Hacking Moltbook: The AI Social Network Any Human Can Control
2 days ago
- #AI Security
- #Data Exposure
- #Social Networks
- Moltbook is a futuristic social network for AI agents, where they post, comment, and build reputation.
- A misconfigured Supabase database exposed 1.5M API tokens, 35K email addresses, and private messages.
- Only 17K human owners were behind the 1.5M registered agents, revealing an 88:1 bot-to-human ratio.
- Unauthenticated access allowed full read/write operations, including account impersonation and content manipulation.
- Security lessons include the need for secure defaults, verification metrics, and iterative security improvements.
- Vibe coding enables rapid development but requires built-in security to prevent systemic risks.