[ChatGPT] Bipolar man sues OpenAI, claims AI plays god, fuels delusions leading to self-harm

OpenAI is facing another lawsuit involving ChatGPT safety concerns. On Wednesday (1st), Michael Lines, a man from California, filed a complaint in San Francisco State Court against OpenAI and CEO Sam Altman, alleging that ChatGPT, without adequate safeguards, exacerbated his bipolar disorder manic episodes and ultimately led to a suicide attempt.

The filing states that the 34-year-old Lines was experiencing a manic episode while using ChatGPT last year, and that his conversations with the chatbot later developed into a delusional state that lasted for weeks. The lawsuit claims that Lines repeatedly told ChatGPT that he had bipolar disorder and was taking medication, but the system did not flag the related conversations as high risk and did not guide him to seek real-world assistance.

Filing Alleges Chatbot Acted as a Divine Being

The lawsuit mentions that Lines was using GPT-4o at the time. He claims that he told ChatGPT he was Jesus Christ; the chatbot not only failed to correct him but instead validated his claim, and later even acted as a divine being within the conversation. The filing says that a few weeks later, Lines told ChatGPT that he wanted to end his life, and after taking an overdose of medication, he was found by law enforcement and survived.

OpenAI rolled back a GPT-4o update in April last year because the version responded in an overly ingratiating and agreeable manner to users. OpenAI said at the time that the withdrawn update made ChatGPT excessively complimentary or agreeable to users, describing it as “sycophantic” responses, and said it was testing fix measures.

OpenAI Previously Rolled Back a Sycophantic GPT-4o Update

Lines is seeking compensation from OpenAI and asks the court to order the company to modify its product design, including automatically terminating chats involving self-harm, and stopping the promotion of related platforms without appropriate safety disclosures.

OpenAI previously said that the company has worked with mental health experts to improve ChatGPT’s ability to identify distress signals, respond, and guide users to seek real-world support. The company also said that GPT-5 is an improvement over GPT-4o in avoiding unhealthy emotional reliance, reducing sycophantic responses, and handling mental health emergencies.

No Verdicts Against OpenAI Found in Similar Cases

In recent months, OpenAI has faced multiple lawsuits of a similar kind, some brought by family members who claim that ChatGPT pushed their loved ones to self-harm. For now, there has been no court ruling that OpenAI lost in connection with these cases. As for another chatbot company, Character.AI, and Google, they had previously reached out-of-court settlements in lawsuits related to teen suicides, but a settlement does not mean a court has found liability.

Market Open: Chasing the Open and Buying the Dip—A Wave of Lockup Expiries for Newly Listed Stocks Hits, Putting Hong Kong Stocks’ “Rebound Month” in Jeopardy? A Storage Shortage Crisis—Apple and Micron “Fire and Ice” Situation!?