Lovable Apps Exposed Passport Numbers and 200,000 Chat Messages in Security Investigation

A security investigation by German cybersecurity firm Crissy Field has identified sensitive information exposed through thousands of applications built with Lovable, the AI-powered app development platform. Researchers examined 159,041 projects and found that 7,403 websites had database records containing email addresses accessible without logging in. Exposed information in individual cases included passport details, payment-related records, customer information, and more than 200,000 chat messages.

What Happened to Applications Built With Lovable?

The investigation was conducted by Hamburg-based security firm Crissy Field in collaboration with German business newspaper Handelsblatt.

Researchers examined applications created using Lovable, a platform that allows users to build websites and full-stack applications using natural language instructions rather than traditional coding.

The investigation identified security problems in applications connected to Supabase, a popular backend database service used by many developers.

In several cases, database access permissions were incorrectly configured, allowing visitors to retrieve information that should have been restricted to authorized users.

Importantly, the investigation did not report a compromise of Lovable’s own infrastructure. Instead, it identified security weaknesses in individual applications built on the platform.

How Many Lovable Applications Were Affected?

Crissy Field analyzed 159,041 Lovable projects identified through Common Crawl, an open web archive covering the period from December 2025 to September 2026.

The researchers reported the following findings:

  • 159,041 Lovable projects were identified and analyzed.
  • 34,545 projects were using Supabase.
  • 24,664 websites returned Supabase database records when their pages loaded.
  • 7,403 websites exposed database records containing email addresses.
  • The exposed records appeared across 10,560 database tables.

The researchers found email-containing records on approximately 30% of the 24,664 websites they examined more closely.

However, this does not mean that 30% of all Lovable applications were compromised. Crissy Field noted that its sample was not representative of every Lovable application and that some email addresses could have been public contact information or test data.

What Sensitive Information Was Exposed?

The investigation identified several examples of applications exposing personal and confidential information.

Were Passport Numbers and Identity Details Exposed?

In one particularly serious case, a service provider exposed more than 1,000 records containing identity document information.

The accessible records included passport numbers, expiration dates, and nationalities.

This type of information can create significant privacy and identity theft risks if accessed or misused by unauthorized individuals.

Were More Than 200,000 Chat Messages Accessible?

Researchers also discovered a German lifestyle forum that exposed more than 80,000 user profiles.

The profiles included full names, email addresses, dates of birth, and gender information.

More than 200,000 chat messages associated with the forum were also accessible.

Did Restaurants and Event Organizers Expose Customer Information?

Several restaurants and event organizers had database records accessible without authentication.

The exposed information included:

  • Customer email addresses and phone numbers.
  • Postal addresses.
  • Food and ticket orders.
  • Payment-related information.
  • Ticket codes and transaction records.

This could create risks for customers, including phishing, fraudulent ticket usage, and targeted scams.

What Other Business Information Was Accessible?

The investigation also identified exposed business documents and internal communications.

Examples included contracts, invoices, personal photo albums, internal WhatsApp communications, and AI-generated summaries of internal emails containing sensitive information about tenants, property owners, and business partners.

In another case, researchers found approximately 280,000 emails sent to a service provider’s customers.

Crissy Field stated that it could not establish whether the exposed information had actually been misused.

Why Did These Security Problems Occur?

According to Crissy Field, many of the issues were related to missing or incorrectly configured database access permissions.

Lovable makes application development easier by generating much of the underlying code automatically. However, applications that handle customer information still require properly configured authentication, authorization, and database security.

One important component is Row Level Security (RLS), a feature available in Supabase that controls which database records users can access.

For example, consider an application that allows customers to create accounts and manage their personal information.

Without correctly configured access policies, one user might be able to retrieve records belonging to another user.

A website can look completely functional while its backend database remains accessible to unauthorized visitors.

This is why security testing must cover both the visible website and the underlying database.

What Was Lovable’s Response?

According to the reporting by Handelsblatt, Lovable was contacted about the findings.

The company’s head of security described the issue as a known problem across the industry and emphasized that application security is a shared responsibility.

The broader issue is not limited to AI-powered development platforms. Similar database configuration mistakes can occur in traditionally developed applications.

However, AI app builders introduce another challenge: people without extensive software engineering experience can now launch applications that collect and process real customer information.

That makes security reviews particularly important before an application goes live.

How Can Developers Secure Their Lovable Applications?

Developers should not assume that an application is secure simply because it works correctly or passes basic functionality tests.

Before collecting real customer information, review the following areas:

1. Configure Supabase Row Level Security

Enable RLS on tables containing private information and create policies that restrict access based on authenticated users and their permissions.

2. Review Public Database Access

Check whether anonymous visitors can retrieve customer records, internal documents, orders, or other restricted information.

3. Protect API Keys and Secrets

Never expose service-role keys, private API credentials, database passwords, or other privileged secrets in frontend code.

4. Test Authentication and Authorization

Test the application using different accounts and unauthenticated sessions. Confirm that users can only access information they are authorized to view.

5. Audit Database Permissions Regularly

Review database policies whenever new features, tables, integrations, or user permissions are introduced.

6. Conduct Independent Security Testing

For applications handling passports, financial information, customer records, or confidential business documents, arrange a professional security assessment before launch.

Automated security tools can help identify common mistakes, but they should not replace a proper review.

What Does This Mean for the Future of AI App Builders?

The investigation highlights an important challenge for the growing AI development industry.

Platforms such as Lovable make it possible to build sophisticated applications without writing large amounts of code. This lowers development costs and gives entrepreneurs, agencies, and small businesses access to capabilities that previously required dedicated development teams.

However, faster development does not automatically mean safer development.

As AI-generated applications become more common, security, database permissions, authentication, and data protection must become part of the development process.

For businesses, the lesson is straightforward: an AI-generated application should go through the same security checks as any professionally developed software before handling sensitive information.

Final Thoughts

Lovable and other AI app builders are changing how quickly people can turn ideas into working applications. But the Crissy Field investigation shows why application security cannot be overlooked.

The exposed information was not simply a technical configuration issue. It involved real customer records, identity documents, business communications, and potentially sensitive personal information.

AI can accelerate development, but it cannot remove the responsibility to protect the people who use the applications.

As AI-powered software development continues to grow, security testing should be treated as a core part of launching any application, not something to consider after a problem occurs.


0
Show Comments (0) Hide Comments (0)
Leave a comment

Your email address will not be published. Required fields are marked *