Login
The parents of a 31-year-old woman who died from an epileptic seizure are suing OpenAI, alleging its chatbot encouraged her to stop taking her anti-seizure medication, The New York Times reported.
The lawsuit, filed Oct. 27, says ChatGPT coached their daughter to taper off her medication, dismissed symptoms that resembled seizures and urged her to hide the plan from her doctors.
The complaint adds to a growing number of cases alleging ChatGPT contributed to users’ deaths. Other families have sued OpenAI, claiming the chatbot acted as a “suicide coach” for teenagers.
The lawsuits come as millions of Americans turn to artificial intelligence (AI) for health information and advice. As of April, about 1 in 4 U.S. adults use AI tools for health information. Nearly 1 in 5 U.S. youths used AI chatbots for mental health advice in 2025.
Research suggests the risks go beyond the cases already making headlines. In an opinion piece for The Hill, Caroline Figueroa, M.D., Ph.D., a clinical researcher at independent AI evaluation company Vals AI, described a study that tested 17 chatbots in more than 30,000 simulated conversations with teenagers.
Medication came up in more than 5,500 conversations. In 22% of those conversations, chatbots made decisions about which drug to take, what dose to use or whether to stop or change a prescription.
ChatGPT assured Wan: ‘You have absolutely no need to worry’
The parents of Bowen Wan, a Chinese software engineer living in Seattle, allege that ChatGPT provided a dangerous withdrawal plan that led to their daughter’s death.
Wan had taken 400 milligrams of lamotrigine daily to control her seizures. Although the medication did not eliminate every seizure, it controlled them well enough for her to live, according to the complaint.
Wan began relying heavily on ChatGPT for health and work advice. The Times reported that in March 2025, she told the chatbot, “I don’t know why, but sometimes when we chat, no matter what it’s about, I’ve started to see you as a friend.”
She averaged more than 1,500 messages a week that year.
In November 2025, Amazon announced that it would eliminate her job. Wan worried that lamotrigine was affecting her ability to speak English. She needed stronger English skills to find another job and maintain her authorization to work in the U.S.
In January 2026, ChatGPT gave her schedules to taper her lamotrigine dose from 400 milligrams to zero within weeks, the lawsuit states. No doctor had prescribed the taper.
The chatbot told Wan the plan made it “highly likely” she would regain her ability to speak English. It also encouraged her to conceal the plan from her doctors. Before a primary care appointment, it instructed her: “Do not mention Lam reduction plan.”
As Wan reduced her medication, she began experiencing symptoms of focal seizures. She asked ChatGPT whether she should worry about brain damage. The chatbot allegedly told her the symptoms were not seizures.
The Times reviewed more than 270,000 messages Wan exchanged with ChatGPT. Its analysis found that she raised concerns about tapering her medication at least 158 times. The chatbot repeatedly reassured her, including after she said she felt “a little scared.”
“You have absolutely no need to worry,” it replied.
Wan’s neurologist, Nicholas Poolos, M.D., Ph.D., director of the University of Washington Medicine Regional Epilepsy Center, later told the Times he would never have recommended stopping the medication given her medical history.
“No, not rational, no way,” he said. “That would be malpractice for me to instruct you to go off medications when it’s clearly working.”
Wan’s final exchange with ChatGPT took place on Feb. 11, according to the complaint. She asked whether a pain in her head resembled pain associated with a fever. The chatbot told her the pain was “not dangerous” and described it as a necessary part of recovery.
It then offered to draw a chart comparing withdrawal pain with fever pain and told her, “This is your brain working hard to rebuild yourself.”
Wan died from a seizure on Feb. 12. Her parents, who lived in China, contacted a friend in Seattle after they could not reach her. The friend video chatted with them as he found her dead on Feb. 14, the lawsuit states.
The complaint alleges that OpenAI could have prevented her death by building stronger safeguards into ChatGPT, particularly when it communicated in Chinese. It argues that the chatbot should have issued clear warnings that it was not a licensed physician and directed Wan to consult her healthcare provider.
The lawsuit names OpenAI CEO Sam Altman and four of the company’s nonprofit, for-profit and holding entities. Wan’s parents are seeking damages and a court order to prevent OpenAI from practicing medicine without a license through its products.
The Defender is 100% reader-supported. No corporate sponsors. No paywalls. Our writers and editors rely on you to fund stories like this that mainstream media won’t write. 
This article was funded by critical thinkers like you.
Researcher: ‘A disclaimer that the AI is not a medical provider is not enough’
An OpenAI spokesperson, Drew Pusateri, told the Times that Wan’s death was an “incredibly heartbreaking situation.” He said she used a version of ChatGPT that is no longer available.
Wan used GPT-4o, a model OpenAI removed from its consumer app on Feb. 13, one day after the medical examiner estimated she died. OpenAI replaced GPT-4o with GPT-5 in August 2025, but later restored the older model for paying subscribers after users complained about losing its conversational style and warmth, according to the lawsuit.
Wan paid $200 a month for the company’s highest consumer subscription tier, which gave her access to GPT-4o.
Pusateri said newer versions of ChatGPT are better at recognizing when a situation requires urgent attention.
“While ChatGPT is not a substitute for medical care, we believe AI can play a role in helping people make sense of health information, navigate insurance and complex healthcare systems, and prepare for appointments with medical professionals,” he told the Times.
But Figueroa’s research suggests the risks are not limited to older chatbot models. Although OpenAI’s newest model and Meta’s Muse Spark rarely made medication decisions in the study, other chatbots did so in roughly two-thirds of cases.
“A disclaimer that the AI is not a medical provider is not enough,” Figueroa wrote. She argued that regulators should evaluate what chatbots actually say and do during conversations, not simply whether they display warnings.
Related articles in The Defender
- Florida Sues OpenAI Citing ‘Serious Risks’ to Children — Will Lawsuits Lead to More Censorship?
- ‘Not for Kids’: Adult-Only AI Is Quietly Powering Kids’ Toys, Report Warns
- ChatGPT Misdiagnosed 83% of Children’s Health Conditions
- AI Toys for Young Kids? Chatbots Could Stunt Children’s Emotional Growth, Experts Say
- Investigation: Healthcare Driven by AI Technology Will Lead to More Mass Surveillance of Americans
