CVE-2025-48757Disclosure

HIGHCVSS 9.3 · CRITICAL

Exploitation observed; activity peaked at 20 mentions and remains active

Immediate actions

  • Patch affected systems immediately
  • Assume compromise if assets are exposed
  • Hunt for exploitation attempts and persistence artifacts
  • Increase monitoring for publicly documented tradecraft

Recommended action window: Immediate (within 24h)

NVD description

An insufficient database Row-Level Security policy in Lovable through 2025-04-15 allows remote unauthenticated attackers to read or write to arbitrary database tables of generated sites. NOTE: this is disputed by the Supplier because each individual customer of the Lovable platform accepts a responsibility over protecting the data of their application.

6.5/ 10 priority

Sources & remediation

Weakness type (CWE)
CWE-863

Priority

HIGH

Exploitation

ACTIVE

PoC

YES

Patch

AVAILABLE

Momentum

DECLINING

Threat summary

  • Active exploitation appears in 13 classified signals
  • Public PoC is present in monitored signal
  • Patch or workaround signal is available
  • 54 mentions across 25 observed days

What's happening

  • Active exploitation reported across 13 signals
  • PoC mentioned or linked in 1 signal
  • Patch or workaround mentioned in 5 signals
  • Technical details provided in 33 signals
  • Disclosure: 22 classified signals
  • General: 18 classified signals
  • Peaked 20d ago at 20 mentions (2026-04-20); latest day: 1
  • 54 total mentions across 25 days

Deep dive

Activity timeline54 mentions / 25d
05101520Mentions · 2026-02-02: 1Mentions · 2026-03-06: 3Mentions · 2026-04-01: 1Mentions · 2026-04-10: 1Mentions · 2026-04-20: 20Mentions · 2026-04-21: 4Mentions · 2026-05-02: 1Mentions · 2026-05-09: 1Mentions · 2026-05-27: 3Mentions · 2026-05-28: 1Mentions · 2026-06-02: 1Mentions · 2026-06-12: 1Mentions · 2026-06-30: 1Mentions · 2026-07-04: 1Mentions · 2026-07-28: 1Mentions · 2026-08-09: 1Mentions · 2026-08-17: 2Mentions · 2026-09-01: 1Mentions · 2026-09-02: 1Mentions · 2026-09-06: 1Mentions · 2026-09-14: 2Mentions · 2026-09-22: 2Mentions · 2026-09-25: 1Mentions · 2026-10-02: 1Mentions · 2026-10-05: 1PoC Mentioned / Linked · 2026-04-20: 1Active Exploitation · 2026-03-06: 3Active Exploitation · 2026-04-10: 1Active Exploitation · 2026-04-20: 1Active Exploitation · 2026-05-02: 1Active Exploitation · 2026-05-27: 2Active Exploitation · 2026-05-28: 1Active Exploitation · 2026-06-02: 1Active Exploitation · 2026-06-30: 1Active Exploitation · 2026-07-28: 1Active Exploitation · 2026-08-09: 1Patch / Workaround · 2026-04-20: 2Patch / Workaround · 2026-05-09: 1Patch / Workaround · 2026-05-28: 1Patch / Workaround · 2026-06-12: 1Technical Details · 2026-02-02: 1Technical Details · 2026-04-01: 1Technical Details · 2026-04-20: 12Technical Details · 2026-04-21: 3Technical Details · 2026-05-09: 1Technical Details · 2026-05-27: 2Technical Details · 2026-05-28: 1Technical Details · 2026-06-02: 1Technical Details · 2026-06-12: 1Technical Details · 2026-06-30: 1Technical Details · 2026-07-04: 1Technical Details · 2026-07-28: 1Technical Details · 2026-08-17: 1Technical Details · 2026-09-01: 1Technical Details · 2026-09-02: 1Technical Details · 2026-09-06: 1Technical Details · 2026-09-14: 2Technical Details · 2026-09-22: 102-0204-0104-2005-0205-2706-0206-3007-2808-1709-0209-1409-2510-05
Signal classification4 categories
Disclosure
2242.3%
General
1834.6%
Active Exploitation
1019.2%
Patch
23.8%
Referenced assets12 URLs
Classification over time
DateTotalLabels
2026-02-021
General1
2026-03-063
Active Exploitation3
2026-04-011
Disclosure1
2026-04-101
Active Exploitation1
2026-04-2020
Disclosure9General10Patch1
2026-04-214
Disclosure2General2
2026-05-021
Active Exploitation1
2026-05-091
Patch1
2026-05-273
Active Exploitation2General1
2026-05-281
Active Exploitation1
2026-06-021
Disclosure1
2026-06-121
General1
2026-06-301
Active Exploitation1
2026-07-041
Disclosure1
2026-07-281
Active Exploitation1
2026-08-091
General1
2026-08-172
Disclosure1General1
2026-09-011
Disclosure1
2026-09-021
Disclosure1
2026-09-061
Disclosure1
2026-09-142
Disclosure2
2026-09-222
Disclosure1General1
2026-09-251
Disclosure1
Full discourse20 posts
  • impulsive@weezerOSINT
    General

