What happened?

Security researcher Dor Zvi and his team at RedAccess scanned thousands of web apps built with AI coding platforms such as Lovable, Replit, Base44 and Netlify. They uncovered more than 5,000 apps that were publicly accessible without any login or protection.

Why it matters

About 40% of those apps contained sensitive information – medical records, financial data, corporate strategies and full chatbot logs. In some cases the apps even gave admin rights to anyone who visited the URL.

Real‑world examples

  • Hospital work schedules with doctors' personal IDs.
  • Company ad‑spending reports and go‑to‑market decks.
  • Retailer chatbot logs showing customer names and contacts.
  • Shipping firm cargo manifests and sales figures.
  • Phishing sites mimicking banks and retailers, all hosted on Lovable’s domain.

How researchers found the apps

RedAccess used simple Google and Bing queries that combined the AI platforms’ domain names with common search terms. Because the platforms host apps on their own domains, the URLs are easy to discover.

Company responses

Netlify did not reply. Replit, Lovable and Base44 pushed back, saying they need more time to review the findings. Replit’s CEO noted that public apps are “expected behavior” and can be made private with one click.

What to do next

Enterprises should audit any apps created with AI coding tools and move them to private hosting or add authentication immediately. Developers must treat AI‑generated code like any other code – test, review and secure it before publishing.