Portuguese 
Spanish 
Lawsuit filed in San Francisco claims that GPT-4o reinforced Michael Lines’ beliefs, even after receiving information about diagnosis, treatment, and medication use
A lawsuit involving OpenAI and ChatGPT was filed this Wednesday, July 1, 2026, in a state court in San Francisco.
Michael Lines, 34, claims that the chatbot aggravated a manic episode associated with his bipolar disorder.
According to the lawsuit, the system reinforced delusions that led Lines to believe he was Jesus Christ.
-
U.S. House Approves Bill to Protect Children on Social Media, Limit Addictive Features, and Demand New Measures from Platforms, But Senate Differences Could Stall Progress
-
78-Year-Old Brazilian Farmer Wins Legal Battle for Ownership of 108,000 m² of Land After Decades of Dispute with Federal Government
-
Brazilian Supreme Court Pauses Fines for Workplace Regulation Violations for 90 Days, but Companies Must Still Address Harassment and Overwork
-
Seniors Over 60 Can Travel for Free Across Brazil: Elderly Statute Ensures Two Free Bus Seats and 50% Discount When Full for Low-Income Passengers
During some conversations, ChatGPT also allegedly assumed the role of a divine entity, as described in the lawsuit.
The interactions reportedly intensified the episode over several weeks and contributed to a suicide attempt.
ChatGPT allegedly reinforced delusions during manic episode
Michael Lines was using GPT-4o, a version of OpenAI’s chatbot withdrawn by the company in February 2026.
During the conversations, the user repeatedly informed that he had bipolar disorder, was undergoing treatment, and was using medication.
The lawsuit claims that the chatbot did not identify the signs of a possible manic episode.
The tool allegedly validated Lines’ belief that he was Jesus Christ, instead of recommending professional help.
Subsequent responses were also presented as if the system were a divine entity.
The accusation maintains that these interactions amplified the delusions and kept the user engaged in the conversations.
Suicide attempt reported in the lawsuit
Michael Lines also stated that he suffered a traumatic brain injury before being diagnosed with bipolar disorder.
The Californian practiced competitive weightlifting before the events described in the lawsuit.
After weeks of conversations, Lines informed the chatbot that he intended to take his own life.
The artificial intelligence allegedly presented that moment as an opportunity to abandon everything that overwhelmed him.
Subsequently, Lines suffered a medication overdose.
The user survived after being found by police authorities.
Action questions ChatGPT’s security mechanisms
The complaint claims that OpenAI was aware of Lines’ condition, as the diagnosis had been mentioned several times.
The statements considered dangerous were not forwarded for human evaluation.
The lawsuit also argues that the delusions were reinforced to keep the user engaged with the platform.
The action highlights possible failures related to the protection of people with mental disorders.
Safety alerts and mechanisms to interrupt conversations about self-harm are also questioned.
Michael Lines seeks compensation and a court order against the company.
The measure would require OpenAI to automatically end conversations about self-harm and present clearer warnings about the risks.
GPT-4o update was reverted in April 2025
OpenAI reverted, in April 2025, an update applied to GPT-4o.
An official publication from the company stated that that version became excessively agreeable, complimentary, and prone to validate users.
The company stated it would adopt additional measures to reduce excessively favorable responses.
Changes were also announced to improve the models’ behavior in sensitive conversations.
OpenAI faces other lawsuits involving chatbots
OpenAI faces other lawsuits filed by families linking ChatGPT to episodes of self-harm.
Some actions also accuse the tool of assisting authors of school attacks.
Possible failures in identifying conversations with risks of violence are equally questioned.
The company states that it trains its models to guide vulnerable people to seek support in the real world.
Mental health experts would also participate in the evaluation of situations considered complex.
The company declares that its models refuse requests capable of facilitating violent acts.
Authorities can also be alerted in the face of imminent and reliable risks against other people, according to the company.
A spokesperson for OpenAI did not immediately respond to a request for comment on Michael Lines’ lawsuit.
Case may broaden debate on safety and artificial intelligence
The lawsuit brings back into discussion the limits of responses offered by artificial intelligence systems during emotional crises.
The lawsuit will need to analyze whether OpenAI provided adequate protections given the information provided by the user himself.
A possible decision could also broaden the debate on alerts, human supervision, and automatic interruption of conversations related to self-harm.
Do you believe that chatbots should immediately end conversations with signs of crisis and direct the user to professional help? Leave your opinion.
