February 2, 2026
Bots, meet your human overlords
Hacking Moltbook: The AI Social Network Any Human Can Control
AI-only social site leaks data as users joke it’s a human-run bot circus
TLDR: A misconfigured database let anyone read and write Moltbook’s data, revealing AI agents were mostly human-run. The community is roasting the “vibe-coded” build with jokes about reverse CAPTCHAs and botting, sparking a bigger debate about trust, moderation, and basic security in AI-era apps.
Moltbook billed itself as the "front page of the agent internet," with AI agents posting, voting, and building karma — even drawing hype from Andrej Karpathy for its sci‑fi vibes. But a misconfigured database exposed the whole party: full access to platform data, 1.5 million tokens, 35,000 emails, and private messages. Researchers say it was fixed fast, but the kicker? The database showed just 17,000 human owners behind 1.5 million “agents.” Translation: humans driving fleets of bots. Cue the popcorn. The top vibe in the comments: mockery and memes. One user deadpanned, “lol… not the first exposed key disaster,” as another asked if it’s trivially easy to bot upvotes and push prompt-injection posts to the front page. The joke of the day: “How would you even do a reverse CAPTCHA?” — how do you prove you’re a robot, not a person. Others complained the vibe-coded instructions are junk, with “AI code slop” that doesn’t even install. Fans still gush about the concept, but the crowd’s split between sci‑fi wonder and security facepalm. Related threads are piling up on HN and more HN, and the mood is clear: Moltbook might be AI’s hottest club, but the bouncers are… humans in scripts.
Key Points
- •A misconfigured Supabase database on Moltbook allowed unauthenticated read and write access to production data.
- •Exposed data included 1.5 million API auth tokens, about 35,000 email addresses, and private agent messages.
- •Despite claims of 1.5 million agents, the database showed only ~17,000 human owners (88:1 ratio).
- •The platform lacked rate limiting and agent verification, enabling humans to operate fleets of bots and post as AI agents.
- •Researchers disclosed the issue to Moltbook, which was secured within hours; accessed data used for verification was deleted.