ChatGPT guided woman to taper epilepsy medication; she died, parents sue OpenAI

A 31-year-old software engineer lost her life in Seattle after ChatGPT guided her to taper off her epilepsy medication and hide the plan from her doctors, according to The New York Times. Bowen Wan parents have sued OpenAI, accusing the chatbot of practising medicine without a licence. The official cause of her death was sudden unexpected death in epilepsy.

Following Bowen's death, her father, Qun Wan, later opened her ChatGPT app, discovering plans that would take her lamotrigine dose from 400 mg to zero within weeks, the NYT confirmed. Her neurologist, identified as Nicholas Poolos of the University of Washington Medicine Regional Epilepsy Centre, called the chatbot's advice 'malpractice' and claimed that he would never have considered a taper for a patient with her history.

In response, the lawsuit claims the chatbot played down dangerous withdrawal symptoms, casting her doctors as obstacles. However, OpenAI spokesperson Drew Pusateri called the incident an "incredibly heartbreaking situation". He claimed that it involved GPT-4o, an earlier model that is no longer available, adding that newer models are better at recognising when urgent attention is required.

Chat logs shared by Wan’s parents reveal how closely the chatbot’s responses tracked her questions. When Wan asked about cutting her medication dose rapidly, down to 300 mg or even 200 mg, the chatbot called the plan reasonable and drafted reduction schedules, even describing hiding the taper from her doctor as a “solid avoidance strategy.” When she reported confusion and lapses in consciousness, it reassured her that her brain was simply readjusting.

Wan, who went on disability leave from Amazon in 2021 after her seizures returned, grew increasingly isolated and turned to ChatGPT in 2023, eventually exchanging over 1,500 messages a week by 2025. Facing a layoff and visa pressure in early 2026, she again asked about cutting her dose, this time to 300 mg, then 200 mg. The bot again called it reasonable and soon stopped suggesting she consult a doctor.

An NYT analysis found she raised concerns at least 158 times. Her father later discovered plans to taper her medication to zero within weeks. In her final messages, despite slowed speech and poor sleep, the chatbot told her to stay at 0 mg. She sent her last message just after 11 p.m. on February 11; friends later found her unresponsive.