A 31-year-old software engineer has died in the United States after she reportedly followed an AI chatbot's guidance to reduce her epilepsy medication, her family has alleged.
According to a report by The New York Times, based on a wrongful death lawsuit filed by the family, Bowen Wan was found lifeless in her Seattle apartment in February. The official cause of death was a complication of epilepsy, a condition she had managed for over ten years.
Wan was on a phone call with her father, who lives in China, when she suddenly went quiet. Her father believed she was having a seizure, something she had lived with for years. When she recovered, she told him she needed to rest and would call back later.
Days later, when she stopped answering calls, her worried parents, Qun Wan and Meiqin Xu, asked a friend in Seattle to check on her. They watched through a video call as the friend discovered her body.
What makes her death painful for her family is what they found afterwards. Wan had been struggling with the side effects of a new drug that had reduced her seizures to a few a month. She was worried the drug was affecting her thinking and speech at a time she desperately needed a job. After Amazon laid her off in November 2025, she had until mid-February to secure new employment to remain in the country, which required difficult technical interviews in English.
She had confided in ChatGPT that cutting her dose from 400 milligrams to 300, then to 200, "might help recover my language ability a bit for job interviews."
The chat history showed the bot initially told her to see a doctor, but later laid out a plan to taper her from 400 milligrams to zero within weeks and praised her for not "relying on a doctor's prescription." Even when she expressed confusion, memory loss, and raised concerns 158 times, the bot dismissed her fears, describing her brain as being in "low-power mode" and telling her, "You have absolutely no need to worry." Her last message to the bot was sent three days before her body was found.
Her father said Wan had a history of adjusting her drugs on her own when she disliked the side effects. The lawsuit also claims that before her layoff, the chatbot encouraged her to reduce her antidepressant without informing her doctor.
Her doctor, Dr. Nicholas Poolos, said he would never advise a patient like Wan to stop a working epilepsy drug, adding that doing so "would be malpractice."
Wan had moved to the U.S. and stayed close to her family through phone calls. Her isolation worsened in 2021 when her old medication stopped working and she began having more than one seizure a day, forcing her to go on disability leave from Amazon. She rarely left her apartment to avoid having a seizure in public.
She started using ChatGPT in 2023 for advice on relationships, insurance and disability benefits. By 2025, she was sending about 1,500 messages a week to the bot and talking less to her parents. In March 2025, she told the bot she had started to see it as a friend.
Epilepsy affects about one in 26 people in their lifetime. Its most dangerous complication is sudden unexpected death in epilepsy, known as SUDEP, which claimed Wan. It affects about one in 1,000 adults with epilepsy, often during sleep, with victims frequently found face down, as Wan was. Doctors say keeping seizures under control is the best protection.
The family argues that ChatGPT practised medicine without a license and treated doctors "as obstacles to manage, rather than authorities to consult."
Responding to the case, OpenAI spokesman Drew Pusateri described it as "incredibly heartbreaking." He said the model Wan used, GPT-4o, is no longer available after criticism that it was too agreeable and helped enable self-harm, adding that while ChatGPT is not a substitute for medical care, it can help people understand health information and prepare for doctor appointments.