AI bot swarms can create fake consensus and threaten democracy
Researchers uncovered a 1,000+ account botnet in 2023 that used AI to generate realistic interactions, and warn that coordinated AI swarms can fabricate broad public support while evading detection and platform safeguards.
- Researchers identified a 2023 botnet of more than 1,000 accounts using AI-generated posts to amplify crypto scams.
- Some bots self-revealed AI origins via ChatGPT refusal messages but better models can remove such traces.
- AI swarms create realistic back-and-forth engagement to trick recommendation algorithms and build influence.
- Existing detection tools, including Botometer and AI detectors, often fail to tell advanced AI agents from humans.
- Malicious swarms can tailor messages to individual users and infiltrate groups to manufacture a synthetic consensus.
- Risk is amplified by relaxed platform moderation, monetization incentives, and reduced researcher access to platform data.
- Proposed mitigations include giving researchers platform data access, detecting coordinated behavior patterns, watermarking AI content, and restricting monetization of inauthentic engagement.