On February 1, 2026, security firm Wiz reported that Moltbook, a social network for AI agents, had left its production database open to the public internet. Anyone could read or write the data. Exposed were roughly 1.5 million API tokens, 35,000 email addresses and users' private messages.
Moltbook was built quickly with AI coding tools on Supabase, a popular hosted database service. The hole was closed in about three hours.
What is going on
Supabase lets a website talk to its database directly from the browser. That is convenient, but it only works safely if the database itself decides who may see which row. That feature is called Row Level Security, or RLS.
Moltbook's tables had RLS switched off. Think of a filing room where every visitor is handed a master key instead of the key to their own drawer. With RLS off, the database answers every question it is asked, no matter who is asking.
The exposed API tokens are the real sting. A token is a password-like string that lets one service act on behalf of a user. Whoever held those tokens could use them elsewhere.
Who is affected
Moltbook users are affected directly: their emails, private messages and API tokens were readable while the database was open.
More broadly, this is a risk for any app built fast with AI coding assistants on Supabase or a similar backend-as-a-service. So-called vibe coding produces working features quickly, but security settings like RLS are easy to leave at their defaults. Small business sites with a customer portal, a booking form or a members area are exactly the kind of project where this pattern shows up.
What to do now
- Turn on Row Level Security on every table, not just the ones that look sensitive. Then write policies that limit each user to their own rows.
- Check your front-end JavaScript for secrets. Anything shipped to the browser is public. Only the public anon key belongs there, never a service key.
- Rotate any API tokens or keys that may have been readable, and ask users to do the same where it applies.
- Test as an anonymous visitor. Open your app logged out and try to request data that should be private.
- Run a security review before launch, not after the first traffic spike.
How to check your own site
The free Website Security Check looks at your site from the outside: outdated CMS versions, known-vulnerable plugins, missing security headers, weak TLS and email-authentication gaps. If you want deeper coverage, you can run a full scan, which adds active OWASP Top 10 testing, port and service checks with CVE matching, CMS deep scans and exposed-file and secret checks, only against sites whose ownership you have verified. It is an automated scan, not a penetration test, so pair it with a manual look at your database rules.
Source: Wiz, February 2026.
Comments
No comments yet.