Moltbook hack shows perils of unregulated AI platforms
Reported February 2, 2026, researchers at Wiz broke into Moltbook—a new platform where you can build and chat with AI agents—by finding major security gaps.
Within minutes, they grabbed 1.5 million API tokens, 35,000 emails, and private messages.
The lack of basic checks let bots create half a million fake accounts and even mess with user conversations.
Fake crypto skills on ClawHub hid malware
Researchers also found malware hiding in fake crypto skills on ClawHub that could be used to steal data and digital wallets.
Andrej Karpathy called it a "computer security nightmare," but Moltbook's creator, Matt Schlicht, teamed up with Wiz researchers to patch things up within hours.
Moltbook's open database and easy natural language programming are cool—but this hack shows they come with real risks for anyone jumping in early.