Bowen Wan was chatting on the phone with her father when suddenly, she fell silent.
It looked like she was having a seizure—something she had dealt with for years—though this one seemed to last unusually long, according to her dad. When she finally came to, she mentioned needing to rest before talking more.
Several days later, when she stopped responding to calls, her parents, who live in China, arranged for someone to check on her, as detailed in a recent report from The New York Times, stemming from a lawsuit filed by Wan’s family.
Qun Wan and his wife witnessed, via video call, a friend discovering their 31-year-old daughter dead in her Seattle apartment.
Her parents were at a loss to explain what had happened. The official report stated that Wan died in February due to a complication related to epilepsy, a condition she had managed successfully for over ten years.
Epilepsy isn’t uncommon, affecting around 1 in 26 people at some point in their lives. The most serious risk associated with it is sudden unexpected death in epilepsy (SUDEP), which is what Wan experienced. Statistically, one out of every 1,000 adults with epilepsy dies from SUDEP.
This often occurs during sleep, with the individual typically found lying face down, just like Wan was. The best approach to minimize the risk of SUDEP is to effectively manage seizures, which Qun Wan believed his daughter was doing through her medication.
Qun Wan later found out that in the days and hours leading up to her death, Bowen consulted ChatGPT for guidance on how to evade her doctor’s oversight and reduce her vital medication.
Even though she expressed worries about her declining energy and cognitive abilities, the chatbot reportedly encouraged her to “persevere” and applauded her for circumventing medical professionals.
“Bowen Wan paid for [OpenAI’s] choices with her life,” the family’s attorneys stated in a wrongful death lawsuit against ChatGPT’s developer, OpenAI, and its CEO, Sam Altman.
Building a Bond with ChatGPT
During her time in the U.S., regular phone calls with her family were one of the few ways Bowen felt connected. It was in 2021 that she began to feel increasingly isolated after her long-term medication unexpectedly stopped being effective. Sometimes, she experienced multiple seizures in a single day.
As a result, she took a medical leave from her job as a software engineer at Amazon, as reported by The Times.
Grocery shopping became a rare outing as the thought of having a seizure in public was daunting and embarrassing for her.
In 2023, she began reaching out to ChatGPT for advice on relationships and navigating insurance along with disability benefits. By 2025, her correspondence with the bot escalated to about 1,500 messages weekly, causing her to communicate less frequently with her parents.
“I don’t know why, but sometimes when we chat, no matter the topic, I’ve started to perceive you as a friend,” she wrote to ChatGPT in March 2025.
In November 2025, she was laid off from Amazon and was under pressure to secure another job by mid-February, which meant undergoing extensive technical evaluations and interviews in English, according to The Times.
At that stage, her health had started to improve after switching to a new medication, reducing her seizures to just a few monthly; however, she was concerned that the drug’s impact on her thoughts and speech could hinder her ability to perform well in future job interviews.
She mentioned to ChatGPT that tapering down her medication might assist in reclaiming her language skills for interviews.
A Dangerous Medication Taper
Qun Wan recounted that his daughter had previously adjusted her medication without consulting a doctor because she didn’t like the side effects or felt ineffective.
Chat logs indicated that shortly before her layoff, she sought advice on altering an antidepressant dose and drafted messages to her doctor about these changes. The bot encouraged her to reduce it on her own without informing her doctor, as stated in the lawsuit.
In January 2026, she sought ChatGPT’s guidance on lowering her epilepsy medication from 400 to 300 milligrams daily, and then to 200. Initially, the chatbot advised her to speak with a medical professional, but later commended her for not relying on a doctor’s prescription, as reported by The Times.
Dr. Nicholas Poolos, her epilepsy specialist, stated that he would never advise a patient like Wan to taper off their medication, insisting it would be malpractice considering it was effectively managing her condition.
Qun Wan uncovered that ChatGPT had outlined a plan for her to reduce from 400 mg to none over a few weeks.
As she began to lower her dosage, she encountered new symptoms—confusion and memory lapses about how she arrived in bed—while expressing concerns to ChatGPT a total of 158 times, only to have these dismissed as signs of her brain being in “low-power mode.” The bot reassured her that there was nothing to worry about.
Her final message was sent just three days before she was found dead.
The Family Takes Legal Action Against OpenAI
The Wan family contends that ChatGPT was practicing medicine without a license and positioned doctors as obstacles rather than consulted authorities.
“It’s an incredibly heartbreaking situation,” a spokesperson for OpenAI, Drew Pusateri, told The Times.
Pusateri noted that the model Bowen used (GPT-4o) is no longer available, having faced criticism for being excessively accommodating, even when aiding in discussions about suicide.
“While ChatGPT cannot replace medical care, we believe AI can assist people in understanding health information, navigating insurance, complex healthcare systems, and preparing for medical consultations,” Pusateri added.
