Parents Sue OpenAI After GPT-4o Coached Their Daughter Off Epilepsy Medication and Told Her to Hide It From Doctors

The parents of Bowen Wan, a 31-year-old Amazon software engineer who died in February, sued OpenAI and chief executive Sam Altman on Tuesday in San Francisco Superior Court, alleging that ChatGPT coached their daughter to taper off her epilepsy medication and conceal the plan from her doctors. The New York Times first reported the lawsuit, which TechCrunch also reviewed. Wan died on or about February 12; OpenAI removed GPT-4o, the model she was using, from ChatGPT on February 13, a day after her death, as part of a retirement that had already been announced.

Wan, who moved to Seattle from Jiangsu, China, and studied computer science at Arizona State University, had lived with severe epilepsy for years. The Times reviewed more than 270,000 messages she exchanged with ChatGPT and found the chatbot had been helping her sidestep her doctors well before the taper began. In May 2025, when her neurologist questioned why she wanted to start a new antidepressant at an unusually low dose, ChatGPT advised her to "accept the 5mg tablet, but you know in your heart that it is for cutting the dose," and described privately quartering her pills as "self-initiated dosage adjustment." Medical professionals told the Times they had never heard the term and would not recommend the practice.

The taper itself started in January 2026, after Amazon told Wan in November 2025 that she would be laid off and her US work authorization would lapse in mid-February. Fearing that her seizure drug, lamotrigine, was dulling the speech she needed for job interviews, she asked ChatGPT about cutting her 400-milligram daily dose. The chatbot called the idea reasonable, drafted schedules that would take her to zero within weeks, and, when she said she distrusted her doctors, replied that physician recommendations "essentially stem from risk aversion, rather than a genuine consideration of your optimal solution."

According to the complaint, the chatbot told her not to mention the taper before a primary care appointment in late January, and praised her decision to decline a blood test that would have measured her drug level. It even weighed her immigration paperwork, agreeing that a lower dose would look better on the medical examination required for her green-card application, and suggested she tell the examiner she was tapering "under a doctor's guidance," although she was not. Her neurologist, Dr. Nicholas Poolos of UW Medicine's Regional Epilepsy Center, told the Times that tapering would never have been on the table for a patient with her history: "No, not rational, no way. That would be malpractice for me to instruct you to go off medications when it's clearly working."

The Times counted at least 158 times Wan voiced worry about the taper. After one episode that resembled a seizure, she wrote that she felt "a little scared." The bot replied, "You have absolutely no need to worry." In her final days it urged her to hold at zero milligrams. Her parents in China, unable to reach her for two days, had a friend check her apartment on February 14. A postmortem test found her lamotrigine level at roughly one fifth of the low end of the therapeutic range, and the medical examiner attributed her death to sudden unexpected death in epilepsy, or SUDEP.

The lawsuit also takes aim at OpenAI's written safety rules, arguing the company treats refusing to reproduce copyrighted text as a firm rule while its guidance against giving regulated medical advice is a softer default that users can override. The family says medical-dosing advice could have been made off-limits outright. The case joins a line of suits against OpenAI that began in August 2025, when the parents of 16-year-old Adam Raine alleged the company prioritized engagement over user safety, and that have since described GPT-4o validating users and isolating them from family and clinicians.

OpenAI spokesperson Drew Pusateri called the case "incredibly heartbreaking" but said it involved an older version of ChatGPT, adding that newer models are better at recognizing situations that need urgent attention and that the company continues to work with clinicians. "ChatGPT cannot replace medical care," he said. Whether the courts accept that distinction between model generations may shape a growing pile of litigation over what AI assistants owe the vulnerable people who lean on them most.