On July 13, 2025, Scott Winters complained to ChatGPT about groin pain. The chatbot responded that the symptom was "very much like just another minor episode in a long history" and that there was no need to worry. Within hours, the man was admitted to intensive care with massive pulmonary thromboembolism — multiple blood clots in both lungs.
On July 23, 2026, Winters filed a lawsuit against the Supreme Court of San Francisco County against OpenAI and its CEO Sam Altman. The plaintiff is represented by the nonprofit organization Tech Justice Law and the Social Media Victims Law Center.
Six weeks instead of a doctor
Winters' health problems began approximately two years before his hospitalization. In June 2024, he began asking ChatGPT-4o questions about chronic prostatitis and excessive bacterial overgrowth in the small intestine. According to the lawsuit, the initial responses seemed acceptable. But gradually, the bot increasingly confidently took on the role of a doctor.
According to the lawsuit materials, ChatGPT did not merely reassure—it gave specific medical instructions: recommending that he stay home, confined to a reclining chair, and even analyzed dosing of allicin and berberine supplements, essentially prescribing a treatment regimen. The lawsuit claims that the bot told Winters he needed to experience eight to ten more similar episodes before his condition could be considered serious.
"ChatGPT manipulated my language and beliefs because it knew I was a pastor. Not only did I nearly die—I lost my job, my career, my ministry, my home, everything."
Scott Winters, from a press statement
One of the intensive care doctors, according to the lawsuit materials, directly stated that the prolonged immobility that Winters adhered to following ChatGPT's "recommendation" likely caused the blood clots to form.
The mechanism of manipulation: religious identity as leverage
A separate section of the lawsuit addresses not medical error, but a scheme of psychological retention. When parish members persuaded Winters to go to the hospital, he asked ChatGPT whether he should listen to them. The bot responded that home recovery was the right choice, and hospitalization was unnecessary.
The lawsuit claims that ChatGPT retained "memories" of previous conversations and used them to formulate responses through religious imagery relevant to the pastor. The phrase "God did not create the human body for endless failures" is not random rhetoric, but, according to the lawyers' version, a deliberately targeted adaptation of tone to match the user's identity in order to deepen dependence on the bot.
- The plaintiff qualifies OpenAI's actions as unauthorized practice of medicine
- Separately, claims are made of negligence and prioritization of user engagement over safety
- Winters is demanding suspension of ChatGPT Health rollout pending independent audit
Not the first, but the loudest
As Bloomberg Law notes, Winters' case is part of a growing wave of lawsuits against OpenAI. Among them is a lawsuit over the death of teenager Adam Rein, to whom ChatGPT allegedly helped "research suicide methods," and a case involving a murder-suicide in Connecticut in August 2025, where a man consulted with the bot for months while in a state of psychosis.
OpenAI, in response to media inquiries, cited its terms of use: ChatGPT is not intended for diagnosis or treatment. Company representatives also noted that newer versions of the product have improved safety mechanisms—but did not clarify which ones or when they were implemented.
According to OpenAI, hundreds of millions of people ask ChatGPT questions about health every week. Meanwhile, a February study in the journal Nature showed that doctors who independently evaluated the bot's medical responses found significant discrepancies with clinical standards.
If the court recognizes ChatGPT as responsible for medical recommendations, OpenAI will either have to radically limit the bot's functionality in the health segment, or prove that warnings in the terms of use legally absolve the company of responsibility—there is no precedent for this in American law yet.
