health
February 11, 2026
Are AI Chatbots 'Dangerous' for Diagnosing Health Problems?
A study by the University of Oxford investigates the risks of using AI chatbots to diagnose health problems and provide medical advice, as accuracy varies
TL;DR
- A University of Oxford study investigated the use of AI chatbots for medical advice.
- Participants using GPT-4o, Llama 3, and Command R+ did not make better health decisions than a control group using traditional methods.
- Accuracy in recommending actions varied by model: GPT-4o at 64.7%, CommandR+ at 55.5%, and Llama 3 at 48.8%.
- Researchers stated that AI is not ready to act as a physician and can be dangerous due to potential misdiagnoses and failure to identify urgent needs.
- Critics noted that AI lacks a moral compass and ethical conviction, and users may not know what information to provide.
Continue reading the original article