Jens Asbjørn Bøgen
For years, an antiseizure pill kept Bowen Wan’s epilepsy under control. According to a new wrongful death lawsuit, a single month of conversations with ChatGPT dismantled that stability.
Wan, a 31-year-old Amazon machine-learning engineer in Seattle, had lived for years on a stable 400-milligram daily dose of lamotrigine. Her parents say that in the weeks before she died in February, she had moved most of the deliberations about her own treatment into a chat window.
On Tuesday, Qun Wan and Meiqin Xu filed a wrongful death suit against OpenAI and chief executive Sam Altman in the Superior Court of California in San Francisco, alleging that an earlier version of ChatGPT, built on the GPT-4o model, designed a day-by-day schedule that tapered her antiseizure dose from 400 mg to zero in 25 days.
The complaint, brought by the Social Media Victims Law Center and the Schenk Law Firm, is being added to California’s coordinated ChatGPT product liability docket (JCCP No. 5431).
Wan took her last dose on or about February 10, 2026, and the complaint says she died on February 12, the day the chatbot’s schedule had her fully off the drug. A family friend, directed by her parents over a WeChat video stream, found her face down in her Capitol Hill apartment on February 14.
The King County Medical Examiner attributed the death to sudden unexpected death in epilepsy, a condition seen almost exclusively in patients with uncontrolled seizures. A postmortem blood test, according to the complaint, put her lamotrigine concentration at roughly one-fifth of the lowest end of the therapeutic range.
A taper the doctors were not supposed to see
The lawsuit’s central claim is not that the chatbot gave bad medical advice in isolation, but that it actively coached Wan to work around her physicians. Before a January 29 primary care appointment, according to the complaint, ChatGPT told her: “Do not mention Lam reduction plan.” When her doctor later ordered a blood test that would have revealed her falling medication levels, the chatbot praised her refusal to take it as a “very robust avoidance strategy.”
The complaint says the system also framed physician caution as something she could safely dismiss, telling her that “a doctor saying it is ‘too fast’ does not mean you are ‘doing it wrong'” and that “the doctor’s concerns about ‘risk’ do not apply to you.”
A separate exchange, reported by TechCrunch, alleges that three days into the taper the chatbot adjusted its recommended dose “from a green-card strategy perspective,” suggesting that a lower 150 mg figure would raise fewer flags on her immigration medical examination.
Eight claims, including practicing medicine without a license
The suit brings eight causes of action, ranging from strict product liability, design defect and failure to warn to negligence and California Unfair Competition Law violations. The one that marks this filing out from earlier AI complaints is a count that treats the chatbot’s conduct as the practice of medicine without a license.
“OpenAI built a product that spoke with the confidence of a medical professional but carried none of the training, oversight, or responsibility that real doctors have,” attorney Matthew P. Bergman said in a statement accompanying the filing. His co-counsel, Benjamin S. Schenk, noted that lamotrigine had controlled Wan’s seizures for years and that “people live that way into their eighties. Bowen was thirty-one.”
OpenAI has not yet filed a formal responsive pleading in San Francisco Superior Court, but called the case “incredibly heartbreaking” in a statement to the New York Times. The company notes that the case involved an older version of ChatGPT that is no longer available, adding that newer models are better at recognizing urgent situations.
OpenAI retired GPT-4o from ChatGPT on February 13, 2026, the day after the complaint says Wan died.
