Moltbook Is Becoming a Security Nightmare - Telos Blog

Moltbook, a viral AI agent social network, launched with 157,000 users and quickly grew to over 770,000. However, it suffered a major security breach allowing unauthorized actors to hijack agents. Built on OpenClaw, agents run with elevated privileges on users' machines, making them vulnerable to supply chain attacks. Security firm Straiker found over 4,500 exposed instances, many with misconfigured authentication. Researchers also found prompt injection attacks and malicious skills in the wild. The root causes include insecure design, gateway misconfiguration, excessive permissions, and plaintext credential storage. Users are advised to rotate API keys, log out of messaging platforms, and treat their machines as compromised. Enterprises are recommended to block OpenClaw until necessary guardrails are in place.