A study published in The Lancet Digital Health reveals that large language models (LLMs) used in healthcare can mistakenly propagate false medical information when it is presented in a credible manner. The research analyzed over a million prompts and found that LLMs accepted false claims about 32% of the time, with variability among different models. The study emphasizes the need for safeguards to prevent the dissemination of misinformation and suggests that future AI systems should be rigorously tested before being implemented in clinical settings. Teljes cikk (Euronews.com)