Skip to content

It works. That does not mean it is safe.

It took one person a few hours to find sixteen security holes, six of them critical, in an education app that Lovable had shown as an example of what its tool can build. They shared the results on Reddit. Records for 18,697 users were open to anyone, including student grades, and a stranger could delete accounts or change grades without logging in. The app worked perfectly for its real users the whole time.

This is the main risk with apps built by AI tools. They do what you asked, but nobody asked them to lock the doors. If any of these sound familiar, your app needs a check before more people use it.

  • You built it with Lovable, Bolt, v0, Cursor or Replit and it handles real user data
  • Each new feature you add seems to break an old one
  • You are not sure who can read your database
  • It runs fine on your laptop but fails or slows down when it goes live
Team Work

Vibe-code rescue and MVP development services

We fix the app you have. We only rebuild the parts that cannot be made safe.

How often AI-built apps have security holes

Stories like the one above are not rare. A research team from Microsoft and the CISPA Helmholtz Center for Information Security in Germany published a study of vibe-coded apps in June. From a collection of 10,517 open-source apps built with AI tools, they checked 200 that were live on the web.

180 of the 200 apps, or 90%, had at least one security hole. In total the team confirmed 1,471 problems, and 76.7% of them were rated critical or high. The most common was broken access control, where users can see or change data that is not theirs, at 36% of all problems. Next came weak protection of passwords and secret keys at 20.7%, and unsafe handling of user input at 17.7%. The researchers also found that better AI models and better prompts reduce these problems, but do not remove them.

The good news is that most of these problems can be fixed without starting over. After the audit, you choose how far to go. This is how the three options usually compare.

Option Best when What you get
Audit only You or your own developer will do the fixes A ranked fix list in 3 working days, for a fixed $299
Fix in place The app works and the problems are in certain areas Security and bug fixes on your existing code, with a fixed quote
Rebuild a part One area, often logins or the database, cannot be made safe That part rebuilt properly while the rest of the app stays

Once the app is safe, it needs to stay that way. New features and new AI-written code can open new holes, so we offer regular checks and updates through our care plans.

Frequently asked questions

Which AI tools do you work with?

Apps built with Lovable, Bolt, v0, Cursor, Replit, Windsurf, Claude and similar tools. Most use React or Next.js on the front, with Supabase, Firebase or Node behind.

Do we have to rebuild everything?

Rarely. Most AI-built apps can be made safe and stable as they are. We only suggest rebuilding parts that cannot be fixed safely.

How fast can you start?

The audit usually starts within a few days of booking and is delivered in 3 working days. If your app has a live data leak, tell us and we will look at it first.

Is our code kept private?

Yes. We sign an NDA on request and only access the parts of your app we need to review.

I have not built anything yet. Can you help?

Yes. Our AI-built websites service designs and launches a secure site for you, and we build app MVPs the same careful way.

Guides from our blog

Plain-English guides that answer the questions clients ask us most about this service.

Not sure your AI-built app is safe to launch?

Book a fixed-price audit for $299. Within 3 working days you will know exactly what needs fixing, and in what order.