How do I know if my vibe-coded app is secure?
Five checks a non-engineer can ask for on an AI-built app, what each one means in plain words, and why only a written review of the code can tell you for certain.

You can't tell from the outside, because an app can look finished and work perfectly for its owner while another customer's data is open to anyone who asks for it. What you can do is ask five specific questions about access, payments, secrets, backups and logging, and treat any answer you can't get in writing as unknown. A written review of the code, with read access to the hosting, is the only way to turn "unknown" into "checked".
Why can't I just click around and see if it's secure?
Because security problems are rarely visible on a screen you are logged into. When you test your own app you see your own data, so a missing rule that would let customer B read customer A's records never shows up. The most common category of web application flaw is exactly that kind: OWASP, the open web security project, ranks broken access control first in its 2021 Top 10, moved up from fifth place in the previous edition, and describes testing 94% of the applications in its data for some form of it.
None of this says AI tools write worse code than people do, and we wouldn't claim that. It says the part of an app that decides who can see what is easy to leave out when the goal was to get a screen working, whoever wrote the code.
What are the five checks?
Ask whoever built the app, or a developer you trust, to show you each of these in writing:
| Check | The question to ask | What a good answer looks like |
|---|---|---|
| Access | Can one customer ever read or change another customer's records? | A named rule in the database or server code, and a test that proves it |
| Payments | Does the server confirm that payment messages really come from the payment provider? | Signature verification on the webhook, with the secret stored outside the code |
| Secrets | Where are the passwords and API keys, and could a visitor's browser see them? | Held on the server or in the host's secret settings, never in files sent to the browser |
| Backups | When did you last restore a backup, not just take one? | A dated restore test |
| Logging | If something went wrong last Tuesday, could anyone find out what happened? | Error and access logs that are kept and that somebody reads |
What does "access" mean for a database like Supabase?
It means every table that holds customer data has a rule saying who can read and write each row. If your app stores data in Supabase, that rule is called row-level security. Supabase's documentation says to "enable RLS on every table in an exposed schema", and warns that a table without it is readable and writable by any role with a grant on it. It also says that once RLS is on, no data is accessible through the API with a publishable key until policies are created, so the work isn't finished when the switch is flipped: someone has to write the policies, and someone has to check they say what you meant.
You can ask for a plain list of your tables with one column saying whether RLS is on. Any "no" next to a table that holds customer details is the first thing to fix.
How do I know my payment flow can't be faked?
Your server has to verify every payment message before it acts on it. Stripe's webhook documentation says that without verification "an attacker could send fake webhook events to your endpoint to trigger actions like fulfilling orders, granting account access, or modifying records", and that Stripe signs each event in a Stripe-Signature header you check against your endpoint's secret. A flow generated to make payments work can skip that step, because it works perfectly in testing without it. The question for your developer is a short one: "Do we check the Stripe signature on every webhook?"
Where should passwords and API keys live?
On the server, or in your host's secret settings, and not in any file the browser downloads. A key that was pasted into the code early on to make something work is easy to forget, and it tends to stay in the project history even after it is removed from the current files. If you can't say where each key is kept, or if one has ever been shared in a chat or a screenshot, the safe step is to replace it with a new one. Rotating a key is routine and takes minutes for most services.
Does this matter legally in the UK?
If your app holds personal data, yes, because the UK GDPR requires it to be processed with appropriate security. The Information Commissioner's Office security guidance quotes Article 32, which asks for "a level of security appropriate to the risk", and says the law doesn't set out specific measures, so what counts depends on your circumstances and the data. It also describes vulnerability scanning and penetration testing as ways to stress-test what you have. We aren't lawyers, and this is a pointer to the source and not legal advice.
What does a Rescue audit do about all this?
It reads the code and the hosting setup and tells you in writing what it found. The We UC Rescue audit is a fixed £2,500 + VAT, deducted from the first phase if you go ahead with us on the fix, and it covers security, data, hosting, code, payments and integrations, and running costs. You get a written report, a priced plan and a walkthrough call. Nothing is changed during the audit, so we need read access to your code and hosting, and the report is yours either way.
We UC is a UK software company, and the team doing the review is the team that runs Olatti, a UK business phone network, and Quenta, an app that posts receipts into Xero, in production. Two QA engineers test our work, and an engineer reviews every change before it ships. If you are still deciding what to do with what you find, our piece on whether to fix or rebuild a vibe-coded app covers that decision, and the one on what an audit costs covers what you should expect to pay.
Common questions
Is a vibe-coded app less secure than one written by hand?
We can't say it is in general, and we won't pretend to have a figure. Any app can leave out the rules about who sees what, and an app built quickly to prove an idea is more likely to have skipped them. Reading the code is the only way to know about yours.
Can I run a security scanner myself?
You can, and it will catch some things, but a scanner looks at what it can reach from outside. The access rules inside your database and the way your server handles payment messages usually need someone to read the code and test them with more than one account.
Will the audit tell me off for using AI tools?
No. You got a working product in front of real users, which is further than most ideas ever get, and the audit is about what the app needs next.
Who owns the code after an audit?
You do. Your code, your repository and your accounts stay yours.
If your app has paying users or customer data and you aren't sure about any of the five checks, you can book a Rescue audit, or tell us what you've built and we'll say whether an audit is the right first step.