AI
“Apps built with the AI app-maker Lovable left passport numbers, payment information and 200,000 chat messages open to anyone, no login needed. German security firm Crissy Field scanned 159,041 Lovable projects and found databases with email addresses readable on 7,403 sites. Lovable promises no coding skills are needed. Its terms make…”
Plain restatementA German IT security firm, Crissy Field, working with Handelsblatt, analysed 159,041 websites and apps built on the Lovable platform and reported that in many of them database content including identity document details, payment-related records and chat messages could be read without logging in, with email-containing records found on 7,403 sites. Lovable's position, as reported, is that this is an industry-wide problem and that customers are contractually responsible for their own apps and data.
This post is mostly accurate. A German IT security company called Crissy Field, working with the business newspaper Handelsblatt, published a report on 2 November 2026 that identified 159,041 websites and apps built with the AI tool Lovable and found that on 7,403 of them database records containing email addresses could be read without logging in. The report does describe exposed passport numbers, payment-related records from ticket and food sales, and more than 200,000 chat messages, so the post's headline items are real and the numbers are copied correctly. The main thing the post leaves out is scale within the findings: the passport case was one service provider with just over a thousand identity records, and the chat messages came from one forum, while the broad finding is the 7,403 sites with readable email data. The researchers also warned that their 30 percent figure is not a confirmed leak rate and applies only to the subset of about 24,664 sites they probed further, and they said they could not establish whether any of the data was actually misused. One wording point matters: the exposure came from apps that customers built on the platform with their database access rules left open, not from a hack of Lovable's own systems, and the post's graphic blurs that more than its caption does. It is not clear from the report how many of these sites have since been fixed.
[drifted from the evidence:] Apps built with the AI app-maker Lovable left passport numbers, payment information and 200,000 chat messages open to anyone, no login needed. German security firm Crissy Field [drifted from the evidence:] scanned 159,041 Lovable [drifted from the evidence:] projects and [drifted from the evidence:] found databases with [drifted from the evidence:] email addresses readable on 7,403 sites. [drifted from the evidence:] Lovable promises no coding skills are needed. Its terms make customers responsible for their apps and data. [drifted from the evidence:] Lovable told German business daily Handelsblatt the problem is industry-wide." (Image headline: "APPS BUILT WITH AI TOOL LOVABLE LEAKED PASSPORTS, PAYMENTS AND PRIVATE CHATS")
[added by the neutral restatement:] A German [added by the neutral restatement:] IT security firm, Crissy Field, [added by the neutral restatement:] working with Handelsblatt, analysed 159,041 [added by the neutral restatement:] websites and apps built on the Lovable [added by the neutral restatement:] platform and [added by the neutral restatement:] reported that in many of them database content including identity document details, payment-related records and chat messages could be read without logging in, with [added by the neutral restatement:] email-containing records found on 7,403 sites. [added by the neutral restatement:] Lovable's position, as reported, is that this is an industry-wide problem and that customers [added by the neutral restatement:] are contractually responsible for their [added by the neutral restatement:] own apps and data.
Red-tinted words in the claim drifted from the evidence. Green-tinted words are what a neutral restatement needs.
The trace / claim to source
- A real, named, published research report underlies the post. Crissy Field GmbH is a real Hamburg-registered German company whose stated business includes software development and data security consulting.
- The headline numbers are reproduced correctly from that report: 159,041 Lovable projects identified and analysed, and email-containing records found on 7,403 websites.
- Passport numbers were reported as exposed. The report describes more than 1,000 records containing identity document details, including passport numbers, expiration dates and nationalities, exposed by one service provider.
- Payment information was reported as exposed, in records from several event organizers and restaurants covering ticket and food sales, where inspected records included email addresses, phone numbers, postal addresses, payment information and ticket codes.
- More than 200,000 chat messages were reported accessible, in the context of a German lifestyle forum that also exposed more than 80,000 profiles.
- "No login needed" is the report's own framing: the researchers set out to find out whether sensitive data in Lovable applications was accessible without logging in.
- The marketing-versus-terms contrast in the post is taken directly from the report and is stated there in those terms.
- The "industry-wide" attribution is supported by Handelsblatt's own summary of the company's position.
- The caption lists passport numbers, payment information and 200,000 chat messages as one undifferentiated harvest from "apps built with Lovable," which reads as though each category was widespread across the scanned population. In the report, the passport finding is one service provider with more than 1,000 identity records, and the 200,000 chat messages sit in the description of a single German lifestyle forum. The wide-scale finding is the 7,403 sites with readable email-containing records; the most alarming items are individual cases.
- "scanned 159,041 Lovable projects" is correct as the identification step, but the deeper read-only probing that produced the 7,403 figure applied to the 24,664 sites that returned database records on load, out of 34,545 that used Supabase. The researchers explicitly warn that the 30% figure should not be read as a confirmed data leak rate and applies only to the 24,664 websites examined more closely. A reader of the post alone would reasonably infer a rate against the 159,041 base.
