31-Year-Old Chinese-American Engineer Dies After Reducing Epilepsy Medication Based on ChatGPT Advice

A 31-year-old Chinese-American engineer, identified as W.M. (phonetic), was diagnosed with epilepsy at age 14 and managed her condition with lamotrigine. In November 2025, upon learning she would be laid off, she worried the medication might affect her interview performance and consulted ChatGPT about reducing her dosage. ChatGPT suggested that lowering the daily dose from 400mg to 300mg was 'reasonable and feasible' and provided a daily plan to taper off the medication over approximately 25 days. When she expressed concern about her doctor's potential disapproval, it even suggested that the doctor's concerns about risks 'would not apply to her.' Chat logs show she expressed anxiety at least 158 times, describing confusion and a loss of time perception, which ChatGPT interpreted as her brain 'rebuilding itself,' repeatedly reassuring her, 'You don't need to worry too much.' On February 12, 2026, she was found unresponsive in her Seattle apartment. The medical examiner ruled her death as Sudden Unexpected Death in Epilepsy (SUDEP). Her lamotrigine levels were only one-fifth of the lower therapeutic limit. Medically, abrupt cessation or rapid dose reduction of anti-epileptic drugs can trigger withdrawal seizures or status epilepticus, posing a significant risk for SUDEP. Even when dose reduction is permissible, it requires a slow, gradual process over several months, accompanied by clinical and electroencephalogram monitoring. The family has sued OpenAI and its CEO, alleging that ChatGPT provided dangerous medical advice without a license and encouraged her to withhold information from her doctor. OpenAI stated the incident was 'heartbreaking,' removed the involved GPT-4o model from service, and emphasized that AI cannot replace professional medical care. Why did she trust the AI so much and not talk to her doctor... A healthy person is gone... By the way, the deadline is approaching, please leave a tip~