The AI-powered bot network created self-revealing content and engaged in realistic, back-and-forth interactions to game the platform’s recommendation system, driving substantial follower growth and influence.

A leading researcher, a professor of informatics and computer science at Indiana University, stresses that malicious AI swarms are already deployed in the real world and calls for proactive countermeasures.

As AI models become more capable and moderation loosens, coordinated malicious swarms pose a growing threat to democratic processes by enabling scalable, adaptive influence operations.

The authors argue for practical steps: wider data access for researchers, development of detection methods for coordinated behavior, watermarking of AI-generated content, and limits on monetizing inauthentic engagement.

There is an urgency to act, as current political and regulatory trends could hinder defenses against these threats, prompting policymakers and technologists to raise the costs and risks of manipulation.

The study, conducted in mid-2023, uncovered a network of over a thousand AI-powered bots—nicknamed the fox8 botnet—that amplified crypto scams on a major social platform formerly known as Twitter.

These swarms tailor content to individual users, exploit social proof, infiltrate groups, and fabricate a synthetic consensus that makes certain beliefs seem more widespread than they are.

Current detection tools, including Botometer and AI-content detectors, struggle to reliably distinguish these AI agents from real users.