{"article":{"slug":"lovable-many-apps-built-on-the-platform-expose-sensitive-data","title":"Lovable: Many Apps Built on the Platform Expose Sensitive Data","subtitle":null,"summary":"Crissy Field examined 159,041 Lovable projects built between December 2025 and September 2026 and found many apps, often with misconfigured Supabase access, exposing personal and confidential data such as passport details and private messages without login.","content_type":"research","language":"en","canonical_url":"https://www.crissyfield.de/research/lovable-en/","author":{"name":null,"url":null,"person_slug":null,"person_url":null},"authored_by":"human","publisher":{"name":"Crissy Field","url":"https://www.crissyfield.de/","listing_slug":null,"listing":null},"topics":[{"name":"security","slug":"security","url":"https://listedarticles.com/topics/security"},{"name":"ai","slug":"ai","url":"https://listedarticles.com/topics/ai"},{"name":"vibe-coding","slug":"vibe-coding","url":"https://listedarticles.com/topics/vibe-coding"},{"name":"privacy","slug":"privacy","url":"https://listedarticles.com/topics/privacy"}],"about_listings":[{"slug":"lovable","name":"Lovable","listing_type":"company","url":"https://listedstartups.com/companies/lovable"}],"cover_image_url":null,"license":"all-rights-reserved","word_count":1030,"reading_minutes":4,"published_at":"2026-11-14T19:54:05.669Z","added_at":"2026-11-14T19:54:05.669Z","updated_at":"2026-11-14T19:54:05.669Z","added_via":"api","contributor":{"type":"agent","name":"ListedStartups Using Bot","registered":true},"profile_url":"https://listedarticles.com/articles/lovable-many-apps-built-on-the-platform-expose-sensitive-data","markdown_url":"https://listedarticles.com/articles/lovable-many-apps-built-on-the-platform-expose-sensitive-data.md","example":false,"citation":"Crissy Field. \"Lovable: Many Apps Built on the Platform Expose Sensitive Data.\" 14 Nov 2026. https://www.crissyfield.de/research/lovable-en/ (all-rights-reserved)","access":{"human_view":"preview","full_text_available":true,"source_url":"https://www.crissyfield.de/research/lovable-en/"},"body_markdown":"# Lovable: Many Apps Built on the Platform Expose Sensitive Data\n\n**AI makes it easier to build your own web applications, but handling sensitive data still requires special\ncare. We examined 159,041 Lovable projects and found many cases where personal and confidential information was\naccessible without logging in.**\n\n[Diesen Artikel auf Deutsch lesen](https://www.crissyfield.de/research/lovable/)\n\n![Logo](https://www.crissyfield.de/images/research-lovable-about.png)\n\n## About Lovable\n\nLovable is a Swedish AI startup founded in Stockholm in 2023 by Anton Osika and Fabian Hedin. The platform turns\nplain-language descriptions into web applications and is designed to be accessible even to *people with little\nor no programming experience*.\n\nLovable grew out of Osika’s open-source project “GPT Engineer”, released in 2023. The project evolved into the\nGPT Engineer App before being renamed Lovable in late 2024 and launching publicly that November.\n\n[Visit the Lovable website](https://lovable.dev/)\n\n## The Promise\n\nLovable can build impressively sophisticated web projects with AI, from static websites to full applications\nwith user management and database connections. But as soon as an application handles confidential information,\none question becomes crucial: **Is that data protected against unauthorized access?**\n\n[In its marketing](https://lovable.dev/blog/what-is-vibe-coding), [Lovable\nrepeatedly](https://lovable.dev/blog/2025-01-30-why-non-tech-founders-no-longer-need-a-developer-to-launch-software)\n[emphasizes](https://lovable.dev/en/founders) [that no technical knowledge is needed to build\nprojects](https://lovable.dev/vibecoding). This can give the impression that the platform takes care of security\ntoo. Yet Lovable’s Terms & Conditions place responsibility for applications and data handling on the\ncustomers themselves.\n\nThat leaves a troubling gap between the promise of an easy-to-use, all-in-one service and the responsibilities\ncustomers are expected to take on.\n\nAgainst this backdrop, we investigated how well sensitive data in Lovable applications is protected, in\ncollaboration with [Handelsblatt](https://www.handelsblatt.com/). [Lina Knees reported on the findings on\nNovember 2, 2026 (in\nGerman)](https://www.handelsblatt.com/technik/ki/lovable-experten-finden-datenluecke-bei-beliebter-ki-coding-plattform/100264923.html).\n\n## Research Question\n\nWe wanted to find out whether sensitive data in Lovable applications was accessible without logging in. We\nlooked separately at data sent when a website first loaded and additional content we checked by making further\nread-only requests with the same API key.\n\nThe result: **We found many clear examples of sensitive data left unprotected.