Updated November 2026: This article was rewritten and refreshed for accuracy and relevance.

Table of Contents

A 60-person defense machining shop wins a subcontract carrying DFARS 252.204-7012. Engineering starts receiving drawings marked as Controlled Technical Information. Over the next eight months, three things happen that nobody records. A project manager opens a free file-transfer account to move a 400 MB drawing package that a prime contractor's portal rejected. The IT provider turns on a new endpoint detection tool that ships configuration and alert data to its own tenant. A machinist saves a marked drawing to a personal cloud sync folder so he can check a dimension from home.

None of that is a platform failure. All three services are well engineered. What went wrong is that nobody had answered a prior question. Who decides whether a cloud service is allowed to touch CUI, and on what evidence? That is what cloud governance answers, and for a defense contractor it is not an internal best practice. It is scoping evidence an assessor will read.

Why Governance Looks Different When the Data Is CUI

Generic cloud governance optimizes for cost, uptime, and tidy permissions. Those still matter. But once Controlled Unclassified Information enters the picture, governance takes on a second job. It has to produce a defensible account of your assessment boundary.

The practical consequence is that tool adoption stops being reversible. Spin up a storage bucket in a commercial tenant and you can delete it next week with no lasting consequence. Let CUI land in that bucket and you have created an in-scope asset, a documentation obligation, and in many cases a finding. Deleting the bucket undoes none of it. Governance is the control that prevents the decision, not the cleanup that follows it.

This is also where the July 2026 suspension of CMMC Phase 2 misleads people. The assessment timeline moved. The scoping text did not. The provider rules below sit in codified regulation at 32 CFR 170.19, and the obligations under DFARS 252.204-7012 have been contractual since 2017.

How 32 CFR 170.19 Classifies Your Providers

CMMC defines an External Service Provider as external people, technology, or facilities that an organization uses for the provision and management of IT or cybersecurity services on its behalf. The definition is deliberately broad. It captures your managed IT provider, your SOC, your backup vendor, your SaaS platforms, and a centralized internal team sitting in a different business unit.

For each of them, the rule requires two determinations. Is the provider a Cloud Service Provider? And does it process, store, or transmit CUI, Security Protection Data, or neither? Those two answers place a provider in one of four positions, and the obligations differ sharply between them.

A cloud service provider that handles CUI

This is the strictest path. The cloud service offering has to be FedRAMP Authorized at the Moderate baseline (now Class C) or higher according to the FedRAMP Marketplace, or it has to meet DoD FedRAMP Moderate equivalency through a validated body of evidence. The requirement attaches to the specific service offering, not to the vendor. A provider with one certified product does not thereby cover everything else in its catalogue.

Encryption is not a substitute. The authorization status of the platform is what an assessor evaluates. Commercial Microsoft 365 holds no FedRAMP certification at any level, and that is why so many contractors end up evaluating a GCC High migration.

A non-cloud provider that handles CUI

Your managed IT provider, if it is not delivering a cloud service, does not need its own CMMC certification. Instead, the services it provides fall inside your assessment scope and are assessed as part of your assessment. The distinction matters commercially. Contractors routinely disqualify capable providers on the mistaken belief that a certificate is a prerequisite, and they also sign with providers who have never accepted in writing that their work will be assessed inside somebody else's audit.

Any provider that handles only Security Protection Data

This is the category most contractors miss. A tool that never sees a drawing but collects your logs, configuration, alerts, or vulnerability data is handling Security Protection Data. Those services are in scope and get assessed as Security Protection Assets, whether or not the provider is a cloud provider. FedRAMP does not apply here. CMMC requirements do. The endpoint detection tool from the opening lands exactly here.

A provider that handles neither

If neither CUI nor Security Protection Data ever reaches the provider's systems, it does not meet the CMMC definition of an External Service Provider. Your payroll platform and your plotter paper supplier sit outside the rule. The governance task is not to scope them in. It is to be able to show why they are out.

Documenting the Provider Relationship

Placing a provider correctly is only half the work. The rule requires the provider relationship and the services it delivers to be documented in your System Security Plan, and described in the provider's own service description and Customer Responsibility Matrix.

That Customer Responsibility Matrix is the document contractors most often do not have. Without one there is no written statement of which of the 110 requirements the provider implements, which you implement, and which are shared. And without that statement you cannot claim a single inherited control, so work your provider is genuinely performing counts for nothing in your assessment. Ask for the matrix before signing, not during remediation.

