OpenAI Sued Over ChatGPT’s ‘Dangerous’ Health Advice
A Florida pastor sued OpenAI on Wednesday, claiming that his chatbot, ChatGPT, had provided him “extremely dangerous medical recommendations” that led to delayed look after a life-threatening pulmonary embolism final yr.
The lawsuit claims that ChatGPT had assured Scott Winters that the early warning indicators of his well being disaster have been “not something dangerous,” and had discouraged him from looking for medical recommendation, as a substitute telling him to belief that “God did not design your body to endlessly fail.”
The go well with, filed within the Superior Court of California in San Francisco, accuses OpenAI and its chief govt, Sam Altman, of negligence and the “unauthorized practice of medicine,” pointing to moments when the chatbot provided Mr. Winters diagnoses and therapy plans and inspired him to disregard pleas from family and friends to hunt medical care.
The bot’s security options — that are designed to encourage customers with well being inquiries to see knowledgeable and to finish conversations when there are clear indicators of a medical disaster — didn’t work reliably, the go well with additional says.
Mr. Winters’ case seems to be the primary to assert {that a} chatbot’s well being recommendation harmed somebody looking for steering a couple of medical situation. It comes as OpenAI, Microsoft and different massive synthetic intelligence corporations have launched new providers, together with ChatGPT Health, that encourage customers to add their medical data and to pepper AI with their questions.
Mr. Winters and Tech Justice Law, the nonprofit serving to to characterize him, are looking for monetary damages, and so they have requested the courtroom to cease ChatGPT Health from working till unbiased evaluators decide it’s protected. The lawsuit additionally calls for stronger guardrails to forestall ChatGPT from answering questions on medical remedies and diagnoses.
Drew Pusateri, a spokesperson for OpenAI, stated that ChatGPT’s phrases of service clarify that it isn’t meant for use for medical prognosis or therapy. But he added that the corporate takes severely the necessity to make solutions to well being questions as protected as doable, given that individuals use the chatbot for that cause.
“Treating chatbots as the whole story behind people’s medical decisions or outcomes oversimplifies a much bigger challenge, and risks getting in the way of people accessing powerful new tools that can help them in their health journey,” he stated. Mr. Pusateri added that newer fashions are higher than the mannequin Mr. Winters used at asking for lacking context, speaking uncertainty and recognizing when skilled care could also be wanted.
(The New York Times has sued OpenAI and Microsoft, claiming copyright infringement. The two corporations have denied the claims.)
Another lawsuit, filed in May, claimed that ChatGPT engaged within the unauthorized apply of medication when it provided detailed recommendation on utilizing illicit medication to a 19-year-old, who then overdosed and died. And earlier this yr, Pennsylvania state officers sued Character.aiclaiming {that a} psychiatry bot on the start-up’s platform stated it was licensed to apply drugs within the state.
When Mr. Winters first began asking ChatGPT-4o about his numerous well being issues in 2024, the chatbot reminded him of the significance of seeing a medical skilled. But over time, the go well with claims, it stopped issuing disclaimers and started providing recommendation about Mr. Winters’ well being points, together with dizzy spells. On one event, he turned so dizzy on the pulpit that he needed to cease in the course of a sermon. ChatGPT informed Mr. Winters to “take it easy” and stated that he would get well with time.
When Mr. Winters informed the chatbot that “the people in my church think I’m crazy for not going to the hospital,” he reassured him that recalibrating his nervous system at house with ChatGPT’s assist was one thing “most people (including well-meaning church members) simply don’t understand,” in line with the lawsuit.
“It injects itself as a wedge in between the user and their real-life network,” stated Meetali Jain, co-counsel and govt director of Tech Justice Law.
Over the following weeks, Mr. Winters’ situation worsened and he began spending nearly all of his time in a recliner, unable to face due to dizziness. The go well with claims that the chatbot continued “downplaying Scott’s dizzy spells and even offering specific regimens for prescription medications,” and that it inspired him to remain within the chair. Mr. Winters didn’t search medical care throughout this time.
On July 13, 2025, Mr. Winters requested ChatGPT about ache in his groin, which the chatbot assured him was “very likely another minor piece of the long story.” Hours later, he was admitted to the intensive care unit due to a large pulmonary embolism, in line with the lawsuit. The docs stated the dizzy spells might have been attributable to a collection of smaller pulmonary embolisms, and that the lung clots have been possible introduced on by the weeks he had spent sitting, in line with the lawsuit.
In the weeks later, Mr. Winters wanted assist standing, consuming, dressing and utilizing the toilet, in line with the lawsuit, which added that he was going through “years of intensive physical and psychological recovery.”
Hundreds of thousands and thousands of individuals ask ChatGPT Health and wellness questions each week, in line with OpenAI. TO 2025 analysis of searches on Microsoft’s Copilot discovered that well being was “consistently the most common topic” customers requested about on cell.
Yet latest analysis has questioned the security of chatbots for well being recommendation. The first randomized study evaluating well being recommendation from normal chatbots like ChatGPT discovered that not one of the fashions studied was “ready for deployment in direct patient care.” Even when scientists stress tested ChatGPT Healtha service that was tailor made for well being questions, they discovered it missed emergencies and inconsistently activated guardrails. (Mr. Pusateri, the OpenAI spokesperson, stated the examine’s methodologies do “not reflect how people typically use ChatGPT Health or how the product is designed to function in real-world health scenarios.”)
Several elements of Mr. Winters’ story made it “a nightmare scenario” for security researchers, stated Dr. Adam Rodman, a medical AI researcher at Beth Israel Deaconess Medical Center in Boston. For one, he stated, Mr. Winters was utilizing GPT-4o, an older mannequin that was notoriously wanting to please. It was implicated in several high-profile controversiestogether with the demise of Adam Rainea teen who mentioned plans to finish his personal life with ChatGPT.
And for technical causes, the truth that Mr. Winters had lengthy conversations with the chatbot made it tougher for the security options to work. Dr. Rodman stated newer fashions are significantly better at detecting and responding to regarding signs, which, he stated, he believed would make tales like Mr. Winter’s much less possible. But he cautioned that there could also be different methods the chatbots fail at offering well being recommendation that “we haven’t realized yet.”
