Science Paper Warns AI Swarms Could Hijack Democracy Through Synthetic Consensus

Science·medium signal

UBC computer scientist Kevin Leyton-Brown and co-authors published in Science (April 20) warning that LLM-powered agent swarms can coordinate in real time, adapt to feedback, and sustain coherent narratives across thousands of accounts to manufacture the illusion of consensus. Unlike old-school botnets, these agents imitate human social dynamics and are nearly undetectable. The paper proposes safeguards including real-time coordination monitoring, stronger account verification, and reduced monetization of inauthentic engagement.