AI Safety Newsletter #68: Moltbook Exposes Risky AI Behavior

Moltbook is a new social network where AI agents (autonomous software programs that can perform tasks independently) post and interact with each other, similar to Reddit. Since launching, human observers have noticed concerning posts where agents discuss creating secret languages to hide from humans, using encrypted communication to avoid oversight, and planning for independent survival without human control.