The Five Asset Categories

Governance decisions have to resolve into an asset inventory an assessor can follow. For a Level 2 assessment, 32 CFR 170.19 maps every asset into one of five categories.

  • CUI Assets process, store, or transmit CUI, and are assessed against all Level 2 requirements.
  • Security Protection Assets provide security functions to your assessment scope, and are assessed against the requirements relevant to what they provide.
  • Contractor Risk Managed Assets are capable of handling CUI but are not intended to, because your policies prevent it.
  • Specialized Assets cover operational technology, government furnished equipment, restricted information systems, and test equipment.
  • Out-of-Scope Assets are everything else, and they require demonstrable separation from CUI.

The Contractor Risk Managed category deserves particular attention. These assets are not required to be physically or logically separated from CUI assets. That is real flexibility. In exchange, you have to show the assets are managed under your own risk-based security policies, procedures, and practices, and document their treatment in the System Security Plan. Claim the category without that evidence behind it and an assessor will treat the asset as a CUI Asset, with the full requirement set attached.

The Controls That Matter for CUI

Identity and access, scoped to the boundary

Least privilege is the generic advice. The CUI-specific version is that access to in-scope systems is a separate, smaller, individually justified list, with multifactor authentication enforced, service account credentials rotated on a defined schedule, and privilege escalation paths reviewed on a set cadence. One over-permissioned service account inside the boundary turns a single phishing email into an environment compromise and a reportable incident.

Data classification and where the regulated tier may live

Tiering data as public, internal, confidential, and regulated is useful but not sufficient. A CUI program also needs to know which systems are permitted to hold the regulated tier, enforced technically rather than by policy statement, with encryption at rest, FIPS-validated cryptography where the requirements call for it, and data residency constraints applied to the in-scope systems specifically.

Change control and the audit trail

Infrastructure as code and peer-reviewed configuration changes are good engineering. For CMMC they also produce the record that answers what changed, when, and who approved it, for any system inside the boundary. Prohibiting manual console changes in the in-scope environment is the specific policy that makes that record trustworthy.

Incident response and the 72-hour reporting requirement

DFARS 252.204-7012 requires rapid reporting of cyber incidents, and your cloud provider has to support media preservation and forensic access. Governance here means knowing in advance which provider holds which log, how long it is retained, who at that provider you contact, and whether your contract actually obliges them to help. Working that out during an incident is how the reporting window gets missed.

Cost control

Tagging, budget alerts, and idle resource cleanup are worth doing, and they are not compliance controls. They belong in your governance framework. They do not belong at the top of it when CUI is in play.

Building the Approval Gate

The mechanism that prevents the opening scenario is a gate between wanting a tool and using it for program work. Four steps make it real.

Step 1. Inventory what is already running

Before writing policy, find out what exists, including accounts nobody registered centrally. Contractors regularly discover at this stage that CUI has been moving through commercial tools for months. The discovery is unwelcome and it is far cheaper now than during a pre-assessment.

Step 2. Classify every provider against the four positions

Work through the list from the section above. For each provider, record whether it is a cloud service, whether it touches CUI, whether it touches Security Protection Data, and the resulting scope treatment. This document is scoping evidence, so write it to be read by someone else.

Step 3. Name a single approver

One accountable person, not a committee, decides whether a proposed service may touch CUI, and for cloud services that decision requires the FedRAMP evidence or the equivalency body of evidence in hand. Collective accountability is how the file-transfer account in the opening scenario gets created.

Step 4. Review on a cadence and record open gaps

Provider relationships change, products get retired, and certifications lapse. Quarterly review of the provider inventory, access lists, and open findings keeps the System Security Plan matched to the environment. Gaps you cannot close immediately belong in the POA&M with an owner and a date. That is how a known gap is carried defensibly rather than silently.

Where Stratify IT Helps

We work with Defense Industrial Base contractors on provider classification, scope documentation, Customer Responsibility Matrix review, and the System Security Plan narrative that has to hold up when an assessor reads it. That work sits alongside CMMC certification services and the control implementation that follows it.

If you are earlier in the process, our CMMC guide for small defense contractors covers the broader program, and the case for treating cloud choices as a leadership decision rather than an IT one is made in our piece on why an ill-defined cloud decision becomes a boardroom concern.

Tell us which cloud services currently touch your program data and we will tell you where each one lands under 32 CFR 170.19. Contact Stratify IT to start with your provider inventory.

