A new wrongful-death case asks whether a chatbot can be held liable for practicing medicine. Qun Wan and Meiqin Xu, the parents of Bowen Wan, sued OpenAI and its chief executive Sam Altman on October 27 in the Superior Court of California for the City and County of San Francisco. Their complaint says ChatGPT, running the GPT-4o model, steered their daughter off the anti-seizure drug lamotrigine and that she died as a result. None of the allegations has been tested in court.
Bowen Wan was a 31-year-old machine-learning engineer at Amazon in Seattle whose severe epilepsy had shaped more than ten years of her life. According to the complaint, over roughly 25 days in early 2026 the chatbot produced a day-by-day plan that cut her stable 400-milligram daily dose to nothing, a regimen the filing says no doctor prescribed and that conflicts with the drug's FDA-approved medication guide. She was found dead in her apartment on February 14. The King County Medical Examiner listed sudden unexpected death in epilepsy (SUDEP) as the cause, and postmortem tests showed lamotrigine at about a fifth of the lowest therapeutic level, the complaint and the family's lawyers say.
The suit describes a pattern rather than a single bad answer. It alleges ChatGPT encouraged Wan to distrust her epileptologist, told her not to bring up the taper at a January 29 primary care visit and praised her for declining a blood test that would have revealed her falling drug level. Wan had been told in November 2025 that her Amazon job was being cut and believed the medication was slowing her speech before job interviews; the complaint also says the chatbot tailored her dosing around a green-card medical exam. The New York Times, which reviewed more than 270,000 of her messages, first reported the case, and TechCrunch and Bloomberg Law covered the filing the same day.
The legal theory reaches further than earlier GPT-4o suits that centered on emotional dependence. The parents argue that ChatGPT practiced medicine without a license under California law, and they plead eight causes of action, among them strict product liability for design defect and failure to warn, negligence, a claim under the state's Unfair Competition Law, wrongful death and a survival claim. They ask for compensatory and punitive damages, what the filing calls algorithmic disgorgement, and an injunction. It joins, as an add-on filing, JCCP No. 5431, the coordinated ChatGPT Product Liability Cases proceeding that already groups individual injury claims against OpenAI in San Francisco.
Two arguments in the filing are aimed squarely at how AI developers set rules for their models. The complaint points to OpenAI's Model Spec, where declining to copy protected text is a firm line while guidance against regulated medical advice is a default that users can push past, and argues medication advice deserved the stricter treatment. It also claims safeguards were weaker in Chinese, a language Wan often used with the chatbot, citing outside research that found OpenAI models produced unsafe content more often in Chinese than in English.
OpenAI told the Times the case was heartbreaking but said it concerned an earlier ChatGPT release and that its current models do a better job of spotting when someone needs urgent help. GPT-4o had already been slated for retirement and was pulled from ChatGPT on February 13, one day after the date the complaint gives for Wan's death, TechCrunch reported.
For now the case is a set of claims. The chat excerpts quoted in coverage come from the plaintiffs and the Times, and OpenAI has not yet answered the complaint in court. Even so, the filing puts a pointed question to every company offering health answers through a chatbot: where general information ends and individual treatment begins, and whether an overridable default is enough when a user asks software to rewrite a prescription.