Lovable AI Apps Left Passport and Payment Data Open to Anyone

A scan by German security research group Crissy Field, conducted with business daily Handelsblatt and first reported November 2, found that 24,664 of 34,545 Lovable-built projects running on Supabase returned database records with no authentication required. The team pulled 159,041 Lovable projects from the Common Crawl web archive, covering apps built between December 2025 and September 2026, and found email addresses readable on 7,403 sites across more than 10,500 tables, over 1,000 passport records with expiration dates and nationalities from one service provider's database, and more than 200,000 chat messages plus 80,000 user profiles leaked from a single lifestyle forum. The findings point to Lovable's AI agent generating working apps without enabling Supabase's Row Level Security, while Supabase's public "anon" key grants broad read access unless actively restricted.

A German security firm scanned more than 159,000 apps built with the AI coding platform Lovable and found passport numbers, payment details and 200,000 private chat messages sitting in databases that needed no login to open. You don't need to hack anything when the door was never locked. That's the story behind a new scan from the German research group Crissy Field, working with the business daily Handelsblatt, first reported November 2. The team pulled 159,041 Lovable projects out of the Common Crawl web archive, covering apps built between December 2025 and September 2026, and checked what their databases would hand over to a stranger with no credentials at all. The answer, for thousands of them, was everything. Of the 34,545 projects running on Supabase as their backend, 24,664 returned database records with zero authentication required. Email addresses alone showed up readable on 7,403 sites, spread across more than 10,500 tables. One service provider's database coughed up over 1,000 passport records complete with expiration dates and nationalities. A lifestyle forum built on Lovable leaked more than 200,000 chat messages and 80,000 user profiles to anyone who asked. Lovable markets itself as the tool that lets anyone, no engineer required, describe an app in plain English and watch an AI agent build it, wire up a database and ship it live. That pitch has made it one of the fastest-growing "vibe coding" platforms in the business, and it's exactly why this scan matters more than a routine vulnerability disclosure. The people hit hardest by an open database are the ones who had no idea a database needed locking in the first place. Supabase, the backend Lovable defaults to, ships with a security feature called Row Level Security, or RLS, that's supposed to make sure a user can only see their own rows in a shared table. It has to be turned on and configured correctly to do anything. Crissy Field's findings point to an AI agent that, left to its own defaults, generates working apps without bothering to lock that door, while Supabase hands its public "anon" key broad read access unless someone actively restricts it. Vibe-coded apps are turning startup speed into security debt https://startupfortune.com/vibe-coded-apps-are-turning-startup-speed-into-security-debt/ Thousands of vibe-coded apps reportedly exposed corporate and personal data on the open web. The bigger issue is the security debt created when nontechnical teams use AI coding tools to publish software faster than they understand access controls, secrets, and deployment defaults. - AI coding tools security vulnerabilities in startup applications https://startupfortune.com/vibe-coded-apps-are-turning-startup-speed-into-security-debt/ - how vibe coding creates accidental data exposure risks https://startupfortune.com/vibe-coded-apps-are-turning-startup-speed-into-security-debt/ A single missed checkbox is a bug. The same missed checkbox appearing across thousands of independently built apps is a pattern baked into how the tool teaches itself to ship fast. Every one of those 7,403 exposed sites was built by a different person typing a different prompt, yet they converged on the identical failure, because the AI model generating their backend code converged on it first. That's not 7,403 separate mistakes by 7,403 careless founders. That's one mistake, repeated at scale by the platform that writes the code none of them read. The data behind the headline figure spans far more than passports. Crissy Field's scan turned up tens of thousands of ticket bookings and restaurant orders carrying home addresses and payment information, plus AI-generated email summaries, contracts, invoices and WhatsApp logs sitting in business tools built the same way. One CRM-style project alone exposed more than 280,000 customer emails, waiting-list entries and support tickets. The people exposed never chose this risk Lovable's own users are rarely the ones with anything to lose here. They're builders who trusted a tool that promised them they didn't need to understand databases to run one safely. The actual exposure falls on their customers: the traveler whose passport sat in a travel-booking app's open table, the diner whose card details rode along with a restaurant order, the stranger whose private messages became one of the 200,000 readable on a forum they signed up for in good faith. None of them chose to accept that risk. They just used an app, built by someone who used a tool that sold ease over safety. Lovable has faced scrutiny over data exposure before. A separate flaw tracked as CVE-2025-48757 led the company to add an automated Security Scanner and an AI-driven Security Reviewer meant to catch exactly this kind of misconfiguration before it ships. Crissy Field's scan, conducted on apps built well after those tools existed, suggests they aren't catching enough of it. A warning that still lets 7,403 sites go live with open databases isn't a safeguard. It's a disclaimer. The responsibility for that gap sits with the platform that writes the code, not with the person who typed a sentence and trusted it to come back secure. Also read: Revolut Shows Cheap Unlimited Tokens, Not Smarter AI, Is What Speeds Up Coding https://startupfortune.com/revolut-shows-cheap-unlimited-tokens-not-smarter-ai-is-what-speeds-up-coding/ • Mercado Bitcoin Loses $46 Million to Hackers Days After Binance Heist https://startupfortune.com/mercado-bitcoin-loses-46-million-to-hackers-days-after-binance-heist/ • Mistral Is Racing OpenAI and xAI on a Fraction of Their GPUs https://startupfortune.com/mistral-is-racing-openai-and-xai-on-a-fraction-of-their-gpus/ This article is posted in Technology News https://startupfortune.com/category/technology/ , check it out for more related stories. Join the discussion Open in the community → https://startupfortune.com/community/ Almost there. Sign in and your reply posts straight away.