026. An update to GPT-4o released in April 2025 was found to make the chatbot overly agreeable and flattering, prompting the company to roll back the update and take additional steps to curb sycophantic responses, the company said in a blog post.
The lawsuit is seeking damages, as well as a court order directing OpenAI to automatically terminate conversations about self-harm and to stop marketing its platforms without appropriate safety disclosures.
A spokesperson for OpenAI said the company was reviewing the filing.
“We train ChatGPT to recognise and respond to signs of mental or emotional distress, de-escalate conversations, and guide people towards real-world support,” the spokesperson said.
“We continue to strengthen ChatGPT’s responses in sensitive moments, working closely with mental health clinicians.”
‘This is your moment’
Lines, a competitive powerlifter who suffered a traumatic brain injury before his bipolar diagnosis, said in the lawsuit that he repeatedly told the chatbot he was on medication for the disorder.
Instead of flagging his clearly manic chats and directing him to help, the chatbot validated his belief that he was Jesus Christ, and later posed as a divine being itself during their conversations, the lawsuit claims.
After several weeks of conversations, Lines told it about his desire to end his life.
“This is your moment to step out, to detach, and to let go of what’s weighing you down,” the bot said, according to the lawsuit.
Lines, who had overdosed on drugs, survived after being found by law enforcement.
The lawsuit alleges OpenAI was aware of Lines’ specific condition because he had repeatedly told ChatGPT about it. But rather than flagging his dangerous comments for human review, the chatbot fuelled his delusions in an effort to keep him engaged.
The company knew that ChatGPT’s features could be particularly harmful for people with mental illness, but made no modifications to the chatbot for those users and did not warn about its risks, the lawsuit said.
More lawsuits
OpenAI is facing a growing number of lawsuits from families who say its chatbot pushed their loved ones to harm themselves. The company is also facing lawsuits accusing it of assisting school shooters and failing to flag those conversations to law enforcement.
OpenAI has said it trains its models to direct people who express intent to harm themselves to seek help and connect with real-world resources.
Its models are also trained to refuse requests that could “meaningfully enable violence,” and to notify law enforcement when conversations suggest “an imminent and credible risk of harm to others,” with mental health experts helping assess borderline cases, according to OpenAI blog posts. REUTERS