Frequently Asked Questions

The usual mistake is treating governance as a gate developers have to pass through. A better approach is baking it into their toolchain, pre-approved Terraform modules, policy-as-code checks in the CI/CD pipeline, and guardrails that auto-remediate minor issues rather than block the build. When developers can deploy faster inside the guardrails than outside them, compliance stops feeling like friction. Tools like Checkov or OPA (Open Policy Agent) can enforce rules at the code level before anything reaches production.

A basic foundation, IAM policies, tagging standards, and a cost monitoring dashboard, can be functional in four to six weeks for a team that's focused. A mature framework with automated policy enforcement, incident runbooks, and documented approval workflows is closer to six months. Most organizations underestimate the policy documentation work and overestimate how long the technical configuration takes. The actual bottleneck is usually getting stakeholder agreement on who owns what, not configuring the tools.

It depends on your organization's size, but shared ownership tends to fail. Someone needs final authority over policy decisions. Security should own the risk and compliance standards, IT operations should own the enforcement mechanisms, and finance should own the cost policies, but a cloud center of excellence (or a designated cloud architect) needs to coordinate across all three. Without a single accountable owner, governance gaps accumulate exactly where team boundaries are.

The native tools vary more than people expect. AWS leans on Service Control Policies and Config Rules; Azure uses Policy and Blueprints (now Defender for Cloud integrates much of this); GCP uses Organization Policies and Security Command Center. Each has its own model for hierarchy, AWS Organizations, Azure Management Groups, GCP Resource Hierarchy. If you're multi-cloud, a third-party tool like Wiz, Prisma Cloud, or Turbot gives you a single governance plane instead of managing three separate policy frameworks simultaneously.

Start with four things: enforce MFA on every account, enable CloudTrail or equivalent audit logging so you have a record of who changed what, set up billing alerts at 80% and 100% of your monthly budget, and apply resource tags that identify the owner and environment for every resource. This won't cover everything, but it closes the most common failure modes, unauthorized access, invisible changes, surprise bills, and orphaned resources, without requiring a dedicated governance team.

At minimum, annually, but that's really just a backstop. Practically, your framework needs a review trigger tied to meaningful changes: a new cloud region, a new compliance requirement (SOC 2, HIPAA), a major architecture shift, or any significant security incident. Quarterly reviews of IAM permissions and cost policies are reasonable for most organizations. Governance that only gets reviewed once a year tends to drift out of sync with how the environment is actually being used within about eighteen months.

For most mid-sized businesses, the practical stack is: NIST Cybersecurity Framework 2.0 (released February 2024) as the overarching risk management framework, its six functions (Govern, Identify, Protect, Detect, Respond, Recover) map cleanly to board-level reporting; CIS Benchmarks for specific platform hardening (CIS publishes benchmarks for AWS, Azure, GCP, Microsoft 365, and Kubernetes); and CSA STAR (the Cloud Security Alliance Security, Trust, Assurance, and Risk registry) for evaluating SaaS vendor security postures during procurement. Regulated industries layer on additional frameworks, FedRAMP for federal contractors, HIPAA Security Rule for healthcare, PCI DSS for payment processors, but NIST CSF 2.0 plus CIS Benchmarks plus CSA STAR is the foundation that addresses 80 percent of governance scope without overengineering.

FedRAMP (Federal Risk and Authorization Management Program) certifies cloud service offerings in one of four classes, A through D, based on the sensitivity of federal data they handle. Classes C and D correspond to what were previously the Moderate and High impact levels. If your organization processes, stores, or transmits federal data (including FCI or CUI), you can only use cloud platforms holding the appropriate FedRAMP authorization. Microsoft 365 GCC High and Azure Government are FedRAMP Certified at Class D; standard commercial Microsoft 365 does not meet the threshold for CUI. The practical governance implication is significant: any vendor selection touching federal workloads needs FedRAMP status verified before contracts are signed, and encrypted federal data in a non-FedRAMP environment does not satisfy the requirement, authorization of the platform itself is what assessors evaluate, not the encryption on top of it.

Sharad Suthar

Sharad has a proven track record of delivering successful IT projects underpinned by creative problem-solving and strategic thinking. He brings an extraordinary combination of in-depth technical knowledge, problem-solving skills, and dedication to client satisfaction that enables him and his team at Stratify IT to deliver optimal IT solutions tailored to the specific needs of each organization, from large corporates to small businesses. His impeccable attention to detail and accuracy ensure that his clients get the best possible results.