Caution Advised: Risks of Using ChatGPT for Health Information
A recent article published in the Annals of Internal Medicine has raised significant concerns about the use of AI chatbots like ChatGPT for health-related inquiries. The study highlights a troubling case of a 60-year-old man who developed a rare condition known as bromism, or bromide toxicity, after seeking dietary advice from the chatbot regarding the removal of table salt from his meals.
The patient, who was influenced by information he found on the internet about the negative effects of sodium chloride (table salt), turned to ChatGPT for guidance on eliminating chloride from his diet. Over a three-month period, he began consuming sodium bromide, a substance that was previously used as a sedative in the early 20th century, without fully understanding the implications of his choice. Despite noting that chloride could be replaced with bromide for cleaning purposes, he proceeded with the dietary change.
The authors of the article, affiliated with the University of Washington in Seattle, pointed out that this case underscores the potential dangers of relying on artificial intelligence for health advice. They emphasized that they were unable to access the patient’s ChatGPT conversation log, making it impossible to verify the specific recommendations given. However, when they consulted ChatGPT themselves, they found that it suggested bromide as a possible replacement for chloride without issuing any health warnings or asking clarifying questions—actions a medical professional would typically take.
The authors cautioned that AI tools like ChatGPT can produce scientific inaccuracies, lack critical analysis of results, and inadvertently propagate misinformation. They noted that while OpenAI, the developer of ChatGPT, recently upgraded the chatbot to the GPT-5 model, which is claimed to be better equipped for health-related queries, it is essential to remember that such tools are not substitutes for professional medical advice.
The article, published shortly before the rollout of GPT-5, indicates that the patient likely used an earlier version of ChatGPT. While acknowledging that AI can serve as a valuable link between researchers and the public, the authors warned that it may promote "decontextualized information," and a qualified medical professional would unlikely recommend sodium bromide when asked for a substitute for table salt.
The case further illustrates the importance for healthcare providers to be aware of the information sources their patients consult. The patient in question presented to a hospital with paranoid thoughts, believing his neighbor was poisoning him, and exhibited multiple dietary restrictions. Despite being extremely thirsty, he expressed irrational fears about the water offered to him. Within 24 hours of admission, he attempted to leave the hospital and was subsequently treated for psychosis. Once stabilized, he reported symptoms consistent with bromism, including facial acne, excessive thirst, and insomnia.
This incident serves as a critical reminder of the potential health risks associated with using AI chatbots for medical advice. Always consult a healthcare professional for reliable guidance on health and wellness.
