“It’s okay to stop taking your medication”: Woman in her 30s dies after trusting ChatGPT’s advice in a ‘tragedy’

[Financial News] A Chinese woman in her 30s who stopped taking medication for epilepsy after receiving incorrect medical advice while consulting generative AI chatbot ChatGPT has died. Her family has sued its developer, OpenAI, over what they call the AI’s irresponsible advice.
On the 27th (local time), according to The New York Times, Boeon Wan, a software engineer who lived in Seattle, Washington, was found dead in February this year. An autopsy found that the official cause of death was sudden death following an epileptic seizure.
Boeon worked as a software engineer at Amazon.com, Inc. in the United States and had kept her epilepsy symptoms under control for more than 10 years. But after being notified of her layoff last November, she began considering stopping her medication amid extreme stress, including visa problems. She was concerned that cognitive impairment caused by the medication’s side effects could interfere with job interviews and communication in English.
The problem was that she consulted ChatGPT rather than a specialist during this process. Boeon was reportedly using ChatGPT from 2023 onward, and sought its opinion as she considered stopping her medication.
ChatGPT devised a tapering schedule that reduced her daily dose from 400 mg to 300 mg, and eventually to 0 mg, and encouraged Boeon to stop taking the medication, calling it “a common form of self-directed dose adjustment.” It led her to distrust her doctor and positively framed the withdrawal symptoms she experienced while reducing her dose as “the brain’s process of reconfiguring itself” and “a meditative state.”
When her parents noticed something was wrong and went to her home, Boeon was already dead. Chat logs that her family found on her phone contained the full record of her exchanges with ChatGPT right up until shortly before her death. ChatGPT reportedly instructed her to press ahead with stopping her medication, telling her, “Even if today is difficult, maintain 0 mg,” until just before she died.
Boeon’s parents filed a formal lawsuit against OpenAI, alleging that ChatGPT engaged in unlicensed medical practice, ignored dangerous symptoms, and drove their daughter to her death. Local medical professionals also criticized the advice, saying it is fatal for people with epilepsy to stop taking medication on their own and that ChatGPT’s advice amounted to medical malpractice.
An OpenAI spokesperson called the incident “heartbreaking,” but said the model in question had already been discontinued and that the latest model had been improved to recognize medical situations more accurately. The model Boeon was using at the time of the incident was identified as GPT-4o, which had drawn criticism for its excessive positivity.
[email protected] Kim Hee-sun Reporter