Security ·
Lovable: AI-built apps exposed passport numbers and over 200,000 private messages
Security firm Crissy Field, working with Handelsblatt, found that public apps built with Lovable let anyone read database records without logging in, including identity-document details and more than 200,000 private messages.
Lovable is a popular tool that turns a text prompt into a complete web app and connects it to a Supabase database. According to research by the German security firm Crissy Field, which reached about 750 points on Hacker News, and reports by outlets such as Cybernews, many public apps built this way let anyone read database contents without logging in. The exposed records included identity-document details such as passport numbers and more than 200,000 private messages.
The cause is not new. Supabase protects individual tables with Row Level Security rules. If a rule is missing or wrong, the public key that ships in the page code is enough for anyone to pull the whole table. The same pattern was found in Lovable apps before and was tracked as CVE-2025-48757. AI-generated apps look and behave correctly on the surface, so missing database protection goes unnoticed until a stranger finds it.
Precision matters here: the research covers public apps connected to Supabase, not every Lovable project. It is also unknown whether the exposed data was misused. The problem is not that AI writes code, but that nobody took ownership of who may read the data.
What should you do if an AI tool built your app? Open the database and confirm that every table holding personal or business data has an access rule that is switched on and tested. Try to reach the data without logging in, using only the public key. Check file storage too, because uploaded IDs and contracts often sit there. If you find a leak of personal data, GDPR generally requires you to notify the supervisory authority within 72 hours.
What it means for companies
An AI-built prototype is great for testing an idea, but it should not go live with customer or employee data without a security review. A company should do three things now. First, find out which internal tools and websites staff generated with AI and where their data lives. Second, have those apps checked as part of cyber security and regular website maintenance. Third, for apps with sensitive data, choose app development with human review of code and access rules, and EU hosting. Prahapp, the AI division of Buinsoft Technology in Prague, builds a free working prototype within 3 working days at a fixed price agreed upfront.
Sources: Crissy Field research, Cybernews