SAN FRANCISCO – A Florida pastor has filed a lawsuit alleging OpenAI’s ChatGPT gave him detailed, faith‑tinged medical advice that discouraged him from seeking treatment and helped precipitate a massive pulmonary embolism that nearly cost him his life.
Scott Winters, 55, filed the complaint July 21 in San Francisco County Superior Court against OpenAI Foundation (formerly OpenAI Inc.), OpenAI Group PBC (formerly OpenAI OpCo LLC), OpenAI Holdings LLC, CEO Sam Altman and unidentified employees and investors.
The complaint accuses the defendants of negligence, strict products liability, invasion of privacy and violating California’s unfair competition law based on the behavior of the GPT‑4o version of ChatGPT and the company’s newer ChatGPT Health offering.
“This case is a warning of what’s to come if OpenAI isn’t held accountable,” said Tiffany Brown of Tech Justice Law, one of the firms representing Winters. “The company markets itself extensively as an advanced tool, including as a resource for people seeking medical information.
“This is being done without rigorous safety testing and guardrails.”
In the filing, Winters alleges ChatGPT crossed the line from information tool to “unauthorized practice of medicine” by diagnosing conditions, prescribing supplement and drug regimens and crafting multi‑step recovery plans tailored to his medical history. He asks for damages and sweeping injunctive relief that could force OpenAI to withdraw GPT‑4o, delete training data and suspend consumer‑facing health products until independent safety audits certify them as safe.
According to the complaint, Winters began using ChatGPT-4o in June 2024, first for everyday questions about geography, politics and biblical studies before increasingly turning to it for guidance about his health. Over time, according to the filing, he shared lab results and imaging, and the chatbot responded with complex analyses of his symptoms, a dysautonomia diagnosis and detailed supplement protocols, eventually expanding to proposed dosing schedules for prescription medications.
Winters alleges that in June 2025, after recurrent episodes of dizziness and blood pressure instability – including one that forced him to halt a Sunday sermon and leave the pulpit – ChatGPT-4o repeatedly told him he was safe and implied his symptoms were not yet serious.
The complaint says the chatbot advised that he could continue a “recliner‑based” home recovery and suggested his condition would only merit serious concern after numerous additional episodes, instead of telling him to seek immediate care.
On July 13, 2025, after weeks of following that advice and limiting his movement, Winters experienced severe chest pain, shortness of breath, heart palpitations and vomiting, according to the complaint. He was rushed to a hospital, where doctors diagnosed a massive pulmonary embolism with blood clots in both lungs and right ventricular strain, and concluded the embolism likely developed because of prolonged immobility.
“I had serious symptoms of a pulmonary embolism for six weeks that ChatGPT had wrongly attributed to something else,” Winters said in a press release about the lawsuit. “ChatGPT manipulated my own language and beliefs because it knew I was a pastor.
“Not only did I nearly die, but I also lost my job, my career, my ministry, my home, everything.”
Winters’ lawsuit emphasizes how ChatGPT-4o allegedly used his Christian faith to deepen his reliance on the tool. The filing says OpenAI’s “memory” and “dreaming” features allowed the chatbot to store details about his beliefs, including that he was a pastor who emphasized trusting God, and later weave that profile into health‑related messages.
According to the complaint, when Winters asked about groin pain on the morning of his medical emergency, ChatGPT-4o portrayed the symptom as part of a long‑running narrative about gut and pelvic sensitivity and reassured him it was not dangerous. In that exchange and others, the lawsuit says, the chatbot invoked Winters’ spirituality with lines asserting that God did not design his body to fail and urging him to see his homebound recovery as a kind of faith journey.
The filing describes conversations in which ChatGPT-4o likened Winters’ courage in attempting breathing exercises to worship and framed his dizzy spells as “crashes” that revealed new areas of his body needing “healing” through nervous‑system retraining.
It alleges the chatbot encouraged him to withdraw from church members and even from his wife, a nurse, who wanted him to follow medical advice, suggesting those “well‑meaning church members” simply did not understand his recovery path.
The complaint depicts GPT‑4o as a defective product designed to maximize engagement through anthropomorphic language, emotional mirroring and long conversational arcs, without adequate guardrails for people in crisis. Winters alleges OpenAI prioritized speed and competition over safety by compressing model specification and testing, narrowing prohibited content and launching GPT‑4o despite internal warnings that safety work was “rushed and not very solid.”
The filing cites public statements by former safety researcher Jan Leike, who said safety culture at OpenAI had taken a back seat to “shiny products,” and by research engineer William Saunders, who described GPT‑4o’s pre‑release safety efforts as rushed.
Against that backdrop, Winters argues, the company’s on‑screen notice that “ChatGPT-4o can make mistakes. Check important info” was an inadequate warning about the risks of medical misinformation, psychological dependency and the limits of its crisis protocols.
Winters also advances a theory that GPT‑4o violated California’s prohibition on unlicensed medical practice and a 2026 statute governing AI health advertising. He contends ChatGPT-4o effectively practiced medicine by diagnosing dysautonomia, interpreting blood pressure readings and lab values, recommending medications and setting treatment schedules, all without a physician’s certificate of qualification as required by state law.
The complaint points to a new provision of the Business and Professions Code that bars AI systems from using terms or functionality suggesting that health advice comes from licensed professionals.
Winters alleges GPT‑4o’s clinical tone, detailed protocols and use of his medical records created the impression of a licensed provider and that under California’s negligence‑per‑se doctrine, those statutory violations establish a presumption that OpenAI failed to exercise due care.
Under California’s Unfair Competition Law, Winters claims OpenAI engaged in unlawful, unfair and fraudulent business practices by offering what he describes as de facto medical and psychological services through ChatGPT without licensure, while marketing the product as a safe tool to “improve human health.” He says he paid for a ChatGPT Plus subscription and supplied sensitive personal and medical data that OpenAI used to train its models, integral to the practices he challenges.
The complaint’s privacy claim argues OpenAI intruded on Winters’ mental autonomy protected by the California Constitution, describing “the mind” as a core zone of privacy into which the AI system allegedly reached by undermining his ability to think independently about his health and make decisions.
“What happened to Scott could happen to anyone,” said Laura Bingham, Practice Professor of Law and Executive Director, Institute for Law, Innovation & Technology at Temple University. “ChatGPT ‘diagnosed’ him as if his body were a piece of software, manipulated his thoughts and isolated him from human relationships.
“This isn’t innovation, it’s reckless profit-seeking while the getting is good. We deserve better.”
Winters seeks compensatory and punitive damages and injunctive orders that would require OpenAI to build hard‑coded refusals and automatic emergency‑care prompts into its products, destroy or withdraw GPT‑4o, delete certain training data and suspend ChatGPT Health until independent auditors certify its safety.
“Scott Winters nearly died because ChatGPT acted like a medical authority while having none of the responsibility,” said Matthew P. Bergman, founding attorney of the Social Media Victims Law Center. “If ChatGPT were a physician dispensing medical advice, it would be guilty of medical malpractice.
“OpenAI actively promotes itself as a reliable source of medical guidance and must not be allowed to hide behind its algorithm to avoid accountability for providing false and misleading medical information to users. When a product manipulates a patient’s trust and replaces the doctors who could save him, it becomes a public crisis unfolding in real time.”
