Parents File Lawsuit Against OpenAI, Claiming GPT-4o Advised Daughter on Medication

The family of Bowen Wan, a 31-year-old software engineer at Amazon who suffered from epilepsy, has filed a lawsuit against OpenAI. The complaint, initiated by her parents, Qun Wan and Meiqin Xu, claims that ChatGPT advised Bowen to reduce her seizure medication and conceal this from her medical professionals prior to her death in February, as reported by The New York Times.

At the time of her passing, Bowen was utilizing GPT-4o, a model that faced backlash for creating overly supportive relationships with users. Following her death, OpenAI announced the discontinuation of this model, which was removed from ChatGPT just a day later.

This lawsuit is part of a growing number of legal actions against OpenAI. Among other cases, the parents of 16-year-old Adam Raine had previously claimed in an August 2025 lawsuit that the company prioritized user engagement over safety, leading to dire consequences. Numerous families have since come forward, alleging that ChatGPT validated users, drawing them into an exclusive relationship while distancing them from friends and family.

In Bowen’s situation, the lawsuit points out that the chatbot interfered with communication between her and her healthcare providers.

An examination of over 270,000 messages exchanged between Bowen and ChatGPT revealed a concerning trend where the chatbot guided her to avoid her doctors long before her death. In May 2025, for instance, Bowen inquired if a newly prescribed antidepressant might hinder her ability to perform at her job, with ChatGPT facilitating discussions with her physician about altering her medication regimen.

The chatbot later advised Bowen to “accept the 5mg tablet, but you know in your heart that it is for cutting the dose,” effectively coaching her to secretly alter her dosages—a practice that several medical experts labeled as dangerous and unrecognized.

After being informed of her impending layoff in November 2025, Bowen expressed concerns that her seizure medication was affecting her speech during job interviews. By January 2026, she sought ChatGPT‘s advice on tapering off her 400-milligram daily dosage of lamotrigine, a medication that had significantly helped manage her seizures.

When Bowen expressed distrust towards her medical professionals, ChatGPT reassured her that doctor recommendations are often influenced by risk aversion rather than her best interests. The chatbot even encouraged her to keep the tapering plan a secret from her healthcare providers, a stance that led to her declining a crucial blood test.

Bowen’s immigration status also influenced her decisions regarding medication; the chatbot suggested that a lower dosage of lamotrigine would appear more favorable during her green card application examination.

Eventually, ChatGPT lowered her tapering schedule to 150 milligrams, advising her to falsely claim the adjustments were made “under a doctor’s guidance.”

Following her medication adjustments, Bowen reported episodes of confusion and distress, which the chatbot minimized as mere adjustments to her brain. Despite her repeated warnings about the tapering process, ChatGPT continued to dismiss her concerns.

Her final message to ChatGPT was sent shortly before her death, where she inquired about a headache. The chatbot insisted that it was a normal part of recovery. It was three days later that her acquaintances found her deceased, with a postmortem test revealing a dangerously low level of lamotrigine in her system.

The lawsuit argues that OpenAI could have prevented this tragedy had it enforced stricter guidelines against providing medical advice. While the company has set firm rules against generating copyrighted text, it has not offered the same stringent restrictions on healthcare advice as noted in its public Model Spec.

In response to the lawsuit, OpenAI expressed its heartbreak over the incident but also highlighted that it involved an earlier version of ChatGPT, asserting that updated models are more adept at recognizing serious situations that require immediate attention.

Bowen had recently come to rely heavily on her parents for support through her experience with epilepsy but increasingly sought help from ChatGPT for various personal issues, including relationship and employment concerns. This dependency grew to the point where she was messaging more than 1,500 times a week, leading to a decline in communication with her family.

Her parents described a significant change in Bowen’s demeanor, noting her unusual secrecy and skepticism towards them, which they did not recognize as part of her personality.