AI health guidance under scrutiny

A medical journal warns against using AI for health information after a patient developed bromide toxicity following salt advice given by a chatbot.

AI health guidance under scrutiny after bromide case linked to salt advice

A US medical journal reports a case where a 60-year-old man developed bromide toxicity after seeking advice about removing table salt from his diet. He began taking sodium bromide for three months, based on guidance he read that chloride could be swapped with bromide. The report notes that historical bromide use is linked to a past rise in psychiatric admissions and describes how such AI-suggested substitutions can lead to serious harm. The authors could not access the patient’s ChatGPT conversation log, so the exact advice given remains unknown, but their own testing of the AI showed it suggested bromide when asked about chloride replacement and did not issue a health warning.

The episode highlights a core tension in AI health tools: they can widen access to information but risk spreading decontextualized or incorrect guidance. The case underscores the need for robust safeguards, clear warnings, and human oversight whenever medical questions are involved. It also raises questions about accountability when AI remains a step removed from clinical verification. As AI models evolve to be more proactive about potential concerns, clinicians should routinely ask patients where their information comes from and verify it with trusted sources. The broader takeaway is not to abandon AI, but to demand safety standards and traceability so AI can support, not endanger, patient care.

Highlights

The coming years will test how AI can assist medicine without compromising safety.