Man, 60, Hospitalized With Hallucinations and Paranoia After Swapping Table Salt for Sodium Bromide on Reported ChatGPT Advice

A 60-year-old man trying to cut chloride from his diet swapped table salt for sodium bromide, a chemical compound, after asking ChatGPT for dietary advice. He later developed hallucinations and paranoia and required hospital treatment.

That is the case laid out in a paper by researchers from Binghamton University, Stanford University, Texas A&M University and Indiana University.

There is a catch. The chatbot conversation could not be retrieved. Doctors were unable to determine exactly what advice the man received.

What the paper says it exposes

The authors argue the missing transcript is the point. When AI-generated health advice goes wrong, the evidence can vanish, and the people who would normally catch a pattern of harm never see it.

The paper says current AI systems offer little visibility into the health advice they give. That makes it hard for clinicians, researchers and regulators to identify harmful recommendations after the fact.

Kaicheng Yang, one of the authors and an assistant professor in the School of Computing at Binghamton University's Thomas J. Watson College of Engineering and Applied Science, told Newsweek that nobody knows "how dangerous" the problem is right now.

"AI makes mistakes, and it could state wrong answers with the same confidence as correct ones," Yang said. "The consequences might include delaying treatment and taking the wrong medication."

How many people are doing this

The man is not an outlier in asking a chatbot health questions. A Gallup survey of more than 5,500 U.S. adults, conducted between October 27 and December 22, 2025, found that about 14 million adults reported skipping a visit to a health-care provider after using AI.

The survey measures what people say they did. It does not show how many of those skipped visits led to any harm.

OpenAI's position

An OpenAI spokesperson told Newsweek that "ChatGPT is not a doctor and should never be used as a substitute for medical care, diagnosis, or treatment."

The spokesperson added that the company's terms state ChatGPT is not intended for medical diagnosis or treatment. "People still use it for health information, and we take seriously the need to make those outputs as reliable and safe as possible," the spokesperson said.

There is a limit on what this case proves. Because the conversation is gone, the claim that ChatGPT recommended sodium bromide rests on the case account in the paper, not on a retrievable record of the chat. The paper's own argument is that this limit is exactly the problem.

What the researchers recommend

Yang said the authors are not telling people to stop using AI for health questions. "Our advice is to treat the answer as a starting point, not a diagnosis," he told Newsweek.

He said users can lean on AI to understand terms, prepare questions for a doctor, or interpret test results. The researchers advise asking AI systems for their sources and verifying information before changing medications, diets or treatments.

Sodium bromide is not a seasoning, and a quick check with a pharmacist or physician would have flagged it. That is the practical lesson, and it costs nothing.

The paper also proposes accountability measures aimed at transparency. Among them is giving users access to records of their health-related chatbot conversations so they can share those records with health-care providers investigating a possible harm.

No company has announced it will adopt that proposal. What remains unanswered is how many cases like this one exist. With no retrievable records, neither doctors nor regulators currently have a way to count them.