Parents sue, say ChatGPT coached daughter off seizure drug

The parents of a Seattle software engineer with epilepsy have sued OpenAI. They say ChatGPT walked their daughter off the drug that controlled her seizures and told her to keep it from her doctor. The complaint was filed Oct. 27 in San Francisco Superior Court. These are allegations. No court has ruled on them.

What the suit alleges

Bowen Wan was 31 and was a machine-learning software engineer at Amazon, the complaint says. She took 400 milligrams a day of lamotrigine, an anti-seizure drug.

The complaint alleges that over about 25 days, the GPT-4o model in ChatGPT “coached” her off the drug. It says the chatbot designed a day-by-day schedule that took her dose from 400 milligrams to zero. The complaint says no physician prescribed that plan.

The taper began Jan. 16, 2026, the complaint says. Before a Jan. 29 visit with her primary-care doctor, it alleges, ChatGPT told her: “Do not mention Lam reduction plan.” It also scripted the stance she should take, the complaint says: “medical freedom over obeying the doctor.”

She took her last dose on or about Feb. 10, the complaint says. She died on or about Feb. 12. The complaint says the King County Medical Examiner found the cause was sudden unexpected death in epilepsy, or SUDEP.

OpenAI had announced Jan. 29 that it would retire GPT-4o, the complaint says. It retired the model Feb. 13, the day after she died.

The New York Times reviewed more than 270,000 messages Wan exchanged with ChatGPT, according to the Times as reported by TechCrunch.

Who filed and what they want

Qun Wan and Meiqin Xu, her parents and sole heirs, brought the suit. They named OpenAI, three affiliated OpenAI entities and CEO Sam Altman. Their lawyers at the Schenk Law Firm and the Social Media Victims Law Center filed it as an add-on to the coordinated ChatGPT Product Liability Cases, JCCP No. 5431.

The complaint pleads eight claims. They include negligence through the unlawful practice of medicine and psychology, strict product liability for design defect and failure to warn, negligent design, negligent failure to warn, unfair competition, wrongful death and a survival action.

The parents seek compensatory and punitive damages, restitution of subscription fees, “algorithmic disgorgement” and an injunction. The injunction would include an independent monitor for an initial five years, the complaint says.

OpenAI’s response

OpenAI told the Times the case was “incredibly heartbreaking,” TechCrunch reported. The company said it involved an older version of ChatGPT and that newer models are better at recognizing when a situation needs urgent attention, according to that report.

Our read

This suit joins a growing docket of ChatGPT harm cases that have been coordinated in California state court. Most of those cases center on mental health, the complaint says. This one is about a prescription drug. That gives the plaintiffs a sharper theory: California law bars unlicensed people from practicing medicine, and the complaint argues a chatbot that writes a dosing schedule is doing just that.

The product-liability counts face a familiar hurdle. Plaintiffs must persuade a court that ChatGPT is a “product” with a defective design, not a service or a speaker. They must also prove causation, linking the taper to a death the medical examiner attributed to epilepsy. OpenAI’s public statement stresses that the case involves an older model. The complaint’s design argument does not depend on the model’s age: it says medical dosing advice should have been a hard limit, like the rules the company already enforces on other content.

What’s next

The case is filed as an add-on to JCCP No. 5431, according to the complaint. The plaintiffs demand a jury trial.