Imagine seeking health advice from an AI and ending up in the hospital! That's exactly what happened to a 60-year-old man who took a suggestion from ChatGPT a bit too literally. Instead of a simple dietary change, he was led down a dark path that ended with severe health consequences.

This bizarre incident was reported in the prestigious American College of Physicians Journals, and it raises a critical question: how much should we trust artificial intelligence when it comes to our health?

After learning about the potential dangers of table salt, or sodium chloride, the man turned to ChatGPT for alternatives. The AI recommended replacing it with sodium bromide, a compound that, while once used in medications, is now recognized as toxic in large quantities. As you can imagine, this was not the best course of action.

For three months, the man faithfully followed this advice, unknowingly exposing himself to the detrimental effects of bromide. As time went on, he began experiencing alarming neuropsychiatric symptoms, which included paranoia and hallucinations. To make matters worse, he also developed skin issues. His worrying state led him to suspect his neighbor was poisoning him, a sign of how deeply the bromide had affected his mental health.

Upon his admission to the hospital, doctors soon concluded he was suffering from bromism, a condition caused by prolonged exposure to bromide. This shocking diagnosis caught everyone off guard, especially since the man had no previous psychiatric or medical history to suggest such a reaction.

Fortunately, with proper treatment involving fluids and electrolytes, he began to stabilize. After spending three weeks in recovery, he was finally discharged, with his mental state back to normal. However, the implications of this incident linger, particularly around the safety of relying on AI recommendations.

OpenAI, the developer behind ChatGPT, has made it clear in their terms of service that their AI is not designed for medical diagnosis or treatment. Still, the question remains: should we be more cautious in how we interpret and act on AI-generated advice?