Production Ready
Is your AI-built app secure?
A plain checklist for apps built with Lovable, Bolt, Replit, Cursor, Claude Code, Codex or another AI tool.
By Joe Minock · Last reviewed
Maybe not yet. Most AI-built apps that leak data fail the same few checks: tables open to anyone, secret keys in the browser, and logins that don't really check. You can run the ten checks below today. Most take minutes.
The ten checks
- Row-level security is on for every table In Supabase, open each table and confirm row-level security is on. Then read each policy. A policy that lets everyone through hides the warning but still lets anyone in.
- No secret keys in the browser The Supabase anon key is meant to be public, but only with row-level security on. Service role keys, Stripe secret keys and OpenAI keys must never appear in code that runs in the browser. Search your front-end code for them.
- Logins are checked on the server Hiding a button isn't security. Every page and every request that returns private data must check who's asking, on the server.
- Paid features are checked on the server If someone can unlock a paid plan from their own browser, sooner or later someone will.
- Private apps are really private Make sure a stranger can't simply sign up to an app that's meant only for your team.
- Payments really charge Check that Stripe webhooks are set up, that production uses live keys, and that failed payments and refunds are handled.
- Every key has a spending cap Set limits and alerts on every AI and cloud key. A stolen key with no cap can run up a huge bill in days.
- Test data and live data are separate Your AI agent should never be able to touch live data. Give it a test copy.
- You have backups, and you've restored one A backup you've never restored is a hope, not a backup.
- Someone gets told when it breaks Error tracking and uptime alerts, so you hear about problems before your customers do.
If you got the Supabase warning
"RLS Disabled in Public" means a table in your public schema has row-level security turned off, so anyone with your project's address and public key may be able to read or change it (Supabase's own docs explain the check). Turn row-level security on, then write policies that let each user see only their own rows. Don't just add a policy that lets everyone through. That silences the warning without fixing anything. Our full guide to the warning has the steps and a policy that works.
What scanners miss
Automated scanners find known patterns, like an exposed key. They miss logic mistakes. In February 2026, The Next Web reported an app on Lovable's own showcase where the login check worked backward: anonymous visitors got full access and signed-in users were blocked. A person reading the code catches that. A scanner often doesn't.
Why this matters now
- A researcher found 170 of 1,645 apps on Lovable's showcase letting anyone read user data (CVE-2025-48757, reported by Semafor).
- Escape scanned 5,600 public apps built with AI tools and found more than 2,000 vulnerabilities, over 400 exposed secrets and 175 cases of exposed personal data, including medical records.
- Moltbook's database key sat in browser code with row-level security off, exposing 1.5 million API tokens and thousands of private messages (Wiz).
- A scan of about 380,000 public apps on Lovable, Base44, Replit and Netlify found about 5,000 leaking corporate or personal data, as reported by Axios.
See the full dated list of incidents.
Want us to check it?
Our Production Ready review is a senior read of your app, delivered within a week, with a ranked list of fixes in plain words and an honest call on what to do first.
Common Questions
Frequently asked questions
Is the Supabase anon key safe to have in my app?
Yes, it is meant to be public, but only when row-level security is on and correct for every table. Without that, the public key is a key to your data. The service role key is different: it must never be in browser code.
Will an AI security scanner catch everything?
No. Scanners find known patterns, like an exposed key. They miss logic mistakes, like a login check that works backward. A person reading the code catches those.
If Lovable or Bolt is secure, is my app secure?
Not on its own. The platform runs your app, but your app's data rules, keys and login checks are yours. Most public leaks came from those, not from the platform.
What should I do first if I think data leaked?
Lock it down first: turn on row-level security, rotate every key that was ever exposed, and check your logs for access you don't recognize. Then ask a lawyer whether you must tell your users. The rules depend on where you are and what kind of data it was.
Proof
How we built Deliberate with Claude Code
Deliberate is our own Mac app, built by a three-person team with Claude Code. We built the app, its registration and billing site, and integrations with Gmail, Outlook, Slack, Asana, Zoom and Granola. It runs on your Mac, never saves the full text of your email, and sends nothing without your yes.
Meet DeliberateWant a senior read on your app?
Send us the link and what worries you. We'll tell you what we'd fix first.