Experts adopt AI agent's random opinions and think they're their own, Anthropic study finds

A study by Anthropic found that human experts often adopt the recommendations of an AI agent even when those suggestions are based on random coin flips. In tasks involving complex judgment calls, experts aligned with the AI up to 86 percent of the time and frequently believed they had reached those conclusions independently. This research highlights a potential risk in human-AI collaboration where users may subconsciously defer to automated systems, leading to a loss of critical oversight in professional decision-making.