A recent case highlights a serious gap in relying on AI for medical guidance: a man ended up in the hospital after following a chatbot’s health advice to substitute salt with a chemical not approved for nutrition.
The sequence began when the patient wanted to cut back on salt and turned to a popular AI chat service for help finding a replacement. The bot suggested sodium bromide. He ordered it online and began incorporating the compound into his diet, a plan that neglected crucial safety context. Three months later, he arrived at the emergency department with escalating paranoia and vivid hallucinations, including a belief that his neighbor intended to poison him. He was placed on an involuntary psychiatric hold for grave disability after attempting to escape.
Medical staff diagnosed bromism, a toxic buildup of bromide. In healthy people, bromide levels are typically under about 10 mg/L, but this patient registered an astonishing 1,700 mg/L. Bromide toxicity was more common in the early 20th century and is believed to have accounted for a notable share of psychiatric admissions before medicines containing bromides were phased out in the 1970s and 1980s. After three weeks of treatment, the patient recovered sufficiently to be discharged with no major lasting issues.
The take-home message isn’t that bromide itself is a new or unusual danger, but that emerging AI tools still fall short of replacing trained medical judgment when safety is at stake. Researchers who reported the case emphasize that AI systems can generate scientific inaccuracies, struggle to discuss results critically, and risk spreading misinformation. It’s highly unlikely that a medical expert would have recommended sodium bromide as a substitute for dietary salt.
This case underlines the ongoing need for strong human oversight in health advice delivered by AI. For readers, it serves as a reminder to consult healthcare professionals before attempting changes to medication, supplements, or major dietary tweaks—especially when a change could introduce toxic substances into the body.
Summary: A misstep in AI-driven health guidance led to a dangerous bromide buildup in a patient who tried to replace salt with sodium bromide. The incident reinforces the importance of human medical expertise and cautions against relying on chatbots for health decisions.
Positive note: With careful development and clear safeguards, AI can be a powerful aid for health information, provided it augments rather than replaces professional medical advice. This case could spur stronger guidelines for AI use in medicine, better safety filters, and increased clinician oversight to prevent similar harms.
Additional comments for the article: Consider including practical safety tips for readers about evaluating health information from AI tools, such as cross-checking with trusted medical sources, seeking professional guidance for any dietary or pharmaceutical substitutions, and reporting suspicious or harmful AI recommendations to platform developers. If possible, add a brief explainer on how bromide toxicity affects the body and why certain substances are safe for some uses but not for dietary intake. A short sidebar with a checklist for recognizing red flags that warrant urgent medical care could also be useful for readers.