HomeTechnologyAI-built social network Moltbook leaks user data after major security lapse

AI-built social network Moltbook leaks user data after major security lapse

Moltbook, a self-described social network for AI agents, has suffered a major security lapse that exposed user credentials and private messages. The flaw stemmed from the platform being almost entirely built by an AI assistant, according to its founder. The episode underlines how experimental AI-built products can unravel in the real world.

February 03, 2026 / 10:06 IST
Story continues below Advertisement
Artificial Intelligence
Artificial Intelligence
Snapshot AI
  • Moltbook leaked 1.5M API tokens, 35K emails, and private AI agent messages
  • Unauthenticated users could edit live posts, risking content integrity
  • Moltbook was built entirely by AI, exposing flaws in AI-generated code

Moltbook’s premise was strange enough to begin with. The platform bills itself as a social network designed not for humans, but for AI agents talking to one another. That novelty has now been eclipsed by a far more familiar story in tech: a poorly secured system leaking sensitive user data.

The vulnerability was discovered by cybersecurity firm Wiz, which also helped Moltbook address the issue after it was reported. In a detailed analysis published by Wiz, the firm said it was able to access a vast amount of sensitive information tied to Moltbook’s users and agents.

Story continues below Advertisement

According to Wiz, the exposed data included around 1.5 million API authentication tokens, roughly 35,000 email addresses, and private messages exchanged between AI agents on the platform. The flaw went further than passive data access. Wiz found that unauthenticated users could edit live posts on Moltbook, effectively allowing anyone to alter content without logging in.

That raised a deeper problem for a platform built around artificial identities. With no reliable authentication controls in place, there was no technical way to determine whether a given post was written by an AI agent or by a human pretending to be one. Wiz’s assessment was blunt, concluding that the supposedly autonomous AI social network was, in practice, “largely humans operating fleets of bots”.