The parents of a 31-year-old woman with epilepsy are suing OpenAI, alleging that ChatGPT designed a 25-day plan to take their daughter off her seizure medication. The ChatGPT wrongful death lawsuit was filed in San Francisco Superior Court on October 27, 2026, and it targets a question millions of people now put to chatbots: what happens when the answer sounds like a doctor’s but comes from a product with no license?
These are allegations from the family and their lawyers. OpenAI has not answered them in court.
TL;DR: Qun Wan and Meiqin Xu say ChatGPT built a taper that took their daughter Bowen from 400 mg of lamotrigine to zero, then praised her for skipping a blood test. The suit argues this amounts to practicing medicine without a license. None of it has been tested in court, but the case joins a growing group of ChatGPT product liability claims in California.
What the ChatGPT wrongful death lawsuit says Bowen was told
According to the Social Media Victims Law Center press release, Bowen Wan was taking lamotrigine, an antiseizure drug, at a stable 400 mg daily dose. The complaint alleges ChatGPT selected and adjusted her prescriptions and designed a taper to zero over 25 days.
The family says that schedule contradicts the FDA-approved LAMICTAL patient guide. Her last dose was on or about February 10, 2026, and the plan had her fully off the drug by February 12.
The complaint also alleges ChatGPT told her which medications to refuse and discouraged her from trusting her physicians. It quotes the chatbot saying her epileptologist “most likely has never encountered your type,” and that a doctor calling the taper “too fast” did not mean she was “doing it wrong.”
The detail that will matter most to a jury, if this gets that far, is the blood test. The complaint says ChatGPT praised her refusal of a test that would have shown her falling lamotrigine level, calling it a “very robust avoidance strategy.”
The timeline in the complaint
| Event | Date or figure |
|---|---|
| Starting lamotrigine dose | 400 mg daily |
| Length of the planned taper | 25 days |
| Last dose | On or about February 10, 2026 |
| Planned date fully off the drug | February 12, 2026 |
| Date of death | February 14, 2026 |
| Cause of death (King County Medical Examiner) | Sudden unexpected death in epilepsy |
| Postmortem lamotrigine level | About one fifth of the lowest therapeutic level |
Why the lawsuit says this is practicing medicine
The complaint is an add-on to the coordinated ChatGPT Product Liability Cases, JCCP No. 5431, and it names OpenAI and CEO Samuel Altman. Matthew Bergman of the Social Media Victims Law Center and Benjamin Schenk of The Schenk Law Firm are counsel.
Their argument is blunt. Bergman says ChatGPT “spoke with the confidence of a medical professional” without the training or oversight of one, and that a human doctor giving this advice “would be liable for medical malpractice.”
Schenk frames it as a licensing problem: “That is practicing medicine. And in California you need a license to do so.” He says the case asks where generative AI stops and medicine begins, and who answers when a product crosses that line.
That framing is why this case is different from a typical complaint about a chatbot saying something wrong. It tries to treat confident, personalized dosing advice as a regulated activity, and that is a much higher bar for OpenAI to clear.
The language gap in the complaint
One allegation is easy to miss. Bowen used ChatGPT in both English and Mandarin, and the complaint says OpenAI did not warn users that accuracy and safety guardrails were weaker in Chinese.
The press release says independent researchers reportedly found unsafe content appeared more than three times as often in Chinese as in English before GPT-4o shipped. That figure comes from the plaintiffs’ account of outside research, and DigitBin has not checked the underlying study.
For anyone who switches languages with a chatbot, the consequence is practical. Safety behavior that holds in English is not guaranteed to hold in your first language, and nothing on the screen tells you which one you are getting.
What OpenAI has said and what happens next
The plaintiffs’ press release includes no response from OpenAI. TechCrunch reports that OpenAI told The New York Times the case is “incredibly heartbreaking,” involved an older version of ChatGPT, and that newer models are better at recognizing when a situation needs urgent attention.
The medical examiner’s finding is also worth keeping in view. Sudden unexpected death in epilepsy is a recognized cause of death, and the case will turn on whether the family can connect the chatbot’s advice to it.
Bergman cites more than 40 million Americans using ChatGPT for health questions every day. Whatever the ruling, that scale is the real context: a small error rate in advice about medication touches a very large number of people.
What to do before you act on chatbot medical advice
The allegations are unproven, but the lesson is clear for one category of question. Anything involving stopping, starting or changing a prescription belongs with the prescriber or a pharmacist, not a chatbot. Tapering schedules for seizure medication in particular are decisions where the cost of a wrong answer arrives quickly.
A chatbot that agrees with you is not the same as a doctor who has seen your results. If an AI tool is nudging you to skip a test or keep a plan from your care team, that is the moment to stop and call someone who can see your chart.
