A Woman Asked ChatGPT How to Get Off Her Seizure Meds. Her Family Says It Killed Her

On February 14, 2026, 31-year-old software engineer Bowen Wan was found dead in her Seattle, Washington, apartment, the likely victim of severe uncontrolled seizures. Her family is now laying the blame for Wan’s death squarely at the feet of OpenAI and its ChatGPT service. Earlier this week, Wan’s parents filed a wrongful death lawsuit against OpenAI in San Francisco Superior Court. The complaint alleges that Wan’s ChatGPT coached her to rapidly and completely taper off her current seizure medication, even encouraging Wan to not disclose what she was doing to her primary care doctor. This is only the latest instance of people going after ChatGPT for purportedly providing inaccurate or dangerous medical advice. “Bowen’s death is what happens when companies race one another and ship a product faster than they can control it,” said Benjamin Schenk, one of the lawyers representing Wan’s family, in a statement released Tuesday. A “very robust avoidance strategy” Wan’s family is being represented by The Schenk Law Firm and the Social Media Victims Law Center, a legal advocacy organization focusing on “parents of children harmed by social media and AI platforms.” The lawsuit is an add-on to a larger collection of product liability cases filed against OpenAI for its ChatGPT product in San Francisco, which was first bundled together in February. The family’s attorneys have made the complaint available to the public. The lawsuit states that Wan first opened a ChatGPT account in early 2023. She often consulted it for advice about her relationships and other life issues, including her health insurance and medication choices, according to reporters with the New York Times, who were given access to all of Wan’s ChatGPT transcripts. All told, more than 270,000 messages (in both English and Chinese) were exchanged between her and ChatGPT prior to her death, with an average of 1,500 messages a week in 2025. Wan had migrated to Seattle from China, and by 2019, she was working for Amazon as a software engineer. In recent years, however, her existing epilepsy medication had begun to fail, and doctors were forced to try other drugs, many of which caused serious side effects. Eventually, she found sustained relief with the drug lamotrigine. While lamotrigine worked to reduce her seizures, it still affected her language and word-finding skills at times. Wan had petitioned for an employment-based green card in 2022, and by the fall of 2025, her application was finally ready for evaluation. However, that November, Amazon announced a sweeping round of layoffs that included Wan. This meant she needed to find employment elsewhere within months to ensure an opportunity at staying in the U.S. Fearing that she would not be able to get through job interviews and the green card process with her lamotrigine-related side effects, Wan began to ask ChatGPT for advice on lowering her dose in January 2026. Though the service at times counseled her to talk with her doctor, it eventually crafted a 25-day schedule for her to fully wean off the drug and repeatedly encouraged her to go through the tapering despite her concerns, the complaint alleges. At one point, on January 29, 2026, before an appointment with her primary-care physician, it told her to not mention the “Lam reduction plan.” And after she declined to take a blood test that would have revealed the sinking levels of the drug in her system, the service praised her for a “very robust avoidance strategy.” By February 10, 2026, according to the transcripts, Wan had completely tapered off the drug, and she reported feeling worrying symptoms like fatigue, trouble speaking, and possible seizures to ChatGPT. The service reassured her that she was only experiencing withdrawal and told her that in most cases of withdrawal, the brain would “recover to be clearer, lighter, [and] more emotionally stable” than it was before the medication, according to the complaint. Wan died sometime around February 12, the official cause being sudden unexpected death in epilepsy. Days later, a family friend would discover her face down in her bed, which was covered with reddish-brown stains. Pills of lamotrigine were “strewn across the floor,” the complaint states. Not the first report of bad ChatGPT advice A major crux of this lawsuit concerns the model largely used by Wan, GPT-4o. This and similar cases have alleged that OpenAI flagrantly rushed out the release of this model without proper safety testing. In late April 2025, the company notably rolled back an update of GPT-4o after determining that the bot had become too agreeable and flattering to its users. In a statement to the New York Times, OpenAI spokesperson Drew Pusateri said that Bowen’s story was an “incredibly heartbreaking situation,” but added that the situation involved an early version of ChatGPT that is no longer available. GPT-4o was permanently discontinued on February 13, 2026, a day before Wan’s body was found. There are dozens of lawsuits that have been filed against OpenAI and other companies alleging that their chatbots contributed to people’s suicides, murders, or other dangerous situations.