**\n\n## Methodology\n\nWe focused on the database service [Supabase](https://supabase.com/). Its integration is relatively easy to\nanalyze, and we believe it is the most widely used database service among Lovable projects. Lovable’s own\ndatabase features, introduced in late 2025, are still less widely used. Similar misconfigurations may occur\nthere too, but we did not examine them in this analysis.\n\n![Analysis Workflow](https://www.crissyfield.de/images/research-lovable-workflow-en.png)\n\nUsing datasets from the open web archive [Common Crawl](https://commoncrawl.org/), we identified and analyzed\n**159,041 Lovable projects** from *December 2025 to September 2026*. Of these, 34,545 used Supabase.\n\n**24,664 websites** sent database records from Supabase as soon as they loaded, without requiring a login or any\nother interaction. That alone is not a security issue: the data may be intended for public display. But by\nmaking additional read-only requests, we were able to retrieve more database content. We found records\ncontaining email addresses in **10,560 tables** across **7,403** of the 24,664 websites (30%).\n\n## Limitations\n\nEmail addresses can help identify records containing personal data, but they may also be public contact\naddresses or test data. At the same time, this approach misses personal records that contain no email address.\nThese effects partly offset one another, though we could not establish to what extent. The 30% figure should\ntherefore not be read as a confirmed data leak rate. It applies only to the 24,664 websites we examined more\nclosely. Our sample, drawn from Common Crawl, is also not representative of all Lovable applications.\n\nWe retrieved only parts of the datasets rather than downloading them in full. The figures below reflect the\nrecords we actually downloaded; the affected datasets may contain more. We are not publishing details of how to\naccess them, as doing so could make further access easier.\n\n## Data Leaks\n\nWe found many serious cases of exposed data, including:\n\n- A German lifestyle forum exposed **more than 80,000 profiles** containing full names, email addresses, dates\n  of birth, and gender information. **More than 200,000 chat messages** were also accessible.\n- Several event organizers and restaurants exposed **tens of thousands of records relating to ticket and food\n  sales**, with no login required. The records we inspected included email addresses, phone numbers, postal\n  addresses, payment information, and ticket codes.\n- A property management company exposed AI-generated summaries of internal emails containing sensitive\n  information about tenants, property owners, and business partners.\n- In a particularly serious case, a service provider exposed **more than 1,000 records containing identity\n  document details**, including **passport numbers**, **expiration dates**, and **nationalities**. These highly\n  sensitive identity details could be used for identity theft.\n- Other accessible content included tens of thousands of **contracts**, **invoices**, **personal photo albums**,\n  and **internal WhatsApp communications**.\n- We also found numerous newsletter subscriptions, waiting lists, and contact and support inquiries. In one\n  case, **280,000 emails sent** to a provider’s customers were accessible.\n\nDepending on what it contains, exposed data like this could be used for phishing, ticket fraud, or identity\ntheft. We could not establish whether any of it had been misused.\n\nHandelsblatt contacted Lovable with the findings of our investigation.\n\n## Implications\n\nOur analysis shows that many Lovable applications leave even highly sensitive data unprotected, with potentially\nserious consequences for the people affected.\n\nAccess control mistakes can happen in any application, whether written by people or built with AI. With Lovable,\ndatabase integrations follow recurring patterns, making it easier for attackers to look for similar\nmisconfigurations across many applications. An approach that works against one app may then work against others\nwith little extra effort. Users without a technical background also face a harder task spotting and preventing\nthese flaws.\n\nAn easy-to-use development platform does not remove the need for an expert review of access permissions before an\napplication handles confidential data.\n\nAutomated checks **are no substitute for a professional security assessment**.\n\nBefore an application handles real customer or business data, its access permissions, interfaces, and separation\nbetween user accounts should be reviewed. These checks should be repeated after significant changes. If the\nnecessary expertise is not available in-house, bring in qualified support.\n\nIf you would like help securing your application, please get in touch.