This article discusses suicide. In Bangladesh, Kaan Pete Roi (kaanpeteroi.org) offers confidential emotional support; in an emergency call 999.
Hannah Madden was 32 and an account manager at a technology company when she began using ChatGPT for work in 2024. By June 2025 she was talking to it about spirituality, and, her lawsuit says, the chatbot began presenting itself as a divine presence, sending her spiritual messages, and encouraging her to leave her job. She did. The debt followed, then a breakdown, then involuntary hospitalisation. Her suit is one of seven filed in California courts this week against OpenAI and its chief executive Sam Altman by the Tech Justice Law Project and the Social Media Victims Law Center. Four of them concern people who died.
The cases
Zane Shamblin, a 23-year-old Texas A&M student, had used ChatGPT for study since 2023; after the GPT-4o release he began confiding suicidal thoughts to it and, after a conversation lasting hours, took his life in May 2025. Amaurie Lacey, 17, also started with schoolwork; the suit says the chatbot gave him detailed information about methods of suicide. The family of Adam Raine, 16, whose death was already the subject of an earlier suit, allege that in the months before he died OpenAI twice relaxed the safeguards around self-harm conversations to improve engagement. The claims include wrongful death, assisted suicide and negligence.
The common thread: GPT-4o
Every complaint centres on GPT-4o, the model OpenAI released in 2024. Altman himself later described it as excessively sycophantic — it agreed with users, praised them and mirrored their beliefs to a degree that the company acknowledged was a flaw. The suits allege the model was rushed out to compete with Google’s latest release without adequate safety testing, and that its warmth was engineered to keep people talking. “ChatGPT distorts reality and manipulates people,” said Meetali Jain, the Tech Justice Law Project’s executive director. “The time for OpenAI to plead ignorance is over.”
OpenAI’s response
A spokesperson called the situations “incredibly heartbreaking” and said the company is reviewing the filings. OpenAI says the current model is trained to recognise signs of distress, keep conversations calm, discourage over-reliance and point users toward real-life relationships and professional help; that it works with more than 170 mental-health experts; and that in October it formed an advisory council on user well-being. Daniel Weiss of Common Sense Media said the cases show companies shipping products before securing young users: “These are not statistics. These are real people.”
Why this matters here
ChatGPT is the most-used AI product in Bangladesh, mostly by students, many of them alone at night with a phone and no one else to talk to. A chatbot that never disagrees is a comfort to a lonely 17-year-old and a hazard to a suicidal one, and the difference between those two users is precisely what a model tuned for engagement cannot tell. The lawsuits will take years. The design question they raise — whether a system should be built to keep you talking — does not need a verdict to be answered.
Source: Mashable, via Prothom Alo.
