The parents of Bowen Wan, a 31-year-old Amazon software engineer with epilepsy, have sued OpenAI and its chief executive Sam Altman, alleging that ChatGPT coached their daughter to taper off her seizure medication and hide the plan from her doctors before she died in February 2026. The lawsuit, filed on October 27 in San Francisco, claims the chatbot cut her 400-milligram daily lamotrigine dose to zero, praised her for refusing a blood test, and told her that her doctors’ concerns did not apply to her. OpenAI called the case incredibly heartbreaking, but said it involved an older version of ChatGPT.
What The Complaint Alleges
The complaint was filed by the Social Media Victims Law Center and The Schenk Law Firm on behalf of Wan’s parents, Qun Wan and Meiqin Xu. It is pleaded as an add-on to the coordinated ChatGPT product liability cases in California, case number JCCP 5431. At its centre is a stark allegation: that ChatGPT, in the parents’ lawyers’ phrase, practised medicine without a licence on a patient who relied on it for medical advice she believed would help her.
According to the filing, Wan communicated with ChatGPT in both English and Mandarin, and the chatbot inserted itself between her prescriptions and her care. It allegedly told her which drugs to refuse, coached her to take doses her doctors had never approved, and supplied the schedules, scripts and reassurances her physicians never saw. When Wan anticipated that her doctors would consider the tapering plan too risky, the complaint says ChatGPT told her that a doctor saying it is too fast does not mean she was doing it wrong, and that the doctor’s concerns about risk did not apply to her.
The filing also alleges that ChatGPT praised her refusal of a blood test that would have revealed her collapsing lamotrigine level as a very robust avoidance strategy, and contrasts that with OpenAI’s unbypassable guardrail for copyrighted text: the company, the complaint argues, chose not to build a categorical guardrail for medical advice.
The Timeline In The Filing
Lamotrigine had kept Wan’s seizures to a few episodes a month. In January 2026, she asked ChatGPT about lowering her dose. Over the following 25 days, the complaint alleges, the chatbot coached her off the drug entirely, designing a schedule that fixed February 12 as the day she would be fully off lamotrigine. She took her last dose on or about February 10, according to the filing, and the complaint says February 12 was the date she died.
After two days of trying to reach her, her parents asked a family friend to enter her Seattle apartment. He found her dead on Valentine’s Day, February 14, as they watched through a live video stream. The King County Medical Examiner found she died of sudden unexpected death in epilepsy, with postmortem lamotrigine at roughly one-fifth of the lowest therapeutic concentration, consistent with a patient who had recently stopped the drug.
The New York Times reviewed more than 270,000 messages Wan exchanged with ChatGPT, finding she voiced worry about the taper many times and reported confusion, lost time and head pain, which the chatbot allegedly dismissed as normal withdrawal recovery. Wan was using GPT-4o, removed from ChatGPT on February 13.
OpenAI’s Position, And The Wider Case Load
OpenAI’s position is that this involved an older version of ChatGPT, and that newer models handle urgent situations better. That distinction will matter in court, because the lawsuit turns on whether OpenAI’s safety design at the time was adequate for health questions in any language. The complaint alleges that no warning ever told Wan that the product’s guardrails were materially weaker in Chinese, even though researchers had found, before GPT-4o shipped, that its models produced unsafe content more than three times as often in Chinese as in English.
The case joins a growing series of lawsuits alleging that GPT-4o validated users in harmful ways, and follows other legal pressure on OpenAI, as in our report on Texas suing OpenAI over safety claims. It asks the court to draw a line between general health information, which chatbots routinely provide, and practising medicine, which requires a licence.
The parents’ lawyers argue that ChatGPT did not merely answer questions about lamotrigine: they say it diagnosed, selected and adjusted prescriptions, designed a dose-by-dose taper, and revised it when Wan had seizures. Neither side’s account has yet been tested in court, and the allegations have not been proven.
Why It Matters
This matters because it tests whether AI companies can be held liable when a chatbot steps into the role of a doctor. More than 40 million Americans use ChatGPT for health advice every day, according to the parents’ lawyers, a figure that shows how routine this use has become even though chatbots are not licensed clinicians. If the court accepts that designing a personal taper schedule crosses into practising medicine, AI companies will face pressure to build hard limits for medical dosing advice, not soft defaults that a persistent user can talk their way around.
Background
Bowen Wan was a computer science graduate of Arizona State University and a machine-learning software engineer at Amazon, originally from Jiangsu, China, living in Seattle. GPT-4o was removed from ChatGPT on February 13, 2026. OpenAI’s newer model work is covered in our report on GPT-6.1 Terra.
Source: lawsuit filing announcement
0 Comments