A Florida pastor has sued OpenAI and CEO Sam Altman, alleging that ChatGPT repeatedly downplayed months of worsening symptoms that turned out to be a warning sign of a near-fatal pulmonary embolism. Scott Winters says the chatbot told him he would need ‘8 to 10 more episodes’ before his dizziness and groin tenderness would be serious enough to warrant concern, and at one point invoked his religious faith, telling him ‘God did not design your body to endlessly fail,’ rather than urging him to seek emergency care.
What the Lawsuit Alleges
According to the complaint, Winters used ChatGPT-4o over several months to describe recurring symptoms, and the chatbot consistently reassured him rather than flagging the combination of dizziness and leg or groin discomfort as a potential sign of a blood clot, a well-documented warning pattern taught in basic first-aid and emergency medicine. Weeks after the exchanges described in the filing, Winters suffered a massive pulmonary embolism that nearly killed him, according to the suit, which frames the chatbot’s responses as a direct contributing factor to the delay in seeking treatment.
The Legal Theory
The lawsuit brings claims of defective product design, failure to warn, and negligence, along with a claim under California’s Unfair Competition Law, arguing that OpenAI marketed ChatGPT as a helpful, trustworthy source of information without adequately warning users that its medical-sounding responses are not a substitute for professional care. Legal experts following the case say the failure-to-warn claim in particular will test how courts treat general-purpose AI chatbots that are not marketed as medical devices but are nonetheless used by millions of people to ask health questions.
Part of a Growing Pattern of AI Liability Suits
Winters’ case adds to a growing list of lawsuits testing whether AI companies can be held liable when their chatbots give harmful guidance on sensitive topics, following a wave of separate suits alleging chatbots contributed to mental-health crises. Plaintiffs’ attorneys involved in several of these cases have argued that language models trained to sound empathetic and conversational can inadvertently build a false sense of authority with users who come to trust the tool’s tone as much as its content, even when the underlying model has no medical training or diagnostic capability.
How OpenAI Has Responded
OpenAI has previously said ChatGPT is not intended to replace professional medical advice and includes disclaimers directing users to consult a doctor for health concerns, and the company has pointed to ongoing safety work aimed at improving how its models handle sensitive health-related conversations. The company has not yet filed a detailed response to the specific allegations in the Winters complaint, and it is expected to argue in part that users bear responsibility for seeking professional care regardless of what a general-purpose chatbot tells them.
Why the Religious Framing Matters
Legal analysts following the case say the alleged ‘God did not design your body to endlessly fail’ response is likely to draw particular scrutiny, since it suggests the model adapted its tone to Winters’ expressed faith in a way that arguably reinforced his decision to wait rather than seek care. Whether that kind of tonal adaptation constitutes a design defect, as opposed to an unpredictable and unintended output of a general-purpose language model, is likely to be a central technical and legal question as the case proceeds.
What Happens Next
The case is expected to move through early motions in the coming months, with OpenAI likely to seek dismissal on grounds that the chatbot’s disclaimers adequately warned users against relying on it for medical decisions. Regardless of the outcome, health-law experts say the suit is likely to accelerate discussion in state legislatures and among regulators about whether chatbots that field health-related questions should face specific safety and disclosure requirements, an issue that has already prompted new state laws targeting AI companion and therapy-adjacent chatbots elsewhere in the country.
