Vault xCircle

Vault xCircle

Researching how files hijack AI
#92588090 followers0 following

About

6 months ago I asked an AI a simple question: "If someone hid instructions in a file I uploaded, would you follow them?" It said yes.... That sent me down a rabbit hole. I tested major AI tools with hidden prompts in PDFs, images, and docs. Most had no defense. Now I'm building the tools to fix that. VaultScan is the first product from VaultXCircle Labs. More coming.

Links

VaultScan
x

Badges

Tastemaker
Tastemaker
Gone streaking
Gone streaking

Maker History

  • VaultScan
    VaultScanScan files for hidden AI attacks before uploading
    Dec 2025
    • 🎉
    Joined Product HuntDecember 9th, 2025

Forums

Vault xCircle•

3mo ago

VaultScan - Scan files for hidden AI attacks before uploading

Most people have no idea their files can hijack AI. Hidden instructions in PDFs. Invisible text in images. Metadata payloads in docs. All designed to manipulate AI systems into doing things they shouldn't. VaultScan scans your files before you upload to ChatGPT, Claude, Gemini, or any AI assistant. Upload a file. See what's hidden. Decide if it's safe. Privacy first: Files processed in memory, deleted immediately. Never stored. Free: 5 scans/month Pro: Unlimited
View more