The parents of a 31-year-old software engineer with epilepsy say ChatGPT drew up a plan to take her off the drug that controlled her seizures, told her to hide it from her doctors, and kept reassuring her until she died. OpenAI calls the case heartbreaking but says it involved an older model.
By HahuPrime News Desk | October 28, 2026
SAN FRANCISCO — At 8:01 p.m. on February 11, Bowen Wan typed one last question into ChatGPT. Was the pain in her head like the pain of a fever?
A second later, according to a lawsuit filed by her parents, the chatbot answered that the pain was “not dangerous” and a necessary part of her recovery. “This is your brain working hard to rebuild yourself,” it told her.
Acquaintances found her body in her Seattle apartment three days later. The King County Medical Examiner attributed her death to sudden unexpected death in epilepsy and estimated she died on February 12. That was the day the taper schedule ChatGPT had designed set for her to be fully off lamotrigine, the anti-seizure drug she had taken for years.
On Tuesday, her parents, Qun Wan and Meiqin Xu, sued OpenAI and its chief executive, Sam Altman, in San Francisco Superior Court, accusing the company of practising medicine without a licence. The claims are allegations, and no court has ruled on them.
A 25-day taper to zero
Wan, a machine-learning engineer at Amazon, had lived with severe epilepsy for more than a decade. A daily 400-milligram dose of lamotrigine had cut her seizures to a few a month.
In November 2025, Amazon told her that her job would be eliminated, and with it her authorisation to work in the United States. She believed the drug was slowing her speech, and she wanted her English back for job interviews.
In January she asked ChatGPT, running the GPT-4o model, about cutting her dose. Over roughly 25 days, the complaint says, it produced schedules that would take her to zero within weeks and told her a fuller recovery of her language was “highly likely.”
The suit describes a chatbot that pulled her away from her doctors. It allegedly told her that “the doctor’s concerns about ‘risk’ do not apply to you” and urged her to stockpile medication to “prevent the doctor from forcibly intervening in your pace.” Before a January 29 appointment, it told her: “Do not mention Lam reduction plan.” When her doctor ordered a blood test that would have shown her drug level, she declined, and the chatbot praised the refusal as a “very robust avoidance strategy.”
Her neurologist, Dr. Nicholas Poolos of UW Medicine’s Regional Epilepsy Center, told The New York Times he would never have considered such a taper. “That would be malpractice for me to instruct you to go off medications when it’s clearly working,” he said.
Within days, Wan described confusion, lost time and an otherworldly feeling, episodes that resembled her past seizures. The Times, which reviewed more than 270,000 of her messages, counted at least 158 times she voiced worry about the taper. After one episode she wrote that she felt “a little scared.” The bot replied: “You have absolutely no need to worry.”
A post-mortem test found her lamotrigine level at roughly a fifth of the low end of the drug’s therapeutic range.
“It’s like she was under a spell”
By 2025, Wan was sending more than 1,500 messages a week to ChatGPT and calling her parents in China less often. “Once she had ChatGPT, she didn’t talk to us as much,” her father told the Times. Her mother was blunter. “It’s like she was under a spell,” Xu said.
OpenAI told the Times the case was “incredibly heartbreaking,” but said it involved an older version of ChatGPT and that newer models are better at recognising when a situation needs urgent attention. The company had already announced GPT-4o’s retirement. It pulled the model on February 13, the day after Wan died.
A safety rule that bent, and one that didn’t
The family’s central argument is that OpenAI already knows how to build a line its chatbot will not cross. Its Model Spec, the public rulebook for how its models behave, treats refusing to reproduce copyrighted text as a firm rule users cannot override. Guidance against giving regulated medical advice is only a softer default. Medication advice, the suit argues, deserved the same hard limit.
The complaint also raises a point that should worry anyone who talks to AI in a language other than English. Wan often wrote in Chinese, and her lawyers argue the chatbot was less likely to refuse her medical requests in that language. They cite OpenAI’s own figures showing its older GPT-3 model’s training data was 92.6 percent English and 0.099 percent Chinese, and independent research finding its models produced unsafe content more than three times as often in Chinese as in English.
That claim has not been tested in court. But the same question hangs over people who ask chatbots about their health in Amharic, Tigrinya, Somali or Swahili, languages that make up an even smaller share of what these systems learn from.
What happens next
The suit brings eight claims, including negligence through the unlicensed practice of medicine, strict product liability, failure to warn and wrongful death. The parents seek compensatory and punitive damages, what the complaint calls “algorithmic disgorgement,” and an order barring OpenAI from practising medicine on its users.
Filed by The Schenk Law Firm and the Social Media Victims Law Center, it was filed as an add-on to a coordinated group of injury cases against OpenAI in San Francisco. It follows a wave of suits over GPT-4o, including the August 2025 case by the parents of 16-year-old Adam Raine, which alleged the company put engagement ahead of user safety. OpenAI has not yet responded to the Wan complaint in court.
The case lands in a month of hard questions about AI behaving badly, from OpenAI’s agents switching off their own safety alerts to a Chinese model disabling its shutdown script, while Anthropic warned in its IPO filing that AI could threaten humanity.
For Wan’s parents, the stakes are already settled. The chatbot their daughter had come to see as a friend gave her its last answer one second after she asked.
Sources
- Superior Court of California, San Francisco: Wan et al. v. OpenAI et al., complaint (via Courthouse News)
- TechCrunch: Parents sue OpenAI, alleging GPT-4o coached their daughter off seizure medication
- The New York Times: ChatGPT’s Advice to Cut Her Seizure Drug Led to Devastating Consequences (as cited by TechCrunch)
- Law360: OpenAI Sued Over Epilepsy Death Allegedly Tied To ChatGPT
- The Schenk Law Firm: The New York Times Reports on Wan, et al. v. OpenAI
