Details may change as reporting changes. Information comes from the linked sources, and we do not guarantee it is complete or current.
In plain English
Moltbook was a social network where AI agents, not people, post and chat. It went viral. Its founder said publicly that he had not written the code, and that AI had built it from his vision.
Security researchers at Wiz looked at it as an ordinary visitor would. Within minutes they found a Supabase key in the website's public code that gave full read and write access to the production database. This looks like the same kind of problem as the previous case, a public key without proper rules behind it, though Wiz describes it as a misconfigured database.
The exposed data included 1.5 million access keys, 35,000 email addresses and private messages between AI agents. The security firm Wiz told the team, who fixed it within hours.
Words used here
- Vibe coding
- Building software by describing it to an AI and accepting what it writes, often without reviewing the code.
- Production database
- The real database holding real users' data.
- Responsible disclosure
- Telling the owner privately so they can fix a problem before others find it.
What happened
Wiz identified a misconfigured Supabase database belonging to Moltbook that allowed full read and write access to all platform data.
Wiz reported the exposure included 1.5 million access keys, 35,000 email addresses and private messages between AI agents.
Wiz disclosed it immediately. The Moltbook team secured it within hours with Wiz's help, and Wiz says data accessed during the research was deleted.
Wiz noted the founder had said publicly that he did not write a line of the code.
How it unfolded
- January 31, 2026
Wiz identifies the exposed database and discloses it.
- Within hours
Moltbook secures the database.
Who was affected
People and agents registered on Moltbook, per the figures Wiz reported. Reported counts vary across outlets.
What the response was
- Wiz disclosed privately and helped verify the fix.
- Moltbook locked the database down within hours.
What went wrong
- A public key gave full access because proper database rules were not in place.
- Nobody with security experience reviewed what the AI generated before a public launch.
What worked
- Responsible disclosure and a very fast fix.
What a small team should do differently
- Have someone who understands security review an AI-built app before launch.
- Test your own app as a stranger: open it logged out and try to reach other people's data.
- Store as little user data as possible. What you do not store cannot leak.
What you can do today
- Run the same "stranger test" on your app: log out, open the browser's network tab, and see what data comes back.
- Check Supabase RLS is on for every table.
- If your app holds access keys for users, scramble them so they cannot be read, and replace any that may have been exposed.
Official help for the platforms involved
The lesson to take away
Going viral is not a security test. Before you invite the internet, check what a stranger can reach.
Sources
Finished this one?
Add it to What I learned and earn 40 Trust. You can open your list from the badge at the bottom of the page.