Test Your AI Security Right Now

Paste a prompt your team sends to ChatGPT, Claude or Copilot. The same 54 detection patterns that ship in the product scan it here — in your browser, on this device — and map every finding to a NIST 800-171 control.

No signup. No email. Nothing uploaded.

Generate your CMMC AI risk snapshot

Paste a real prompt your team sends to ChatGPT, Claude or Copilot. HoundShield's detection engines scan it locally, in your browser, map every finding to a NIST 800-171 control, and preview exactly what the full $499 report contains.

Your text is never sent anywhere — the scan runs entirely on this device.
Try a real scenario:
0 / 100,000 characters

What the engine scans for

16 detection engines across 54 patterns — CUI markings, CAGE codes, ITAR terms, PHI and more. The scan above runs this same set locally, so no prompt text ever leaves your network.

CUI markings
CAGE codes
Contract / DoDAAC #
Clearance levels
ITAR / EAR terms
Export-control
SSN / PII
PHI · MRN
ICD / diagnosis
API keys / secrets
AWS / cloud keys
JWT / tokens
PCI / card data
Source code
Trade-secret IP
IP / network data

How to take the test — step by step

1. Paste a real prompt

Use text your team actually sends to an AI tool — a code snippet, a support ticket, a config file, a draft email. The more realistic it is, the more honest the result. It stays on your device either way.

2. Or load a scenario

Four one-click scenarios cover defense (CUI, CAGE codes, ITAR), healthcare (a patient record), legal (an M&A memo) and DevOps (a config paste with live credentials). Each one selects its industry automatically.

3. Read the findings

Every finding names the pattern, the data category, the NIST 800-171 control it implicates, and whether the proxy would block or flag it. Expand any finding for why it matters, a quick fix and a permanent one.

4. Check what is NOT shown

Findings deliberately show the pattern name, never the matched text. Your SSN, key or CAGE code is detected and never echoed back — on screen or in the PDF. That constraint is the product, demonstrated rather than asserted.

5. Generate the PDF

A branded gap-report preview, built by your browser and never uploaded. It is marked as a preview, because a one-paste snapshot is not the tamper-evident artifact an assessor accepts.

6. Get the real thing

The $499 CMMC AI Risk Assessment Report runs the proxy in your own environment for 14 days and returns a signed PDF mapped to NIST 800-171, backed by an immutable audit trail.

This free preview

  • All 54 patterns, on one prompt you paste
  • Runs in your browser — no server, no upload
  • NIST 800-171 control per finding, plus remediation
  • Preview PDF, generated on your device
  • No real-time interception
  • No audit log — nothing an assessor can verify

The $499 report

  • The same 16 engines and 54 patterns, on EVERY prompt for 14 days
  • Inline gateway — blocks before the model sees the data
  • Every finding mapped to a NIST 800-171 control
  • Immutable SHA-256 hash-chained audit trail
  • Signed PDF you can hand to a C3PAO assessor
  • Runs in your environment (Docker) — nothing leaves it
Get the $499 report See full pricing

This demo runs 100% in your browser. No data is sent to any server.

54
Patterns checked
16
Detection engines
Nowhere
Your data sent
$0
Cost