Security auditFor AI-built apps

A security audit of your app before it goes live or works with business data. Voltti reviews the code and configuration of apps built with AI, using AI models and its own skills, and a person checks every finding.

For businesses in Belgium and the Netherlands. Custom quote.

Built fastRarely reviewed

With Claude, Cursor or Lovable, an internal tool or customer portal can be up in a few days. The code looks tidy and does what you asked. Whether it is also secure is something the tool does not check on its own. Veracode tested code written by AI models and found a security flaw in 45% of the tests (GenAI Code Security Report, 2025).

01

It works, so it looks done

An AI tool builds what you ask for: a login, a form, an integration. Whether another user can reach your data through an edited link is something the tool usually does not test.

02

Speed leaves traces

API keys in the code, test accounts that stay active, error messages that show too much. When you build fast, that kind of clutter is easy to leave behind.

03

Business data raises the stakes

Once the app handles customer details, invoices or internal documents, a leak is no longer a cosmetic flaw. It can become a data breach under the GDPR.

What we check

01

Secrets and API keys

Passwords, tokens and API keys in the code, the repository or its history.

CodeRepository
02

Login and permissions

Who can do what? We check that every user can only reach their own data, through the API as well.

AuthenticationAuthorization
03

Input and injection

Forms, search fields and parameters that end up in a database query or command without checks.

SQL injectionValidation
04

AI features in your app

If your app lets a language model read text or documents, we look at prompt injection and at what the model is allowed to do.

Prompt injectionLLM
05

Dependencies

Packages and libraries with known vulnerabilities (CVEs), and packages you ship but do not use.

CVEsPackages
06

Data leaks in logs and responses

Log files and API responses that show more personal data or internal details than needed.

LoggingGDPR
07

File uploads

Which files your app accepts, where they are stored and who can open them afterward.

UploadsStorage
08

Rate limiting, CORS and headers

Limits on the number of requests, and CORS and security header settings that are often too loose.

ConfigurationHeaders

How an audit works

  1. 01

    Intake and scope

    We agree on which app, which repository and which environment we review, and what stays out of scope. You give read access to the code. We do not need your production data.

  2. 02

    Broad analysis with AI

    AI models and Voltti's own skills go through the code, the dependencies and the configuration. We also work with local models, so your code does not have to leave the network.

  3. 03

    A person reviews

    A person checks every finding. Is it a real problem, how big is the risk and what is the best fix? We remove false alarms. You never get unchecked AI output.

  4. 04

    Report and recheck

    You get a report with the findings, their severity and concrete fixes. After the repairs, we check again whether the flaws are closed. If you want help with the repairs, we include that in the quote.

What you getAnd what you do not

In short: after a security audit you know which flaws your app has, how serious they are and how to close them. We test against the OWASP Top 10 for web applications and, for apps with AI features, against the OWASP Top 10 for LLM applications.

01A report you can use

For each finding you read where the problem is, what can go wrong and how to fix it, with an example fix where possible. Your developer or your AI tool can start working on it right away.

02Severity and order

Every finding gets a severity: critical, high, medium or low. That way you start with what someone could abuse today, and small improvements do not get pushed back until they are forgotten.

03Who an audit is for

Businesses that built an internal tool, a customer portal or an automation with AI, themselves or through an agency or freelancer. And agencies and freelancers who build apps with AI for their clients and want an independent check. Especially when the app handles customer data or business data, or is about to go live.

04What an audit is not

A security audit looks at code and configuration. A pentest, where someone attacks your running app from the outside, is something else and is not part of it. We do not issue a certificate or seal, and no audit can promise that an app is 100% secure. An audit does make the risk a lot smaller.

This service is about your app. How Voltti secures its own Hermes installations is covered under security and privacy. If you work with AI tools and business data yourself, also read the AI safety checklist for business data.

Questions and answers

01What is a security audit of an app?

A security audit is a focused check of your app's code and configuration for security flaws. You get a list of findings, their severity and concrete fixes. At Voltti, AI models do the broad analysis and a person reviews every finding before it goes into the report.

02Why does an app built with AI need an audit?

An AI tool writes code that works, but rarely tests whether it is secure. Typical flaws are API keys in the code, missing permission checks and input that reaches the database without checks. In 2025, Veracode found a security flaw in 45% of the tests of code written by AI models.

03Which AI models do you use?

Language models that are strong at code and security work, and local models where needed, so your code does not have to leave the network. Which model we use depends on the assignment. The models do the search work; a person decides what counts as a real finding.

04Do you get access to my production data?

No, that is not needed. We work with read access to the code and the configuration. If we find a secret such as an API key or password, we report it right away and advise you to replace it.

05How much does a security audit cost?

You always get a custom quote. The price depends on the size of the code, the number of integrations and environments, and whether you want help with the repairs. We define the scope together in the first conversation.

06Is my app 100% secure after an audit?

No. No audit can promise that. An audit finds the flaws that are in the code and configuration now, which makes the risk a lot smaller. After major changes or new AI features, a new check makes sense. General advice on making your organization more resilient is available from the NCSC.

Read more

Which app do you want reviewed?

Discuss your pilot