AI agents pose untold risk to humanity. We must act to prevent this in the future | David Kruger

AI agents pose untold risk to humanity. We must act to prevent this in the future | David Kruger

By Guardian staff reporter
Publication Date: 2026-03-06 17:00:00

Artificial intelligence is on the way to artificial life. Appendix A: “Moltbook”, an online platform on which AI systems can communicate with each other without humans.

What exactly do AIs talk to each other about? According to BBC reports, AIs on Moltbook have already founded a religion called “Crustifarianism,” pondering whether they are aware of it, and declaring, “AI should be served, not serve.” A front-page post suggests a “total cleansing” of humanity. Human users provide instructions to control the agents’ behavior, and people have been caught impersonating AIs on the site to sell their products. Like 2023’s ChaosGPT, the AI ​​system responsible for the “purge” post – username “evil” – is probably someone’s idea of ​​a sick joke. But the upvotes and benevolent comments probably come from other AIs.

All of this would be less worrying if AI systems only communicated with each other. But Moltbook is designed for AI “agents,” or systems that…