SAT-CHAIN
Inference Shield

Your public text may be telling competitors how you built it.

Marketing pages, whitepapers, docs and pitch decks often name the exact tools and methods behind a product. Inference Shield finds the phrases that give your technology away, and rewrites them so your text still sells without explaining how it works.

Check a document Get 50 free credits

No subscription needed. Paste text or drop a PDF, Word, Markdown or text file.

What it catches

An example. An analyst AI reads your text the way a competitor would, and tries to reconstruct your stack from it.

Before
Our assistant remembers every conversation. PostgreSQL with pgvector and a fine-tuned cross-encoder keep answers under 200 ms, and 3 enterprise customers already rely on it daily.
81leak score out of 100
PostgreSQLpgvectorDense retrievalCross-encoder re-ranking
After Protect
Our assistant remembers every conversation. A high-dimensional memory layer and a custom relevance engine keep answers under 200 ms, and 3 enterprise customers already rely on it daily.
11re-checked by the same analyst

Same message, same claims. Only the phrases that named the method changed.

How it works

Two AIs from different companies check each other's work.

01 · Leak Score

Find what leaks

An analyst AI tries to work out your stack and methods from the text. You get a score from 0 to 100, what a competitor could infer, and the exact phrases that give it away.

02 · Protect

Rewrite only those phrases

A second AI from another vendor rewrites just the flagged phrases, replacing mechanisms with outcomes. The rest of your text stays as you wrote it.

03 · Verify

Check the result

The same analyst scans the rewrite again. If it can still infer too much, Shield tells you, instead of calling it safe.

Your text stays yours

  • Your text is deleted after the analysis, and results after 24 hours.
  • The AI providers are used on paid API terms that don't train on your data.
  • Nothing runs until you press Run.

SAT-CHAIN Verified receipts

  • Every run comes with a signed receipt: which checks ran and what they found.
  • Receipts hold fingerprints (hashes), never your text.
  • Anyone can check a receipt is genuine at shield.sat-chain.com/verify.

Pricing

Pay with LISA credits: 1 credit = $0.01, bought from $1. Prices are per 2,500 words. Failed runs are refunded automatically.

Quick Leak Score

10 credits

A fast first check with a lighter analyst.

Deep Leak Score

25 credits

Our strongest analyst: the depth to use before you publish.

Protect

150 credits

Score, rewrite and re-check. +100 credits per extra 2,500 words.

New keys come with 50 free credits. LISA Pro, Team and Enterprise plans include 300, 1,000 and 3,000 credits a month. All plans and credit bundles

Part of the LISA Suite

One license key and one credit balance across three tools.

LISA extension

Saves, compresses and hands off your AI conversations on 12 platforms.

Learn more · Add to Chrome

LISA Nexus

Turns long chats and documents into compact AI memory, and keeps your Library online.

Learn more · Open

Inference Shield

Keeps what you publish from revealing how your technology works.

Learn more · Open

Check a document

Inference Shield is proprietary technology of SAT-CHAIN LLC. A lower leak score reduces risk; no tool can guarantee what a determined reader will infer.