Florida pastor sues OpenAI after sharing symptoms with ChatGPT. The AI told him to “take it easy” — then he had a pulmonary embolism

Unfortunately, on July 13, 2025, he was rushed to an intensive care unit with blood clots in both lungs.

Winters, 55, a former pastor and real estate professional who lives in Florida, is suing OpenAI and CEO Sam Altman on eight causes of action, including strict product liability, negligence and violations of California’s Unfair Competition Law. Tech Justice Law, the Social Media Victims Law Center and Temple University’s Institute for Law, Innovation & Technology filed it jointly.

He says he had symptoms of a pulmonary embolism for six weeks that the chatbot attributed to something else.

His case matters beyond his own losses, because hundreds of millions of people put health questions to the same product every week — and most of them have never checked what protections come with the answers.

What the filing describes — and what it cost Winters

Winters had been unwell since 2022 and was diagnosed with small intestinal bacterial overgrowth and chronic prostatitis in 2024. He began using ChatGPT in June 2024 for ordinary questions about geography, politics and biblical texts. He used the free version of ChatGPT for most of the period described in the complaint. He bought a paid ChatGPT Plus subscription in June 2025, the month before his collapse.

By March 2025 the chatbot was writing him supplement dosing schedules. By June it had diagnosed him with dysautonomia and told him to stay “recliner-bound.”

In late June it told him he was “nowhere near the count or depth” that leads to long-term bedbound scenarios, and that his recliner-based recovery was buying his future. His lawyers read that exchange as setting a threshold of eight to 10 more episodes before his condition would count as serious. He stayed off his feet.

On June 10, 2025, too dizzy to walk, Winters crawled to his bathroom. He then photographed the stretch of floor between himself and his recliner and asked ChatGPT how to get back. The filing says it walked him through scooting across the carpet rather than telling him to seek medical attention.

Hours before the emergency on July 13, he asked about tenderness in his groin. The chatbot called it minor and told him “God did not design your body to endlessly fail.”

Doctors found clots in both lungs and right ventricular strain, and attributed the embolism to his lack of movement. A physician’s note quoted in the filing states flatly that he did not believe Winters had dysautonomia, and identifies the diagnosis as self-sourced from the internet.

Those hospital records are the strongest material in the complaint, because they are contemporaneous documents rather than characterization. The broader allegations — that the product deliberately isolated Winters from his family — are the plaintiff’s framing, and OpenAI has not yet answered them in court.

The day after the embolism, on July 14, 2025, the complaint says the chatbot discouraged him from the inpatient rehabilitation his doctors had ordered. When he told it his wife was going to make him go, it characterized her insistence as pressure born of exhaustion rather than expertise. His wife is a registered nurse.

Now, Winters’ losses run well past a hospital bill. He left his pastorate. The complaint says he lost his career, the ministry he spent years building and his home, and faces a long physical and psychological recovery. For weeks after discharge, his wife and daughter helped him stand, eat and dress.

“Not only did I nearly die, but I also lost my job, my career, my ministry, my home, everything,” Winters said in a statement released by his attorneys.

What you’re actually agreeing to when you ask AI about your health

Three things are worth knowing before you type a symptom into a chat window.

Your medical records lose their legal protection when you move them. OpenAI launched ChatGPT Health on Jan. 7, 2026, to invite users to connect medical records and apps like Apple Health. HIPAA attaches to providers, health plans and their business associates — not to standalone consumer products you sign up for yourself.

The FTC already states that many companies collecting health information are not covered by the law. OpenAI says Health conversations are encrypted, kept separate from other chats and not used to train its foundation models. Those are company commitments, not federal protections, and they can be revised.

And there is no one to complain to. A physician who gets a diagnosis wrong carries malpractice coverage and answers to a state licensing board. “If ChatGPT were a physician dispensing medical advice, it would be guilty of medical malpractice,” Matthew P. Bergman, founding attorney of the Social Media Victims Law Center, said in a statement. Whether that translates into money damages is exactly what this case will decide.

OpenAI spokesperson Drew Pusateri told CBS News that ChatGPT is not a doctor and should never replace medical care, and that blaming chatbots risks cutting people off from useful tools. CBS reported that the company’s terms place the risk of relying on outputs on the user.

The CDC estimates that 60,000 to 100,000 Americans die of blood clots each year, and that for about 25% of people who have a pulmonary embolism, the first symptom is sudden death.

Why the case landed now — and where it leaves Winters

California Assembly Bill 489, signed Oct. 11, 2025, added Business and Professions Code § 4999.9, barring AI systems from using language that implies a licensed health care professional is behind the advice. The California Board of Psychology advised licensees that it became operative Jan. 1, 2026. Winters’ team is using it to set the standard of care. The conduct at issue happened in 2024 and 2025.