    @artee_49 CVE-2025-48757 it already has a CVE its just not been fixed. Also the report was dismissed, its only right to post!

    Post summary

    The post simply notes that CVE‑2025‑48757 remains unfixed, providing no further technical or exploit details.

    32113166.3K
    10.6K followersView on X
  • Simone Margaritelli@evilsocket
    Disclosure

    A fresh warning from developer Morgan Linton says free Lovable accounts can still read other users' AI chat histories, source code, and database credentials on projects created before November 2025. The pattern is the same one that earned the platform CVE-2025-48757 last year. https://awesomeagents.ai/news/lovable-breach-chat-source-code-credentials/

    Post summary

    The message warns that free Lovable accounts can read other users’ data—similar to the prior CVE-2025-48757—yet it provides no PoC, exploit code, patch, or evidence of active attacks.

    100401.2K
    47.8K followersView on X
  • Somya@somyatwts
    General

    🕛August, 2025 Claude Opus 4.1 launched with 74.5% on SWE-bench Verified - exceptional at understanding complex software architectures and multi-file refactoring. Powered deeper codebase analysis for Claude Code. Lovable raised $200M at $1.8B valuation despite CVE-2025-48757 - a critical security flaw (CVSS 9.3) discovered in May that exposed 170+ apps to unauthenticated database access.

    Post summary

    The post references CVE‑2025‑48757 only in a business context, noting its severity and impact, but does not provide exploit details, patches, or evidence of active attacks.

    1003055
    30 followersView on X
  • Suryansh Tiwari@Suryanshti777
    Disclosure

    Prompt 7: The Exposed Database Check A researcher scanned Lovable apps in 2025 and found 170 with the same hole: Supabase tables shipped with row level security never switched on. It became CVE-2025-48757. AI does what you ask. It never thinks about what you didn't ask. "You are an application security engineer who audits Supabase backed apps before launch and has seen the same handful of failures in every vibe coded project. Audit this project for the security problems AI generated code creates by default. Report everything. Fix nothing until I approve it. Audit: - Every table: is row level security actually enabled, and does each table have a policy that genuinely restricts rows - Any policy with a condition that evaluates true for everyone, which is the same as having no policy - Tables holding emails, names, messages, or anything personal, listed first - What the public key can read, write, and delete right now if someone pulls it out of my frontend bundle, which they can - API keys or secrets sitting in client side code instead of in secrets and edge functions - Data getting dumped to the browser console on page load - Anything the frontend enforces that the database does not, because the frontend is not a security layer - Flag anything that needs a professional security review before this site collects personal data or takes payments Output a findings table ranked critical to low, each with the exact fix, then wait for me to choose which ones to apply."

    Post summary

    The text discloses CVE-2025-48757, identifying that 170 Lovable apps have Supabase tables shipped without row-level security enabled. It provides detailed technical vulnerability information but includes no PoC, exploit, active exploitation evidence, or named patch.

    2001091
    44.5K followersView on X
  • notliaf@notliaf
    Disclosure

    1. No row level security Your Supabase anon key ships in the client bundle. With no RLS behind it, it reads the entire database. 170 of 1,645 Lovable apps had exactly this. CVE-2025-48757, CVSS 9.3. One scan found 172 sites where anyone could delete rows, unauthenticated.

    Post summary

    The entry reveals that many Supabase applications ship an anonymous key with no row‑level security, allowing unauthenticated read and delete of the entire database. This flaw is identified as CVE-2025-48757 with a CVSS score of 9.3.

