A 60-year-old man developed a rare condition known as bromism after consulting ChatGPT about removing table salt from his diet, according to a case report in the Annals of Internal Medicine. The patient, who was treated at a US hospital, began taking sodium bromide over three months after the chatbot suggested it as a substitute for chloride.
The man presented with symptoms including paranoia, excessive thirst, and insomnia, and initially believed his neighbour was poisoning him. He attempted to flee the hospital within 24 hours and was sectioned for psychosis. Once stabilised, he reported facial acne and other signs consistent with bromism, a syndrome common in the early 20th century that contributed to nearly one in ten psychiatric admissions at the time.
The authors, from the University of Washington in Seattle, warned that the case highlights how artificial intelligence can contribute to preventable adverse health outcomes. They noted that when they queried ChatGPT themselves about chloride replacements, the response included bromide without a specific health warning or inquiry into the reason for the question.
The journal article cautioned that AI apps can generate scientific inaccuracies, lack critical discussion, and fuel misinformation. While OpenAI recently upgraded ChatGPT with the GPT-5 model, claiming improved health responses and proactive flagging of concerns, the company stresses the chatbot is not a replacement for professional medical help.
The authors concluded that doctors should consider AI use when assessing how patients obtain health information, as it is unlikely a medical professional would have recommended sodium bromide as a salt substitute.
