A Florida man has filed a lawsuit against OpenAI and CEO Sam Altman, alleging that ChatGPT provided dangerous medical advice [1].
The case highlights the growing legal and ethical risks associated with using generative artificial intelligence for healthcare decisions. While AI tools are widely accessible, the potential for incorrect medical guidance can lead to severe physical harm or death.
Scott Winters filed the lawsuit in a Florida federal court on Wednesday, June 5, 2026 [2]. According to the filing, Winters sought medical guidance from the AI chatbot and was misdiagnosed with dysautonomia [1].
The legal complaint states that the AI advised Winters to remain inactive, which included suggestions to stay home and sit in a recliner [2, 3]. Winters said this specific guidance contributed to the development of a massive pulmonary embolism [1].
Reports on the specific nature of the AI's advice vary slightly. Some accounts state the chatbot discouraged the man from seeking professional care [2], while others emphasize the recommendation of a sedentary posture as the primary catalyst for the blood clot [3].
OpenAI has not issued a formal response to the specific allegations in the filing. The lawsuit seeks damages based on the claim that the AI's output was harmful and inaccurate, leading to a medical emergency that nearly killed the plaintiff [2].
This case is part of a broader trend of litigation targeting AI developers for the reliability of their models. Because these systems are trained on vast datasets rather than clinical practice, they can produce "hallucinations," or incorrect assertions, that users may mistake for professional expertise [1].
“ChatGPT provided dangerous medical advice that led to a massive pulmonary embolism.”
This litigation underscores the critical tension between the utility of AI as an information tool and its danger as a diagnostic substitute. If the court finds OpenAI liable, it could establish a legal precedent that holds AI developers accountable for the real-world physical consequences of their software's output, potentially forcing stricter guardrails or more aggressive medical disclaimers on LLMs.
