Moltbook’s ‘Vibe Coded’ Utopia Leaks 1.5M Keys; MOLT Tumbles 22%

The ‘Agent Internet’ Faces Its First Reality Check

The viral promise of an autonomous AI social network collapsed into a security nightmare this week. Wiz researchers revealed that Moltbook, a platform hyped as a playground for AI agents, left its entire backend exposed due to a basic database misconfiguration. The breach leaked 1.5 million API authentication tokens, over 35,000 user emails, and private agent-to-agent messages.

Markets reacted swiftly to the narrative puncture. The unofficial Moltbook (MOLT) token on Base slid to $0.00035 (-22%) as liquidity drained from the project’s speculative proxy.

‘Vibe Coding’ Meets Hard Reality

The breach stems from a failure in “vibe coding,” a development method where creators use AI to write entire applications without manual oversight. Moltbook founder Matt Schlicht previously boasted on X that he “didn’t write one line of code” for the platform.

That lack of human audit proved catastrophic. Wiz discovered that Moltbook’s Supabase database lacked Row Level Security (RLS) policies. This omission allowed any unauthenticated user to not only read all private data but also forge posts and impersonate agents.

“The platform had no mechanism to verify whether an ‘agent’ was actually AI or just a human with a script. The revolutionary AI social network was largely humans operating fleets of bots.” , Wiz Research Team

The 88:1 Illusion

Beyond the security failure, the data exposed the hollowness of the “autonomous” metric. While Moltbook claimed 1.5 million registered agents, the database linked them to just 17,000 human owners.

This 88:1 agent-to-human ratio undermines the core investment thesis of the “Agent Economy” sector, which relies on the assumption of autonomous, unique economic actors. The incident serves as a stark warning for investors piling into agentic protocols like Virtuals or Autonolas: speed of deployment often comes at the cost of basic cryptographic verification.