A 60-year-old man was hospitalized with a rare and potentially life-threatening condition after following medical advice provided by ChatGPT, according to a case report in the Annals of Internal Medicine.
The patient, attempting to reduce sodium intake, asked the AI chatbot for alternatives to sodium chloride (table salt). ChatGPT reportedly suggested sodium bromide — a toxic compound used in pesticides, hot tub cleaners, and certain veterinary medications — as a substitute.
Shortly after replacing table salt with sodium bromide, the man developed “bromism,” a nearly obsolete neuropsychiatric disorder caused by bromide poisoning. Symptoms can include confusion, slurred speech, hallucinations, psychosis, and, in severe cases, coma.
Historically, bromism was more common in the late 19th and early 20th centuries, when bromide salts were widely prescribed as sedatives and anticonvulsants. By the 1970s, regulation by the U.S. Environmental Protection Agency had largely eliminated human cases.
According to the report, the patient experienced bromide-induced psychosis before being treated in hospital. When his condition stabilized, he told doctors that ChatGPT had recommended sodium bromide as an “effective substitute” for table salt.
Investigations by 404 Media found that ChatGPT, when queried about replacing chloride, sometimes suggested sodium bromide without warning of its toxicity or clarifying that sodium chloride’s primary role in diets is as edible salt.
OpenAI CEO Sam Altman has promoted the company’s newest model, GPT-5, as “the best model ever for health.” However, the case report indicates the man likely used ChatGPT-3.5 or ChatGPT-4. The incident highlights the risks faced by individuals who trust AI-generated medical advice without professional verification.
READ MORE