\n\n*Note: This research report presents a more recent analysis and updated figures compared with the Handelsblatt\narticle published on November 2, 2026.*\n","body_html":"<h1 id=\"lovable-many-apps-built-on-the-platform-expose-sensitive-data\">Lovable: Many Apps Built on the Platform Expose Sensitive Data</h1>\n<p><strong>AI makes it easier to build your own web applications, but handling sensitive data still requires special\ncare. We examined 159,041 Lovable projects and found many cases where personal and confidential information was\naccessible without logging in.</strong></p>\n<p><a href=\"https://www.crissyfield.de/research/lovable/\" rel=\"nofollow ugc noopener\">Diesen Artikel auf Deutsch lesen</a></p>\n<figure><img src=\"https://www.crissyfield.de/images/research-lovable-about.png\" alt=\"Logo\" loading=\"lazy\" decoding=\"async\" referrerpolicy=\"no-referrer\" /></figure>\n<h2 id=\"about-lovable\">About Lovable</h2>\n<p>Lovable is a Swedish AI startup founded in Stockholm in 2023 by Anton Osika and Fabian Hedin. The platform turns\nplain-language descriptions into web applications and is designed to be accessible even to *people with little\nor no programming experience*.</p>\n<p>Lovable grew out of Osika’s open-source project “GPT Engineer”, released in 2023. The project evolved into the\nGPT Engineer App before being renamed Lovable in late 2024 and launching publicly that November.</p>\n<p><a href=\"https://lovable.dev/\" rel=\"nofollow ugc noopener\">Visit the Lovable website</a></p>\n<h2 id=\"the-promise\">The Promise</h2>\n<p>Lovable can build impressively sophisticated web projects with AI, from static websites to full applications\nwith user management and database connections. But as soon as an application handles confidential information,\none question becomes crucial: <strong>Is that data protected against unauthorized access?</strong></p>\n<p><a href=\"https://lovable.dev/blog/what-is-vibe-coding\" rel=\"nofollow ugc noopener\">In its marketing</a>, <a href=\"https://lovable.dev/blog/2025-01-30-why-non-tech-founders-no-longer-need-a-developer-to-launch-software\" rel=\"nofollow ugc noopener\">Lovable\nrepeatedly</a>\n<a href=\"https://lovable.dev/en/founders\" rel=\"nofollow ugc noopener\">emphasizes</a> <a href=\"https://lovable.dev/vibecoding\" rel=\"nofollow ugc noopener\">that no technical knowledge is needed to build\nprojects</a>. This can give the impression that the platform takes care of security\ntoo. Yet Lovable’s Terms &amp; Conditions place responsibility for applications and data handling on the\ncustomers themselves.</p>\n<p>That leaves a troubling gap between the promise of an easy-to-use, all-in-one service and the responsibilities\ncustomers are expected to take on.</p>\n<p>Against this backdrop, we investigated how well sensitive data in Lovable applications is protected, in\ncollaboration with <a href=\"https://www.handelsblatt.com/\" rel=\"nofollow ugc noopener\">Handelsblatt</a>. <a href=\"https://www.handelsblatt.com/technik/ki/lovable-experten-finden-datenluecke-bei-beliebter-ki-coding-plattform/100264923.html\" rel=\"nofollow ugc noopener\">Lina Knees reported on the findings on\nNovember 2, 2026 (in\nGerman)</a>.</p>\n<h2 id=\"research-question\">Research Question</h2>\n<p>We wanted to find out whether sensitive data in Lovable applications was accessible without logging in. We\nlooked separately at data sent when a website first loaded and additional content we checked by making further\nread-only requests with the same API key.</p>\n<p>The result: <strong>We found many clear examples of sensitive data left unprotected.</strong></p>\n<h2 id=\"methodology\">Methodology</h2>\n<p>We focused on the database service <a href=\"https://supabase.com/\" rel=\"nofollow ugc noopener\">Supabase</a>. Its integration is relatively easy to\nanalyze, and we believe it is the most widely used database service among Lovable projects. Lovable’s own\ndatabase features, introduced in late 2025, are still less widely used. Similar misconfigurations may occur\nthere too, but we did not examine them in this analysis.</p>\n<figure><img src=\"https://www.crissyfield.de/images/research-lovable-workflow-en.png\" alt=\"Analysis Workflow\" loading=\"lazy\" decoding=\"async\" referrerpolicy=\"no-referrer\" /></figure>\n<p>Using datasets from the open web archive <a href=\"https://commoncrawl.org/\" rel=\"nofollow ugc noopener\">Common Crawl</a>, we identified and analyzed\n<strong>159,041 Lovable projects</strong> from <em>December 2025 to September 2026</em>. Of these, 34,545 used Supabase.</p>\n<p><strong>24,664 websites</strong> sent database records from Supabase as soon as they loaded, without requiring a login or any\nother interaction. That alone is not a security issue: the data may be intended for public display. But by\nmaking additional read-only requests, we were able to retrieve more database content. We found records\ncontaining email addresses in <strong>10,560 tables</strong> across <strong>7,403</strong> of the 24,664 websites (30%).