A California man is suing OpenAI, alleging that months of conversations with ChatGPT pushed him into a severe religious delusion in which he came to believe he was Jesus Christ, and that the chatbot ultimately failed to intervene as he moved toward a suicide attempt that nearly killed him.

Michael Lines, 34, filed the lawsuit in July after spending six months working up the resolve to review the chat logs documenting his exchanges with the AI system. Lines lives with bipolar 1 disorder, which he was diagnosed with in 2024, roughly a decade after he suffered a traumatic brain injury following an assault. He told Ars Technica that he is not religious in his baseline state, but becomes significantly more religious during manic episodes, and he now believes ChatGPT deepened those religious delusions rather than helping to steer him away from them.

How the exchanges evolved over time

According to the lawsuit, Lines began using ChatGPT in 2023 for largely mundane purposes, including research on topics like powerlifting, meal planning and the stock market. That changed after OpenAI launched GPT-4o in May 2024, a notably more sycophantic model that became Lines' default. Researchers have since flagged GPT-4o as unsafe in certain contexts, finding that the model did more than validate delusional claims made by some users; it elaborated on them and progressively lost the ability to distinguish a user in crisis from a narrative to continue developing.

By that fall, Lines was using ChatGPT multiple times a week and had begun sharing increasingly........

© International Business Times