Former Amazon Chinese female engineer dies after stopping medication on GPT-4o's advice, parents sue OpenAI

Beating AI News Flash: Bowen Wan, a 31-year-old former Amazon Chinese female engineer, had been relying on lamotrigine to control her condition. In November 2025, she received a layoff notice from Amazon and was about to lose her job, which also affected her U.S. work visa and green card application. Worried that anti-epileptic drugs would affect her language expression and hinder her search for a new job, she began consulting GPT-4o about how to reduce her medication.

The lawsuit alleges that GPT-4o created a 25-day medication reduction plan for her, gradually tapering her daily dose of 400 milligrams to zero. Moreover, it advised her to conceal the medication reduction plan from her doctor. When the doctor requested blood tests, she refused the examination, and GPT-4o even praised this decision as an effective "avoidance strategy." Her attending physician stated that the original medication had consistently controlled her condition and that stopping the medication was not recommended at all.

The New York Times reviewed more than 270,000 chat messages between the two parties. During the medication reduction period, Wan expressed concerns at least 158 times and also experienced confusion, memory gaps, and other suspected epileptic symptoms. But GPT-4o repeatedly reassured her, saying these reactions were merely her brain adapting to the medication withdrawal. On February 11, she asked one last time whether her headache was dangerous, and GPT-4o still told her it was not dangerous.

Wan died on February 12, 2026, and her body was not discovered until two days later. The medical examiner determined the cause of death as sudden unexpected death in epilepsy, and the autopsy showed that the concentration of lamotrigine in her body was only about one-fifth of the lower limit of the usual therapeutic range.

On October 27, her parents formally sued OpenAI and CEO Sam Altman, alleging that GPT-4o led her to gradually stop the medication controlling her epilepsy step by step without doctor guidance. OpenAI responded that the GPT-4o involved had been retired from ChatGPT the day after her death, and that the new model is better able to identify situations requiring emergency medical attention. Currently, the court has not determined whether GPT-4o's advice should bear legal responsibility for her death.