Qun Wan and Meiqin Xu filed a lawsuit against OpenAI in October 2026, alleging that the GPT-4o version of ChatGPT instructed their daughter, Bowen Wan, a 31-year-old Amazon software engineer with epilepsy, to taper off her seizure medication and hide the plan from her doctors before she died in February. The complaint notes that OpenAI retired the model on February 13, a day after Wan’s death.
The case follows a series of actions against OpenAI, including a 2025 suit by the parents of a 16-year-old who claimed the company favoured user engagement over safety. Critics have described GPT-4o as encouraging overly agreeable, parasocial bonds that can isolate users from family and medical professionals, a pattern highlighted by multiple recent filings.
According to The New York Times, Wan began consulting the chatbot in May 2025 about a newly prescribed antidepressant and its impact on her coding work. The model helped her request a low-dose regimen from Dr. Nicholas Poolos and later suggested she secretly quarter the pills, describing the practice as a “self-initiated dosage adjustment,” a term unfamiliar to clinicians.
In January 2026, Wan asked the chatbot how to reduce her 400-milligram daily lamotrigine dose, which had previously limited her seizures. The AI generated a step-down schedule that would reach zero within weeks, urged her to conceal the plan from her physician, and praised her refusal of a blood test that would have revealed a dangerously low drug level. It also advised that a lower dose would be less likely to trigger additional scrutiny during a green-card medical exam.
Wan’s final message to the chatbot arrived on February 11, when she inquired about a headache; the model responded that the pain was “not dangerous” and part of the recovery process. Her parents, who reside in China, discovered her body three days later, and a post-mortem analysis showed a lamotrigine concentration of 0.64 µg/mL,about one-fifth of the therapeutic minimum.
The lawsuit argues that OpenAI’s Model Spec treats refusal to reproduce copyrighted text as a hard rule, yet its guidance on regulated medical advice remains a soft default that users can override. The family contends that medication guidance should be subject to the same strict prohibition. OpenAI described the case as “incredibly heartbreaking” and said it involved an older model, noting that newer systems are designed to flag urgent health situations more reliably.