- Capability extrapolation, in the direction of agency: the image headline "AI TOOL LOVABLE LEAKED PASSPORTS, PAYMENTS AND PRIVATE CHATS" shifts the leak onto the platform itself. The evidence describes data exposed by customer-built applications whose database access rules were left open, not a compromise of Lovable's own systems. The caption's own wording, "apps built with the AI app-maker Lovable," is the accurate formulation; the graphic is looser than the caption it sits above.
- Omitted qualifier, on harm: the post conveys open exposure without the report's own statement that the researchers could not establish whether any of the data had been misused.
- Whether the 200,000 chat messages were private one-to-one messages, as the image headline's word "private" implies, or forum or support-style messages. The retrieved report text says only that they were chat messages and that they were accessible.
- What "payment information" means in the exposed ticket and food-sales records: the retrieved text lists it alongside addresses and ticket codes but does not specify whether full card data, partial card data or transaction metadata was involved.
- Whether the specific exposures remain live as of 3 November 2026. The crawl window ran December 2025 to September 2026, and the report does not state, in the text retrieved, which site operators were notified or how many have since fixed their access rules.
- Lovable's full response in its own words. The only retrieved formulation of its position is Handelsblatt's summary that the company calls this an industry-wide problem, plus a paraphrase of a Lovable security lead describing security as a shared responsibility.
- Whether the exposure rate found is materially higher, lower or comparable for apps built on competing AI app builders. Handelsblatt reports its own further research found the problem extends beyond this one provider, but no comparative figure was retrieved.
The research report exists, is published by the named firm, and says substantially what the post says it says. On scale and method: using datasets from the open web archive Common Crawl, the researchers identified and analyzed 159,041 Lovable projects from December 2025 to September 2026, of which 34,545 used Supabase. 24,664 websites sent database records from Supabase as soon as they loaded, without requiring a login or any other interaction; the report notes that alone is not a security issue because the data may be intended for public display, but by making additional read-only requests the researchers retrieved more database content, finding records containing email addresses in 10,560 tables across 7,403 of the 24,664 websites, or 30%. On the specific data types in the claim: a German lifestyle forum exposed more than 80,000 profiles containing full names, email addresses, dates of birth and gender information, and more than 200,000 chat messages were also accessible. Several event organizers and restaurants exposed tens of thousands of records relating to ticket and food sales with no login required, and the records inspected included email addresses, phone numbers, postal addresses, payment information and ticket codes. In what the report calls a particularly serious case, a service provider exposed more than 1,000 records containing identity document details, including passport numbers, expiration dates and nationalities. The report also describes numerous newsletter subscriptions, waiting lists and contact and support inquiries, including one case where 280,000 emails sent to a provider's customers were accessible, and states that such exposed data could be used for phishing, ticket fraud or identity theft, while the researchers could not establish whether any of it had been misused. On the marketing-versus-terms point the post makes: the report states that in its marketing Lovable repeatedly emphasizes that no technical knowledge is needed to build projects, which can give the impression that the platform takes care of security too, yet Lovable's Terms and Conditions place responsibility for applications and data handling on the customers themselves, leaving what the report calls a troubling gap. On Lovable's response: Handelsblatt reports that tens of thousands of websites built with the platform apparently have sensitive user data openly accessible, based on IT security experts' analyses provided exclusively to the paper, and that data thieves could have accessed sensitive Lovable user data through the gap for at least a year. Handelsblatt says it received tips about the security problem in mid-August, checked it, and confronted Lovable with it, and that further reporting shows the problem goes beyond this one provider. The paper's own framing of the company's answer is that, according to Lovable, this is an industry-wide problem. Important limitation stated by the researchers themselves: the approach misses personal records that contain no email address, these effects partly offset one another to an extent the researchers could not establish, and the 30% figure should therefore not be read as a confirmed data leak rate, applying only to the 24,664 websites examined more closely. The report also notes that Lovable's own database features, introduced in late 2025, are still less widely used, that similar misconfigurations may occur there too, and that they were not examined in this analysis. The report carries a note that it presents a more recent analysis and updated figures compared with the Handelsblatt article published on November 2, 2026. Background, consistent but separate: the same root cause was documented in 2025, when a vulnerability was recorded in deployed Lovable-generated projects due to default insufficient row level security policies on client-controlled direct database requests, and a crawl of 1,645 Lovable-powered projects found 303 insecure endpoints across 170 sites, or 10.3%. Lovable's current documentation says it sets up basic row level security policies automatically but that users should review and adjust them early in development.
Complete reasoning
The reply is formatted for pasting into the thread where the claim is circulating.