German security firm Crissyfield says it examined 159,041 apps built on Lovable, the Swedish AI app builder, and found many cases where personal and confidential data could be read without logging in. Of the 24,664 sites it examined more closely, it found records containing email addresses in 10,560 tables across 7,403 sites, or 30%.
By our arithmetic, that is 4.7% of all 159,041 projects in the sample. Crissyfield says the 30% “should therefore not be read as a confirmed data leak rate”, and that it “could not establish whether any of it had been misused”. Its report was the top story on Hacker News on 14 November.
Business Pill · The Model and the Harness
An evergreen Business Engineer explainer on how the harness around a model shapes what an agent does. It teaches the general idea only and does not discuss Lovable.
The key insight: As we read the report, the exposure sits in configuration rather than in Lovable itself: Crissyfield traces it to Supabase databases that sent records on page load, and it notes Lovable’s terms place responsibility for data handling on customers.
Crissyfield says it used the open web archive Common Crawl to find 159,041 Lovable projects from December 2025 to September 2026, of which 34,545 used the database service Supabase. It says 24,664 of those sites “sent database records from Supabase as soon as they loaded, without requiring a login or any other interaction”, and that further read-only requests with the same API key returned more content.
By our arithmetic, 21.7% of the projects used Supabase and 71.4% of those sent records on page load. Crissyfield says that alone “is not a security issue”, because the data may be meant for public display. It did not examine Lovable’s own database feature, introduced in late 2025, and says its Common Crawl sample “is also not representative of all Lovable applications”.
Funnel from Crissyfield’s Lovable report: 159,041 projects from Common Crawl (Dec 2025 to Sep 2026), 34,545 on Supabase, 24,664 sending records on page load, 7,403 with email-containing tables readable without login (30% of 24,664; not a confirmed leak rate, per Crissyfield).
The Exposed Records
Among the cases it lists is a German lifestyle forum that exposed more than 80,000 profiles and more than 200,000 chat messages. Event organizers and restaurants exposed tens of thousands of ticket and food-sale records, including payment information and ticket codes, and one service provider exposed more than 1,000 records with passport numbers, expiration dates and nationalities.
It also reports accessible contracts, invoices, personal photo albums and internal WhatsApp messages, and one case where 280,000 emails sent to a provider’s customers could be read. Crissyfield names none of the apps and says it is not publishing how to reach the data, “as doing so could make further access easier”.
Who Is Responsible
Crissyfield notes that Lovable’s marketing says no technical knowledge is needed to build apps, while “Lovable’s Terms & Conditions place responsibility for applications and data handling on the customers themselves”. It argues that “database integrations follow recurring patterns, making it easier for attackers to look for similar misconfigurations across many applications”.
The research was done with Handelsblatt, which reported an earlier version on 2 November and, the report says, contacted Lovable with the findings. We found no published Lovable response and did not contact the company. A different study this month found Grok Build left 94.7% of 37,571 live apps open, using a different method on a different platform.
The Structural Read
As we read the numbers, the headline 30% is a rate inside a narrowed sample: it covers the 24,664 sites that already sent records without login, which is 15.5% of the 159,041 projects, and Crissyfield says the figure is not a confirmed leak rate.
As we read the recurring-patterns point, Crissyfield’s concern is scale: it says similar database setups make it easier to look for the same misconfiguration across many apps.
Crissyfield research report on Lovable, November 2026
“Automated checks are no substitute for a professional security assessment.”
Three Implications
FOR APP BUILDERS Crissyfield says access permissions, interfaces and separation between user accounts should be reviewed before an app handles real customer or business data, and again after significant changes.
FOR PLATFORMS Crissyfield says database integrations on Lovable follow recurring patterns, which it says makes similar misconfigurations easier to find across many applications.
FOR USERS OF THESE APPS Crissyfield says exposed data like this could be used for phishing, ticket fraud or identity theft, and that it could not establish whether any had been misused.
The framework argues that “The surface, not the model, governs autonomy”, meaning the harness around a model decides what an agent can do.
As we read it, Crissyfield’s findings point the same way for app builders: the exposure it reports comes from the database setup around generated apps, not from the AI model. The framework does not discuss Lovable.
What Is Not Established
Crissyfield says email addresses can be public contact details or test data, and that its method misses personal records without an email address, so the 30% is not a confirmed leak rate. It also could not tell whether any of the data had been misused. Its figures cover only the records it downloaded, which may be part of larger datasets.
This Week In The AI Supercycle
AI Lost Its Central Banker, the Lab Oligopoly, and the Future of Open Source
This week’s issue: the Nvidia leadership change, read as the AI build-out losing its central banker: who now finances the GPU landlords, backstops the data centres and sponsors open models, and why two labs are becoming the bank.
Crissyfield says that of 24,664 Lovable apps it examined closely, 7,403 left tables with email addresses readable without a login, including one with more than 1,000 passport records, and it calls automated checks “no substitute for a professional security assessment”.
A note on sourcing. The figures, cases and caveats come from Crissyfield’s English research report on Lovable, read in full on 14 November 2026; the report itself notes it updates the figures in Handelsblatt’s 2 November article. Percentages beyond Crissyfield’s 30% are our arithmetic. We did not contact Lovable or Crissyfield for this piece.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
