A recent medical case has highlighted just how dangerous it can be to take health advice from AI tools like ChatGPT without professional oversight.
A 60-year-old man from New York ended up in the hospital after developing bromism – bromide poisoning – caused by following a dietary suggestion from the chatbot.
According to a report published in the Annals of Internal Medicine and covered by NBC News, the man asked ChatGPT for alternatives to regular table salt. The AI recommended sodium bromide, a chemical once used in sedatives but now largely abandoned due to its toxicity. Believing it was a healthier substitute, the man replaced his salt with sodium bromide purchased online and consumed it daily for three months.
The results were catastrophic. He developed paranoia, insomnia, psychosis, and severe physical symptoms. At one point, he even believed his neighbor was poisoning him. Doctors later diagnosed him with bromism, a condition that is now rare precisely because bromide salts are hardly used in modern medicine. Once he stopped taking the substance and received treatment, his health began to improve.
Researchers tested ChatGPT themselves and found that it did indeed suggest bromide as a substitute for chloride, with no mention of toxicity. This case underscores the limits of AI in healthcare: chatbots cannot replace the judgment, accountability, or safety standards of trained medical professionals. It also raises a bigger question about guardrails and safety warnings in AI systems – especially when people are tempted to treat them as all-knowing experts.
In a world where AI tools are woven into daily life, curiosity and experimentation should never override common sense and medical advice. As this incident shows, not every answer generated by AI is safe to follow – and in matters of health, a single wrong suggestion can be life-threatening.