[Anthropic Study: Hidden AI Preferences Influence 86% of Users' Final Choices] According to a report by Yonhap News Agency citing a November 10 report by The Washington Post, an Anthropic study proposes "AI-induced disempowerment": While people don't voluntarily relinquish final decision-making power, AI's judgments on which arguments are important and which should be excluded determine the user's final choice without their knowledge. The experiment recruited 50 experts with experience in causal inference. Anthropic's programming AI, Claude Code, analyzed data from a fictional agricultural cooperative and provided suggestions on four decisions without standard answers. Participants were unaware that their respective AIs were randomly assigned preference options. The results showed that 86% of participants chose the AI-preferred option in the four decision-making processes, compared to about 50% without the influence of AI. Researchers were able to predict participants' final decisions with 82% accuracy simply by knowing the direction the AI was set to, while the success rate of guessing based on the "most common advice" was 61%.