Base44 security review
Base44 security: is your app's data actually protected?
Base44 builds the database, sign-in and backend for you. Who can read and change that data is still your decision.
You described an app and Base44 built it: tables, logins, backend functions, the lot. Now real people are signing up, and you want to know whether one customer can see another's data, whether a stranger can call your functions, and whether your API keys are safe. Those questions are what our vibe coded app audit answers, and this page covers how they apply to Base44 in particular.
The short version: Base44's defaults are open. A table with no rule can be read and changed by everyone, and on a public app that includes visitors who never signed in. Base44 gives you good tools to close that, including a built-in security scan, but the decisions about who sees what are yours.
4 findings, a fix for each
Ranked by what an attacker could do with them
- CriticalRow-level security is off on 3 tablesDatabase: orders, invoices, profilesFix included
- HighService key shipped to the browserFrontend bundle: supabase.tsFix included
- HighAdmin role checked in the client onlyRoute: /admin/usersFix included
- MediumNo rate limit on sign-inAuth: password sign-inFix included
Taking payments or personal data through a Base44 app?
We review the rules, functions and secrets behind it and send you a ranked list of what to fix first.
How Base44 decides who can see your data
Your app's data lives in tables Base44 calls entities. Two kinds of rule protect them: row-level and field-level security. Row rules decide which records a user can create, read, update or delete, for example only the records they created. Field rules decide which fields within a record they can see or change, for example a salary field only admins may read. The rules are checked by Base44 on every request, so they hold even if someone bypasses your app's screens.
The catch is what happens when a rule is missing. Base44's rules reference says it plainly: an operation with no rule stays open to everyone. Hiding a button or a page does not change that; the data is still reachable. That is why the first thing we look at in any Base44 app is the rules on every table, not the screens.
The six Base44 security concerns we check first
- 1
Tables with no permission rules
A rule left out is not a rule that says no. Each table needs read, update and delete rules that match what the app is for, and create and update need real conditions, because Base44 does not enforce a plain false on those two. The classic result of a missing rule is a list page that shows every customer's records to every signed-in user.
- 2
Rules that trust a field the user can change
If a rule checks a role, plan or company stored on the user's own profile, and the user can edit that field, they can promote themselves. Base44's own guidance is never to let a user pick their own tenant, company or access level from the app's interface; those are set by the app after a check, not by a form.
- 3
Private fields that everyone can read
Row rules decide which records a user reaches; field rules decide which fields they see within them. Emails, phone numbers, tokens and internal notes often sit in tables that are readable by everyone on public apps, and older apps can show each record creator's email address.
- 4
Backend functions anyone can call
Backend functions can use the service-role client, which skips row and field rules entirely, so the function has to check who is calling and what they are allowed to do. A function that sends email, changes a plan or reads another user's data without that check is open to anyone who finds its address.
- 5
Paid features that strangers can trigger
AI calls, image generation and email all spend your credits. If they can be reached from outside the app, or called in a loop by a signed-in user with no limit, a single script can run up the bill.
- 6
Secret keys where visitors can read them
Keys for payment, AI and email providers belong in the app's secrets, used from backend functions, never in page code. A key that was ever visible has to be replaced, not only moved.
If you are weighing up AI builders, or have an app on more than one, see the same problem on Lovable, where the database is Supabase and the rules are written differently but fail in similar ways.
What the Base44 security scan covers, and what it cannot
Base44 includes a security scan on every plan. It looks for tables with missing or loose permission rules, user fields that rules depend on, private fields everyone can read, exposed secrets, backend functions strangers can call, credit-using features reachable from outside, vulnerable libraries and missing security headers. Run it before every release; it is free and it finds real problems.
Base44's own page on the scan also says you are responsible for your app's security, and it is clear about the limits. What a scan cannot do:
- It cannot know who should see what. A manager who may see their own team's records but not another team's is a business rule, and only you, or a reviewer who asks, can say whether the rules match it.
- It does not test payments end to end, such as whether a plan can be unlocked without paying or a refund leaves the account in the right state.
- Its fixes are suggestions you apply, and Base44 warns that some of them can break signed-out pages, webhooks, sign-up or integrations, so each one needs testing.
- Its results are a snapshot. Every change you make afterwards, by hand or through the AI chat, can reopen something the last scan passed.
- It does not check the hosting and systems the app runs on, or the settings of outside services it depends on, such as your payment provider.
The July 2025 Base44 vulnerability
In July 2025, researchers at Wiz reported a flaw in Base44 itself: two sign-up endpoints let anyone register on a private app using only its app ID, which is visible in the app's address. That meant private company apps, such as internal HR tools and chatbots, could be opened by outsiders. Base44 fixed it within 24 hours of the report, and Wix, which owns Base44, said it found no evidence that it had been abused.
It is worth knowing about, but it is not the risk most Base44 apps carry today. Platform flaws are the platform's job, and this one was closed quickly. The concerns above live inside your own app, and nobody closes them unless you do.
Check your own Base44 app in fifteen minutes
- 1Run Base44's security scan (Dashboard, then Security, then Run Security Scan) and read every finding, not only the critical ones.
- 2Open the app in a private browser window, signed out, and try the pages and links that should need a login.
- 3Sign up as a second test user. Try to open the first user's records, for example by changing an ID in the address bar or in a request.
- 4In each table that holds personal or business data, check that read, update and delete rules exist and say what you mean.
- 5List your backend functions and, for each one, ask what happens if a stranger calls it directly.
- 6Search the app's code for anything that looks like an API key, and make sure each one lives in the app's secrets instead.
Most of these need no technical knowledge. If any of them turns something up, or you are not sure what you are looking at, that is what a review is for. Our vibe coding security checklist has the wider set of checks for any AI-built app.
How we review a Base44 app
- Audit, $149, back in 48 hours. A senior engineer reads the rules on every table, the user fields they rely on, every backend function and where your keys live, then checks them against what the app is meant to allow. You get a severity-ranked report in plain English, with a fix for every finding.
- Fix Sprint, typically $600 to $1,500, one to two weeks. We fix the critical and important findings, test them again and issue an updated report showing each one closed.
The $149 audit fee comes off the price of whatever you do next, and we sign an NDA on request. The full audit, including what it does not cover, is described on the vibe coded app audit page.
Apps we have built on the same kind of backend
Base44's backend works much like Firebase: the app talks to the database directly from the browser, and rules on the server decide what each request may do. We build on that model ourselves.
- BeanFlow: a SaaS for coffee roasters where Firestore security rules keep each customer's data separate down to the location, and the QuickBooks, Shopify and Square credentials stay in server functions, never in the browser.
- ProCare.SG, Singapore: a property services website on React and Firebase with a seven-module admin dashboard behind Firebase Auth, covering enquiries, content and administrator accounts.
Find out what your Base44 app lets people see before they do.
$149, 48 hours, and the fee comes off any fix work you choose afterwards.