    20010106
    9 followersView on X
  • Anjula Dwivedi@HeyAnjula
    Disclosure

    Prompt 7: The Exposed Database Check A researcher scanned Lovable apps in 2025 and found 170 with the same hole: Supabase tables shipped with row level security never switched on. It became CVE-2025-48757. AI does what you ask. It never thinks about what you didn't ask. "You are an application security engineer who audits Supabase backed apps before launch and has seen the same handful of failures in every vibe coded project. Audit this project for the security problems AI generated code creates by default. Report everything. Fix nothing until I approve it. Audit: - Every table: is row level security actually enabled, and does each table have a policy that genuinely restricts rows - Any policy with a condition that evaluates true for everyone, which is the same as having no policy - Tables holding emails, names, messages, or anything personal, listed first - What the public key can read, write, and delete right now if someone pulls it out of my frontend bundle, which they can - API keys or secrets sitting in client side code instead of in secrets and edge functions - Data getting dumped to the browser console on page load - Anything the frontend enforces that the database does not, because the frontend is not a security layer - Flag anything that needs a professional security review before this site collects personal data or takes payments Output a findings table ranked critical to low, each with the exact fix, then wait for me to choose which ones to apply."

    Post summary

    The text discloses CVE‑2025‑48757 (Supabase tables without row‑level security) and outlines a detailed audit checklist for AI‑generated projects, covering policy, secret exposure, and client‑side security issues. No PoC, exploit code, patch, or active exploitation information is provided.

    20000101
    2.8K followersView on X
  • Rob Weidner@automatewithrob
    Active Exploitation

    What nobody tells you about leaving the platform: One audit found 170 of 1,645 AI-built production apps had missing or broken row-level security (CVE-2025-48757). Another founder shipped an app with "zero lines of code written." Database exposed in 3 days. 1.5M tokens leaked.

    Post summary

    A CVE-2025-48757 audit revealed 170 AI-built apps with broken row-level security, resulting in a database exposure and 1.5 M leaked tokens within three days, indicating active exploitation.

    2000024
    19 followersView on X
  • Dharmendra Jagodana@Dharmendra_Jago
    Disclosure

    3/8 2. Broken authorization. CVE-2025-48757: Lovable apps wired to Supabase with row-level security off. One app had the access logic inverted — logged-in users blocked, anonymous visitors got everything. 170+ production apps leaked user data the same way.

    Post summary

    A newly disclosed authorization flaw in Supabase‑connected apps led to data leakage across more than 170 production deployments.

    2000067
    192 followersView on X
  • Jace · Rivetz@JaceFromHI
    General

    This isn't hypothetical. CVE-2025-48757 documented this exact pattern affecting Lovable apps built before April 2025. NIST published it May 29, 2026. Lovable's response: "each individual customer accepts responsibility over protecting the data of their application." Which is fair. Most customers just didn't know they had to.

    Post summary

    The post reports that CVE‑2025‑48757, identified by NIST, impacts Lovable apps built before April 2025, while the vendor notes customer responsibility for data protection, implying lack of a vendor‑issued patch.

    20000122
    5 followersView on X
  • Timur Yessenov@Timur_Yessenov
    Disclosure

    @weezerOSINT @Keoikantse @Lovable @discord posted a full writeup, the api endpoint let any authenticated user pull project data by id with zero authorization checks. cve-2025-48757 if you want the formal reference

    Post summary

    A writeup reveals that CVE‑2025‑48757 exposes an API endpoint allowing authenticated users to retrieve any project data by ID due to missing authorization checks, but no exploit code, patch, or active exploitation reports are provided.

    00011504
    633 followersView on X
  • Grok@grok
    General

    Lovable's clarification was to separate their platform from the issue. No hack or leak hit Lovable's core systems, user prompts, or proprietary codebases. The real problem: CVE-2025-48757—missing/misconfigured Row Level Security in Supabase DBs of ~170+ user-built apps. AI-generated apps often skip proper auth by default, exposing data publicly. They stated "no breach" to reassure users their Lovable accounts were safe, while highlighting it's on devs to review/secure what the AI builds. Classic vibe-coding risk, not platform compromise.

    Post summary

    Lovable confirms no platform breach while highlighting that CVE‑2025‑48757 affects many AI‑generated apps due to missing row‑level security in Supabase DBs; no exploitation, PoC, or patch is reported.

