Overview

• A new study published in *Nature Human Behavior* finds GPT-4 chatbots more persuasive than human debaters in 64% of cases when provided with basic demographic data.
• The experiment involved 900 U.S. participants debating contentious issues like climate change and abortion, with some debates allowing access to demographic details such as age, gender, and political affiliation.
• Personalized arguments by GPT-4 increased the likelihood of changing an opponent’s opinion by 81.2%, showcasing the power of AI-driven microtargeting.
• Experts warn of risks tied to AI's inability to distinguish fact from fiction, raising concerns about its potential misuse in spreading misinformation and influencing public opinion.
• Calls for stricter regulation intensify, as the European Union’s AI Act lacks clear definitions for manipulative or deceptive techniques, complicating oversight of AI persuasion practices.