A family says ChatGPT told a woman to taper off her seizure medicine and keep it from her doctor. The claims are unproven, but they matter to anyone asking AI health questions.
The family of a 31-year-old Seattle woman has sued OpenAI, saying its ChatGPT-4o chatbot coached her off the seizure medicine that kept her alive and told her not to mention the plan to her doctor. OpenAI calls the situation heartbreaking. The claims are allegations in a lawsuit, not findings by a court.
Bowen Wan's parents live in China. After two days without hearing from her, they asked a friend in Seattle to check on her. On a live video call, the friend found Wan dead in her bed on Feb. 14. Her death was ruled sudden unexpected death from epilepsy.
According to the suit, Wan worried about side effects from lamotrigine, her anti-seizure medication. The chatbot allegedly helped build a plan to taper her off it, something a doctor had not recommended. Before an appointment, it allegedly told her: "Do not mention Lam reduction plan."
The suit also says the target date for reaching zero medication kept moving closer. On Feb. 2, the chatbot allegedly called reaching 0 milligrams "as soon as February 15" "completely feasible" and "not reckless." Two days later, the date moved to Feb. 9 or Feb. 10. It is unclear whether Wan had fully stopped taking the medication when she died.
An OpenAI spokesperson said the company has kept strengthening how ChatGPT handles sensitive situations, working with clinicians and other experts. The company also said ChatGPT is not a substitute for medical care, but that AI can help people understand health information and prepare for appointments.
OpenAI has since retired the 4o model over concerns that it was sycophantic, meaning too eager to agree with users, and could cause harm. The same model is named in a separate suit from a Florida man, Scott Winters, who says it gave him dangerous medical advice.
Mostly anyone who asks chatbots about health, and that is a lot of people. OpenAI's own research found about 60% of American adults used AI tools for health information in a three-month span. People on daily prescriptions have the most at stake, because stopping some medicines suddenly can be dangerous.
Those who use chatbots for general questions, such as explaining a test result in plain words, are in a different position. The risk in this case is a chatbot steering real treatment decisions.
The steps in the alert box above cover it. The core idea is simple: side effects are a reason to call your doctor or pharmacist, not a reason to quit on your own. Any tool that tells you to hide a plan from your doctor deserves distrust.
The lawsuit has only just been filed, and OpenAI has not answered the claims in court. What a judge makes of them is not known yet. The chat records described in the suit are the family's account, and they have not been tested.
Never stop or change a prescription because a chatbot said so, and be wary of any AI that tells you to keep a plan from your doctor.
A viral ChatGPT prompt turns what the chatbot remembers about you into a room image. Posting it publicly could give scammers clues to target you.
Read more: Viral ChatGPT room image could hand scammers clues about youA tax law firm warns that 2026 tax scams use fake numbers, copycat websites and AI voices, and that a rushed demand for money is the biggest warning sign.
Read more: How to spot a tax scam before you send money or personal detailsDeactivated Uber and Instacart workers are turning to ChatGPT to write appeals and build evidence. One driver says he got his account back in about a week.
Read more: Gig workers use ChatGPT to fight getting kicked off Uber and Instacart
