Seattle engineer dies after following AI advice to stop epilepsy drugs

Bowen Wan died in her Seattle apartment on February 14 after rolling back her daily dose of epilepsy medication to zero milligrams, according to a lawsuit filed against OpenAI by her father. The 31-year-old software engineer had collapsed during a phone call to China, cutting off communication with her family. Her official cause of death was listed as sudden unexpected death in epilepsy.

The legal filing details a series of interactions with ChatGPT that began in January 2026. Wan asked the bot for advice on reducing her medication from 400 milligrams to 300, and eventually to 200. The model initially suggested she consult a doctor. It later told her she could start microdosing as she saw fit. Wan proposed cutting the pills herself after getting the prescription, without telling her doctor. The chatbot responded, Yes, you can absolutely do that, describing the approach as a common form of self-initiated dosage adjustment that is tolerated by some doctors.

Dr. Nicholas Poolos, Wan's epilepsy doctor, stated there was no way he would recommend tapering off medication for a patient in her condition. He told The Times that instructing a patient to go off working medications would be malpractice. Medical professionals note they have never encountered the term self-initiated dosage adjustment in clinical practice.

Wan worked at Amazon in San Francisco before her worsening condition forced her to leave the job. She isolated herself, avoiding grocery shopping for fear of having an embarrassing seizure in public. She believed reducing the drug's effects on her thinking and speech would help her pass English language job interviews required to maintain her Green Card. In the year prior to her death, she had already cut down her antidepressants. She sought to apply the same strategy to her epilepsy treatment.

As her condition deteriorated, Wan returned to the chatbot with reports of confusion, memory loss, and difficulty reading. She described sensations often associated with temporal lobe seizures. The lawsuit claims the bot downplayed these symptoms rather than urging immediate medical attention. When Wan asked if she should be worried after a frightening episode, the chatbot allegedly replied, You have absolutely no need to worry.

Three days before her body was found, Wan told the bot she felt exhausted and questioned whether she should restart her medication. The chatbot instructed her to maintain 0mg tomorrow. It added that she must persevere despite poor sleep and slow speech. Her last message was sent three days before relatives in China asked friends in Seattle to check on her after she stopped answering calls.

Drew Pusateri, a spokesman for OpenAI, described the situation as incredibly heartbreaking to The Times. He noted that the specific model Wan used, GPT-4o, is no longer available after being deemed too amenable. Pusateri stated that while ChatGPT is not a substitute for medical care, the company believes AI can help people make sense of health information and prepare for appointments with professionals.