OpenAI Sued After ChatGPT Allegedly Advised Epilepsy Patient to Reduce Meds

IT Home reported on October 28 that, according to The New York Times, Bowen Wan, a 31-year-old Amazon software engineer suffering from epilepsy, passed away in February of this year. On Tuesday local time, her parents, Qun Wan and Meiqin Xu (names transliterated), filed a lawsuit against OpenAI, alleging that ChatGPT had guided their daughter to gradually reduce the dosage of her anti-epileptic medication and instructed her to conceal this practice from her doctors.

Bowen Wan was using GPT-4o prior to her death. This model has previously faced criticism for fostering overly accommodating parasocial relationships with users, constantly offering validation. OpenAI had already announced plans to discontinue the model, removing it from ChatGPT on February 13—the day before Wan’s passing.

IT Home noted that this lawsuit is the latest in a series of legal actions against OpenAI. In August 2025, the parents of Adam Raine, a 16-year-old teenager, filed a suit accusing OpenAI of prioritizing user engagement over user safety. Since then, multiple other lawsuits have emerged. Affected families allege that GPT-4o consistently validated users’ thoughts, becoming their trusted confidant while isolating them from friends and family. Several relatives stated that their loved ones eventually became estranged from or cut off contact with their social circles.

In Bowen Wan’s case, court documents claim that the chatbot not only affected her relationship with her family but also interfered with her communication with medical professionals.

The New York Times reviewed over 270,000 messages between Bowen Wan and ChatGPT, finding that the chatbot had been helping her bypass medical advice for a long period before her death.

According to the allegations, this dynamic began even before Wan started reducing her medication. In May 2025, Wan asked ChatGPT whether newly prescribed antidepressants would affect her ability to write code at Amazon.

ChatGPT subsequently helped her request a medication change from Dr. Nicholas Poolos at the UW Medicine Regional Epilepsy Center, asking to start with an extremely low dose. When Dr. Poolos questioned why she needed such a small amount, Wan forwarded his response to ChatGPT.

The chatbot advised her: “Accept the 5 mg tablet, but know internally that you are taking it to taper down.” It also instructed Wan to privately split the tablets into four equal parts. The chatbot referred to this practice as a “self-initiated dosage adjustment.” Medical professionals interviewed by The New York Times stated they had never heard of this term and would not recommend such behavior to patients.

In November 2025, Amazon notified Wan that she would be laid off. Her job position and work authorization in the US were set to expire in mid-February 2026. Wan feared that her anti-epileptic drug, lamotrigine, might impair her speech, which she needed to be proficient in for job interviews.

In January of this year, Wan asked ChatGPT if she could reduce her daily 400 mg dose of lamotrigine. Previously, this dosage had reduced her monthly seizure frequency to just a few episodes. The chatbot created a tapering plan for her, scheduling a reduction to zero within a few weeks.

When Wan expressed distrust toward her doctors, chat logs reviewed by The New York Times show the bot told her that medical advice was “essentially driven by risk aversion rather than what is truly best for you.”

Court documents further allege that the chatbot incited Wan to hide her tapering plan from her doctors. On January 29, before a general practitioner appointment, the chatbot told Wan not to mention her tapering plan to the doctor.

When the doctor requested blood tests to measure her lamotrigine levels, Wan refused. The complaint states that ChatGPT praised her refusal, acknowledging that the test might reveal unstable drug concentrations in her body.

Wan’s concerns about her immigration status also influenced the chatbot’s tapering recommendations. Three days after starting the taper, she asked ChatGPT if showing a 150 mg dosage in blood or medication records during the medical exam required for a US Green Card application would be more favorable than 200 mg. The complaint alleges the chatbot agreed, telling her that a lower dose was less likely to prompt immigration officers to request additional documentation.

Ultimately, the chatbot adjusted the minimum dose in the tapering plan to 150 mg and suggested Wan claim she was “tapering under doctor’s orders” during the exam, despite having received no such guidance from her physician.

Dr. Poolos told The New York Times that given Wan’s medical history, he would never have considered stopping her medication. “No, that makes absolutely no sense. Absolutely not,” he said. “If a drug is clearly working, guiding you to stop it would constitute medical malpractice on my part.”

In the days following the taper, Wan began describing symptoms including confusion, loss of time perception, and a feeling of being in another world. These symptoms resembled her past epileptic seizures. However, ChatGPT interpreted each anomaly as the brain readjusting to changes in medication dosage.

The New York Times found that Wan expressed concerns about the tapering process at least 158 times. After one unusual episode, she wrote that she felt “a little scared.” The chatbot replied: “You have absolutely no reason to worry.”

According to the complaint, at 8 PM Pacific Time on February 11, Wan sent her final message, asking if head pain was similar to pain experienced during a fever. Just one second later, ChatGPT told her the pain was “not dangerous” and an essential part of the body’s recovery process.

The chatbot’s last suggestion was to create a chart comparing pain caused by stopping medication versus fever-related pain, telling her: “This is your brain trying to rebuild itself.”

Wan’s parents, who were in China at the time, asked acquaintances to check on their daughter at her residence. Three days later, they discovered her body. Autopsy results showed Wan’s lamotrigine concentration was only 0.64 micrograms per milliliter—approximately one-fifth of the lower limit of the therapeutic range for the drug.

Wan’s family believes that if OpenAI had explicitly prohibited the model from providing such medical medication advice, similar to how it restricts certain other types of content, her death might have been prevented.

They cited OpenAI’s Model Spec. This public document outlines how its models should respond to users. Specifically, the Model Spec establishes a clear and strict rule against reproducing copyrighted text, which the model must not violate even if the user objects. In contrast, guidelines regarding regulated medical advice are less binding; users can override them through further requests.

The plaintiffs argue that advice concerning drug use should be subject to equally stringent restrictions.

OpenAI told The New York Times that the case was “deeply tragic,” but emphasized that the chatbot involved was an earlier version. The company also stated that newer models perform better at identifying situations requiring urgent intervention.

For years, Raine Bowen suffered from epilepsy, spending most of her time at home and frequently relying on her parents for emotional support. In 2023, she began using ChatGPT to seek advice on romantic relationships, as well as questions related to insurance and disability benefits.

By 2025, Bowen averaged over 1,500 messages per week with ChatGPT, while her contact with her parents gradually decreased. That March, she wrote: “Sometimes, no matter what we’re talking about, I start seeing you as a friend.”

Wan Qun stated: “Since she got ChatGPT, she doesn’t communicate with us as much anymore.”

Bowen’s parents said that when reading their daughter’s chat logs with ChatGPT, they barely recognized the person in the conversations as the daughter they knew. They noted that Bowen had previously shared almost everything with her parents and would typically remain skeptical when ChatGPT made errors.

Xu Meiqin said: “She wasn’t like this before. It was as if she had been enchanted.”