Skip to content
ATAI Today Brief
HomeNewsConceptsGuidesToolbox
AboutSubscribeUA
Subscribe

AI Today Brief

The daily AI-engineering brief. Built in public. EN · UA.

XTelegramLinkedInYouTubeRSS

Follow AI Today Brief on LinkedIn for daily AI-engineering updates and the weekly “5 shifts that changed how developers work” PDF.

Explore

NewsDigestsConceptsGuides

Company

SubscribeAdvertiseAbout

Legal

Editorial policyAI disclosurePrivacyTerms

© 2026 AI Today Brief. All rights reserved.

  1. Home/
  2. News/
  3. Tools & releases/
  4. JFrog Exposes Batch of Fabricated SQLite CVEs Generated by LLMs
Tools & releases

JFrog Exposes Batch of Fabricated SQLite CVEs Generated by LLMs

August 3, 2026· 5 min read
OKCurated by Oleksandr Kuzmenko, AI Product Engineer·Updated August 3, 2026·Sources cited on every story
AI-assisted · editor-reviewed·How we use AI
JFrog Exposes Batch of Fabricated SQLite CVEs Generated by LLMs

Security researchers discovered that over 50 GitHub vulnerability advisories for SQLite were entirely AI-generated LLM slop. The cited code, functions, and patch references did not exist in the target versions.

Impact: Medium

Why it matters

Security teams must independently verify incoming CVE advisories to avoid wasting engineering hours patching non-existent bugs.

TL;DR

  • 01LLMs are being used to generate fake CVE advisories at scale.
  • 02Always cross-reference CVEs with official vendor security pages.
  • 03Automated vulnerability ingestion pipelines are vulnerable to poisoned metadata.

Key facts

Fabricated Advisories
54 out of 55 tested
Initial Score
10.0 Critical (later downgraded to 7.6 High)

The Rise of LLM Slop CVEs

A newly created GitHub repository (programmervuln/cveadvisory-) recently published a batch of SQLite vulnerability advisories alongside 50+ other CVEs. NVD flagged them as critical and CISA ADP agreed. However, JFrog security researchers dug in and found the claims to be completely fabricated LLM slop.

Isolated Testing Workflow

To verify the reports, researchers established an isolated testing workflow:

  • Source Inspection: Cloned official repositories and checked out target tags (version-3.41.0, version-3.51.2, version-3.51.3).
  • Clean Environment Build: Compiled official SQLite releases inside isolated Docker containers.
  • PoC Execution: Fed advisory SQL statements verbatim into compiled binaries under AddressSanitizer (ASan).
  • Metadata Audit: Evaluated CPE patterns across NVD and GHSA feeds.

Specific Findings

  • CVE-2026-51302: Claimed a heap UAF via sqlite3ReleaseTempReg() dereferenced by exprComputeOperands(). Investigation showed exprComputeOperands() did not exist in version 3.41, and register release mechanics do not use heap deallocation.
  • CVE-2026-51303: Claimed ExprListDelete() failed to clear back-references, referencing a nonexistent patch diff between 3.51.2 and 3.51.3.
  • CVE-2026-51296: Reported a UAF in jsonRemoveFunc at lines 3555 and 3575 in json.c, a file that was only 2706 lines long in the target version.

What to do today

  • →Verify all critical CVE alerts against official vendor commit history.
  • →Do not rely solely on automated NVD severity scores for rare dependencies.
#SQLite#NVD#CISA#Docker#AddressSanitizer

Sources

  • SQLite Critical CVEs or LLM Slop?
ShareShare on XShare on LinkedIn
← Previous storyShieldFont Defends Web Content Against AI Scraping via Font Glyph ObfuscationNext story →Copilot Worm Demonstrates Self-Propagating Prompt Injection in Word Documents

Related stories

  • Tools & releasesShieldFont Defends Web Content Against AI Scraping via Font Glyph Obfuscation
  • Tools & releasesAurora: Open-Source Self-Hosted AI Gateway in Go for Multi-Provider Routing
  • Tools & releasesCursor Removes Dollar Cost Data from Usage Dashboard and CSV Exports
  • Tools & releasesDeepSeek V4-Flash API Public Beta Launches with Native Codex Integration

Email digest

Get the morning AI brief

One email a day — the stories that matter for engineers, founders and tech leads. Human-edited, with links to primary sources.

  • ✓120+ sources scanned daily
  • ✓Edited by a human
  • ✓1 email per day
  • ✓EN + UA

By subscribing you agree to the privacy policy.