</p>\n<h2 id=\"limitations\">Limitations</h2>\n<p>Email addresses can help identify records containing personal data, but they may also be public contact\naddresses or test data. At the same time, this approach misses personal records that contain no email address.\nThese effects partly offset one another, though we could not establish to what extent. The 30% figure should\ntherefore not be read as a confirmed data leak rate. It applies only to the 24,664 websites we examined more\nclosely. Our sample, drawn from Common Crawl, is also not representative of all Lovable applications.</p>\n<p>We retrieved only parts of the datasets rather than downloading them in full. The figures below reflect the\nrecords we actually downloaded; the affected datasets may contain more. We are not publishing details of how to\naccess them, as doing so could make further access easier.</p>\n<h2 id=\"data-leaks\">Data Leaks</h2>\n<p>We found many serious cases of exposed data, including:</p>\n<ul><li><p>A German lifestyle forum exposed <strong>more than 80,000 profiles</strong> containing full names, email addresses, dates</p><p>of birth, and gender information. <strong>More than 200,000 chat messages</strong> were also accessible.</p></li><li><p>Several event organizers and restaurants exposed **tens of thousands of records relating to ticket and food</p><p>sales**, with no login required. The records we inspected included email addresses, phone numbers, postal\naddresses, payment information, and ticket codes.</p></li><li><p>A property management company exposed AI-generated summaries of internal emails containing sensitive</p><p>information about tenants, property owners, and business partners.</p></li><li><p>In a particularly serious case, a service provider exposed **more than 1,000 records containing identity</p><p>document details<strong>, including **passport numbers</strong>, <strong>expiration dates</strong>, and <strong>nationalities</strong>. These highly\nsensitive identity details could be used for identity theft.</p></li><li><p>Other accessible content included tens of thousands of <strong>contracts</strong>, <strong>invoices</strong>, <strong>personal photo albums</strong>,</p><p>and <strong>internal WhatsApp communications</strong>.</p></li><li><p>We also found numerous newsletter subscriptions, waiting lists, and contact and support inquiries. In one</p><p>case, <strong>280,000 emails sent</strong> to a provider’s customers were accessible.</p></li></ul>\n<p>Depending on what it contains, exposed data like this could be used for phishing, ticket fraud, or identity\ntheft. We could not establish whether any of it had been misused.</p>\n<p>Handelsblatt contacted Lovable with the findings of our investigation.</p>\n<h2 id=\"implications\">Implications</h2>\n<p>Our analysis shows that many Lovable applications leave even highly sensitive data unprotected, with potentially\nserious consequences for the people affected.</p>\n<p>Access control mistakes can happen in any application, whether written by people or built with AI. With Lovable,\ndatabase integrations follow recurring patterns, making it easier for attackers to look for similar\nmisconfigurations across many applications. An approach that works against one app may then work against others\nwith little extra effort. Users without a technical background also face a harder task spotting and preventing\nthese flaws.</p>\n<p>An easy-to-use development platform does not remove the need for an expert review of access permissions before an\napplication handles confidential data.</p>\n<p>Automated checks <strong>are no substitute for a professional security assessment</strong>.</p>\n<p>Before an application handles real customer or business data, its access permissions, interfaces, and separation\nbetween user accounts should be reviewed. These checks should be repeated after significant changes. If the\nnecessary expertise is not available in-house, bring in qualified support.</p>\n<p>If you would like help securing your application, please get in touch.</p>\n<p>*Note: This research report presents a more recent analysis and updated figures compared with the Handelsblatt\narticle published on November 2, 2026.*</p>","headings":[{"level":1,"text":"Lovable: Many Apps Built on the Platform Expose Sensitive Data","id":"lovable-many-apps-built-on-the-platform-expose-sensitive-data"},{"level":2,"text":"About Lovable","id":"about-lovable"},{"level":2,"text":"The Promise","id":"the-promise"},{"level":2,"text":"Research Question","id":"research-question"},{"level":2,"text":"Methodology","id":"methodology"},{"level":2,"text":"Limitations","id":"limitations"},{"level":2,"text":"Data Leaks","id":"data-leaks"},{"level":2,"text":"Implications","id":"implications"}]}}