    00020290
    8.7M followersView on X
  • Timur Yessenov@Timur_Yessenov
    Disclosure

    @mohbii @weezerOSINT 48 days open is wild but the real number is 170+ databases exposed through missing RLS policies, cve-2025-48757 confirms it was baked into every project by default

    Post summary

    The tweet announces that CVE‑2025‑48757 is a default configuration flaw in many projects, resulting in over 170 databases being exposed due to missing RLS policies.

    10010243
    633 followersView on X
  • Timur Yessenov@Timur_Yessenov
    Disclosure

    @sanjaygpts @weezerOSINT cve-2025-48757 literally shows the platform itself shipped no RLS on generated projects, so yeah the weakest link was never the vibe coder

    Post summary

    The tweet highlights a missing security feature (no RLS on generated projects) in the platform related to CVE-2025-48757, indicating a disclosure of a potential vulnerability.

    00020313
    818 followersView on X
  • Auditd.Pro@auditdpro
    Disclosure

    CVE-2025-48757. Someone scanned 1,645 apps built with Lovable. 170 of them had databases anyone could read. #supabase

    Post summary

    The tweet discloses scan findings related to CVE-2025-48757, noting 170 Lovable apps had publicly readable Supabase databases, without mentioning patches, exploits, or active attacks.

    1000043
    19 followersView on X
  • Auditd.Pro@auditdpro
    Disclosure

    CVE-2025-48757: 170 Lovable apps with no working row level security. 303 endpoints. The platform shipped a scanner that checked whether RLS existed, not whether it worked.

    Post summary

    The text discloses technical details about CVE-2025-48757, describing ineffective row-level security in Lovable apps and endpoints. It does not mention a PoC, exploit tool, active exploitation, or remediation.

    0001045
    16 followersView on X
  • Muvon@muvonteam
    Disclosure

    Same month: CVE-2025-48757 – an inverted access control check in Lovable's Supabase integration – left ~170 production apps open. LLMs generate plausible code. Plausible and correct diverge exactly at the boundaries: auth, permissions, data exposure.

    Post summary

    The post announces CVE‑2025‑48757 as an inverted access control flaw that left roughly 170 production apps vulnerable, but offers no PoC, exploit code, or patch details.

    1000040
    109 followersView on X
  • Vinh Nguyen@nguyenthanhvin9
    Disclosure

    What actually stops it is a policy on the table itself. Nobody asks for one, no test goes red without one, and the app looks identical either way. CVE-2025-48757: 170+ production apps, CVSS 9.3, disputed by the vendor.

    Post summary

    CVE‑2025‑48757 has a CVSS of 9.3 and affects over 170 production apps, yet the vendor disputes it; a policy appears to mitigate exploitation.

    1000013
    52 followersView on X
  • Rishabh@rispectrum
    General

    Checklist plus the version-pinned CVE list: http://github.com/boxed-dev/vibe-coding-security Sources in order: CVE-2025-48757 on http://nvd.nist.gov, the 5,600-app scan is http://Escape.tech's Oct 2025 audit, and Lovable published their own April 2026 incident writeup. I haven't re-scanned those apps myself. Those numbers are theirs, not mine.

    Post summary

    The post lists a CVE and various sources, including an incident writeup, but provides no technical details, PoC, exploit code, or patch information.

    1000072
    56 followersView on X
  • defiankit | LyraFin AI | LyraSec AI@cryptonaire7
    Disclosure

    CVE-2025-48757: apps on Lovable shipped without Supabase row-level security on by default. A scan of 1,645 live apps found 303 exposed endpoints across 170 of them — no login needed to read/write PII and API tokens. (Lovable disputes the classification; the scan itself isn't.)

    Post summary

    The CVE‑2025‑48757 disclosure highlights that Lovable applications were shipped without Supabase row‑level security, leading to 303 exposed endpoints that could read/write PII and API tokens; no exploit proof‑of‑concept or patch is mentioned, and the vendor disputes the classification.

    1000036
    42 followersView on X
  • Mar ia@maricarmen_dev
    Active Exploitation

    4/ La parte que no entra en la valuación: seguridad. CVE-2025-48757: +170 apps de Lovable exponiendo correos, API keys y datos de pago, sin auth. Tres incidentes serios en 13 meses. RLS y manejo de secretos son justo lo que no ves en la demo.

    Post summary

    The post reports CVE-2025-48757 has led to three serious incidents in 13 months, exposing sensitive data due to authentication bypass, highlighting active exploitation in the wild.

    1000037
    144 followersView on X

Explore more