HoundShieldHoundShield
Products
Products by Industry

One firewall · every compliance framework · one deployment

TechnologySOC 2

Engineers pasting API keys & source into Copilot.

HealthcareHIPAA

Clinicians pasting PHI into AI for documentation.

DefenseCMMC L2

DoD contractors leaking CUI into proposal tools.

Legal & FinancePCI

Privileged client data shared with AI assistants.

Five EyesAUKUS

Allied suppliers navigating DISP & Essential 8.

GovernmentSoon

FedRAMP / FISMA — agency AI governance.

SOC 2 · HIPAA · CMMC L2 · 16 engines · <10msAll capabilities
Features
Core Capabilities

Inside the HoundShield firewall engine

AI Prompt Interception

Every LLM request inspected before it leaves the network.

16 Detection Engines

CUI, PII, IP, PHI, secrets, CAGE codes, clearances.

Immutable Audit Trail

SHA-256 tamper-evident logs. C3PAO-ready.

Live Threat Dashboard

Real-time blocked prompts, risk & posture.

Pricing
Pricing

All frameworks included in every plan

Free

Up to 1,000 prompts/mo

$0/mo
Pro

CMMC suite + AI gateway

$199/mo
Growth

PDF reports + C3PAO coord

$499/mo
Enterprise

On-prem · air-gapped

$999/mo
Compare all plans
Partners
Partner Program

Build & grow with HoundShield

RPO / MSP Referral

Co-brand the $499 report · keep the margin.

MSP / Agency

40% per report · 20% recurring · white-label.

Integrations

Drop-in proxy for ChatGPT, Copilot, Claude.

Docs
Documentation

Live in under 5 minutes · no code changes

Quickstart

One URL change → full compliance.

API Reference

Gateway, classifier & audit endpoints.

FAQ

Searchable answers — pricing, HIPAA, CUI.

16 engines · <10ms scan
Sign inStart free

NIST 800-171 Controls · Access Control

AC.2.004 — Separate Duties to Reduce Risk

CMMC Level 2SPRS if unmet: -1MEDIUM priority~4h to implement

Separate the duties of individuals to reduce the risk of malevolent activity without collusion.

What AC.2.004 means in plain English

Don't let one person have complete control over something important from start to finish. For example, the same person who approves purchase orders shouldn't also be the one who processes payments. Or the person who manages IT accounts shouldn't be the only one who reviews those accounts. Splitting up critical tasks means one bad actor alone can't cause a major problem.

The assessment question

“Are critical functions — such as approving financial transactions, creating user accounts, and reviewing system logs — divided between two or more people so no single individual has end-to-end control without oversight?”

How to implement AC.2.004

  1. Make a short list of your most sensitive functions: who approves contract-related purchases, who manages IT admin accounts, who reviews audit logs. Ensure at least two people are involved in each critical process.
  2. In Microsoft 365, ensure the Global Admin role is held by at least two people (owner and backup) and that the IT manager's account activities are reviewable by the business owner.
  3. Set up a simple two-person rule for financial approvals: the person who requests payment is different from the person who approves it — document this in your purchasing procedure.
  4. Use Microsoft 365 Privileged Identity Management (PIM) if available to require approval before elevated admin roles are activated.
  5. Document your separation-of-duties assignments in a simple one-page matrix: task, primary responsible person, secondary reviewer/approver.

Evidence your assessor will ask for

  • Separation of duties matrix listing critical functions and assigned personnel
  • Screenshot of Microsoft 365 admin roles showing multiple admins with limited roles
  • Written purchasing or financial approval procedure showing two-person requirement
  • Interview notes or attestation from owner confirming SOD is practiced

Does AI prompt monitoring help with this control? Honestly, no.

AC.2.004 is met through the remediation steps above, not through AI traffic controls — an AI prompt firewall neither satisfies nor substitutes for it. We map AI monitoring only to the controls it genuinely evidences (flow control, boundary protection, audit, and incident support); for the full picture of where it does help, see the mapping guide linked below.

Full mapping: which 800-171 controls AI prompt monitoring evidences →

More Access Control controls

AC.1.001

Limit System Access to Authorized Users

AC.1.002

Limit System Access to Authorized Transaction Types

AC.2.003

Control CUI Flow per Authorizations

AC.2.005

Employ Least Privilege

AC.2.006

Use Non-Privileged Accounts for Non-Security Functions

AC.2.007

Prevent Non-Privileged Users from Executing Privileged Functions

Score yourself against all 110 controls

The free ShieldReady assessment walks every NIST 800-171 requirement, computes your SPRS score, and shows exactly which gaps cost the most points.

Start the free assessment $499 AI risk assessment
HoundShieldHoundShield

Local-only AI compliance firewall for CMMC Level 2, HIPAA & SOC 2. Prompt content never leaves your network.

CMMC LVL 2HIPAASOC 2NIST 800-171DFARS 7012
Product
FeaturesHow it worksPricingCompareDashboardChangelog
Compliance
CMMC Level 2HIPAASOC 2NIST 800-171DFARS 7012
Company
PartnersDocumentationFAQContact salesAboutSecurity
© 2026 HoundShield. All rights reserved. · Privacy · Termshoundshield.com · local-only · zero data exfiltration