Lovable apps expose passport details and 200,000 chat messages, researchers find

Lovable apps expose passport details and 200,000 chat messages, researchers find

Security researchers, working with the German paper Handelsblatt, examined 159,041 projects built on Lovable, the Swedish AI platform that turns plain-language descriptions into web applications. They found many cases where personal and confidential information could be read without logging in. Handelsblatt reporter Lina Knees published a German-language article on the findings on November 2, 2026; the research report carries a more recent analysis and updated figures.

Lovable was founded in Stockholm in 2023 by Anton Osika and Fabian Hedin. It grew out of Osika's open-source project GPT Engineer, which became the GPT Engineer App and was renamed Lovable in late 2024, launching publicly that November. The researchers point to a gap. Lovable's marketing repeatedly says no technical knowledge is needed, which can suggest the platform handles security too, yet its Terms & Conditions place responsibility for applications and data handling on customers.

The method: the team used datasets from the open web archive Common Crawl to identify 159,041 Lovable projects from December 2025 to September 2026. They focused on the database service Supabase, which they believe is the most widely used database among Lovable projects. Of the projects, 34,545 used Supabase. Of those, 24,664 websites sent Supabase database records as soon as they loaded, with no login or other interaction. That alone is not a security problem, since the data may be meant for public display. But by making additional read-only requests with the same API key, the researchers retrieved more database content. They found records containing email addresses in 10,560 tables across 7,403 of the 24,664 websites, or 30%.

The examples they list: a German lifestyle forum exposed more than 80,000 profiles with full names, email addresses, dates of birth and gender, and more than 200,000 chat messages were also accessible. Several event organizers and restaurants exposed tens of thousands of records on ticket and food sales, including email addresses, phone numbers, postal addresses, payment information and ticket codes. A property management company exposed AI-generated summaries of internal emails with sensitive details about tenants, property owners and business partners. In what the researchers call a particularly serious case, a service provider exposed more than 1,000 records with identity document details: passport numbers, expiration dates and nationalities. Other accessible content included tens of thousands of contracts, invoices, personal photo albums and internal WhatsApp communications, plus numerous newsletter subscriptions, waiting lists, and contact and support inquiries. In one case, 280,000 emails sent to a provider's customers were accessible.

The researchers say exposed data like this could be used for phishing, ticket fraud or identity theft, but they could not establish whether any of it had been misused. They did not download the datasets in full, so the affected datasets may hold more than the figures show, and they are not publishing details of how to access the data. Handelsblatt contacted Lovable with the findings.

Their conclusion: access control mistakes can happen in any application, whether written by people or built with AI. With Lovable, though, database integrations follow recurring patterns, so an approach that works against one app may work against others with little extra effort. Users without a technical background also find such flaws harder to spot and prevent. The report recommends reviewing access permissions, interfaces and separation between user accounts before an app handles real customer or business data, and repeating the checks after significant changes.

Key facts

  • Researchers analysed 159,041 Lovable projects found in Common Crawl (December 2025 to September 2026); 34,545 used Supabase and 24,664 sent database records on page load.
  • Records with email addresses turned up in 10,560 tables across 7,403 of those 24,664 websites (30%), a figure the authors say is not a confirmed leak rate.
  • One German lifestyle forum exposed more than 80,000 profiles and more than 200,000 chat messages; a service provider exposed more than 1,000 records with passport numbers, expiration dates and nationalities.
  • Other exposed content included tens of thousands of ticket and food sales records, contracts, invoices, photo albums, internal WhatsApp communications and, in one case, 280,000 emails to a provider's customers.
  • The researchers could not establish whether any exposed data was misused; Lovable's Terms & Conditions put responsibility for data handling on customers.

Why it matters

AI platforms promise that anyone can build a working web app from a plain-language description. This study shows what can follow when the apps handle real personal data: sensitive records readable without a login, including passport details. The researchers stress that access control mistakes happen in any application, human-written or AI-built. Their specific worry about Lovable is that its database integrations follow recurring patterns, so an approach that works against one app may work against others with little extra effort. They also point to a gap between marketing that says no technical knowledge is needed and Terms & Conditions that make customers responsible for data handling.

Who it affects

First, the people whose data sat in these apps: forum members, event and restaurant customers, tenants and property owners, and people whose identity document details were exposed. Second, the non-technical builders who use Lovable to create apps that collect such data and who, per the report, find it harder to spot and prevent these flaws. Third, businesses that put customer or business data into AI-built tools. The affected apps in the report include a German lifestyle forum, event organizers, restaurants, a property management company and a service provider.

How to use it

The report's advice is for anyone building on an AI app platform. Before an application handles real customer or business data, have its access permissions, interfaces and separation between user accounts reviewed. Repeat the checks after significant changes. Automated checks are no substitute for a professional security assessment, and if the expertise is not available in-house, bring in qualified support. In the authors' words: "An easy-to-use development platform does not remove the need for an expert review of access permissions before an application handles confidential data."

How solid is it

The report is specific about its method: Common Crawl datasets, Supabase-backed projects, and additional read-only requests with the same API key. It was produced in collaboration with Handelsblatt, which published its own article on November 2, 2026; this report has more recent figures. No affected company or app is named, so the individual cases cannot be checked from the text. The researchers' own organisation is not named in the article text, and the report ends with a call to get in touch for help securing applications. Lovable's response is not given; the source says only that Handelsblatt contacted Lovable with the findings.

Risks and caveats

The authors say the 30% figure should not be read as a confirmed data leak rate. Email addresses may be public contact addresses or test data, while records without emails are missed; these effects partly offset each other, but the extent is unknown. The 30% applies only to the 24,664 websites examined more closely. The Common Crawl sample is not representative of all Lovable applications. Only parts of the datasets were retrieved, so affected datasets may contain more than the reported figures. The more than 200,000 chat messages belong to a single forum. Lovable's own database features, introduced in late 2025, were not examined. The researchers could not establish whether any exposed data was misused. The source does not say whether the exposed data has since been secured.

“An easy-to-use development platform does not remove the need for an expert review of access permissions before an application handles confidential data.”

— Report authors, crissyfield.de