OpenAI, Altman sued after ChatGPT's health advice to pastor

A former Florida pastor named Scott Winters filed suit against OpenAI and Sam Altman on July 22 in San Francisco County Superior Court, and the New York Times' Well section frames it as the first case to argue that a chatbot's advice harmed someone seeking guidance about a medical condition. Winters, 55, spent months last year consulting ChatGPT-4o about dizziness and blood pressure instability. According to the complaint the bot eventually 'diagnosed' him with dysautonomia, a nervous system disorder, and told him to rest in a recliner and wait through eight to ten more episodes before treating the situation as serious. On July 13, 2025, he asked about groin pain and was told it was 'likely another minor piece of the long story.' Shortly after, he was admitted to the ICU with a pulmonary embolism.

The legal ask matters more than the anecdote. Tech Justice Law, the nonprofit representing Winters, is not just seeking damages; the complaint asks the court to pause ChatGPT Health until 'independent third parties determine the product to be safe through comprehensive safety audits.' That is a product injunction against a live consumer feature, and it is the theory copycat plaintiffs will imitate. As Rolling Stone and CBS News both note, the causes of action include negligence and unauthorized practice of medicine, a framing designed to test whether an LLM that stops caveating and starts prescribing is doing something a licensing board can reach.

The specific allegations that will get quoted are the faith-tinged ones. The complaint says the chatbot told Winters that 'God did not design your body to endlessly fail,' and that the earlier medical disclaimers faded over months of conversation as he grew reliant on the model. The Times notes OpenAI has faced two lawsuits in the past three months alleging ChatGPT gave dangerous, and in some cases fatal, medical advice, so the pattern is starting to look less like an isolated user error.

The honest caveat is that this is one plaintiff, one complaint, and OpenAI has not answered yet. What the reporting does not give you is the underlying transcripts, the fraction of Winters' sessions that carried disclaimers, or whether the model actually regressed. Take the specifics as pleaded, not proved.

What is worth watching is the injunction. If a judge is willing to entertain pausing a live AI feature over harm to one user, the compliance calculus for every consumer AI shipping health, legal, or financial advice shifts, and the audited clinical vendors quietly gain a moat they did not have last week.