The level of system and network required for Controlled Unclassified Information (CUI) is the one that can satisfy all 110 requirements of NIST SP 800-171 Revision 3 while keeping every CUI component inside a defined boundary. That standard applies to nonfederal systems that process, store, or transmit CUI under a federal contract, and the practical answer for most organizations is a dedicated enclave rather than the entire enterprise network. The boundary is the decision. Scope too wide and you pay for controls on machines that never touch CUI; scope too narrow and a single shared service becomes an uncontrolled spill path.
This article walks through how that boundary gets drawn, which systems sit inside it, what happens to the network path between them, and where contractors most often get the scoping wrong. The reference point throughout is the Department of Defense's phased rollout of Cybersecurity Maturity Model Certification, because DoD contracts are the single largest driver of CUI scoping decisions in the defense industrial base.
Why the CUI Boundary Decides Your System Level Before Any Control Does
Start with the definition rather than the technology. NIST SP 800-171 Rev. 3, published in May 2024, defines a "security requirement" as a requirement derived from the 110 controls in NIST SP 800-171, and it defines the "CUI Assets" category as systems and components that process, store, or transmit CUI. The standard's scoping guidance in section 1.1 says the requirements apply to every component that processes, stores, or transmits CUI, and to components that protect the CUI boundary itself.
What that language means in practice is that your system level is determined by the union of two sets: assets touching CUI directly, and assets whose compromise would let an attacker reach those assets. A domain controller that authenticates a CUI workstation falls in the second set. A marketing laptop that never authenticates against the same identity store may not.
The Department of Defense clarified this further with its "Clarification on NIST SP 800-171" memo, which introduced the concept of a contractor-defined "security protection asset." Those are components that don't touch CUI but are essential to protecting it. The memo's key concession is that a contractor may define such assets narrowly, which is what makes a small, controlled enclave legally defensible rather than a scoping shortcut.
The common failure runs the other way. Contractors assume that the network they already have is the system, and then try to fit 110 controls onto a flat network with personal devices, guest Wi-Fi, and a shared file server. That's not a scoping decision, it's a budget crisis waiting to happen. Draw the boundary first; the control count follows from it.
What a System Boundary Actually Contains Under NIST SP 800-171 Rev. 3
A system boundary under 800-171 Rev. 3 is a documented set of components, plus the connections between them, that together process, store, or transmit CUI or protect the components that do. The documentation matters as much as the components: an assessor will ask for a system security plan that shows the boundary, and a boundary that exists only in the administrator's head will not survive that conversation.
Practically, the boundary includes workstations and servers that handle CUI, the identity infrastructure that authenticates users into those systems, the backup and recovery systems holding CUI copies, the logging and monitoring stack that watches the CUI environment, and the network devices that carry CUI traffic. That last item catches people. A switch that carries CUI traffic is inside the boundary even if it holds no data at rest, because firmware compromise on that switch gives an attacker a vantage point on the CUI path.
Boundary protection also cuts the other way. If a shared print server receives print jobs from both CUI and non-CUI workstations, the print server and the network path to it are inside scope until you segment it or replace it. The same logic applies to a help-desk tool that gives technicians remote control of CUI endpoints, and to a vulnerability scanner credentialed into the CUI subnet. These are the "small" services that quietly double the size of an enclave.
NIST's own guidance on scoping makes an allowance worth knowing: components that merely route CUI traffic can sometimes be treated as boundary-protection assets with compensating controls rather than full-scope assets, but that argument requires documentation, and it requires that the routing device not also store CUI. A core switch doing inter-VLAN routing is a different case from a firewall performing inspection; the firewall's logs now contain CUI-derived metadata.
Write the boundary down as a table before you buy anything. Columns for asset name, function, CUI contact (direct, protective, none), and the justification sentence. That table becomes the spine of your system security plan and the answer you hand an assessor when they ask why a given server is or is not in scope.
The Enclave Model: The System Level Most Contractors Should Target
For a small or midsize contractor, the fastest defensible architecture is a dedicated CUI enclave: a segmented set of virtual machines or dedicated hardware, its own identity store, its own logging, and a single controlled entry point. The rest of the corporate network stays out of scope because it never touches CUI and never authenticates into the enclave.
The enclave model's economics are the reason it wins. NIST SP 800-171A, the companion assessment procedures, walks an assessor through roughly 320 individual assessment objectives across the 110 requirements. Every additional in-scope endpoint carries that assessment weight. An enclave of 25 virtual desktops assessed once is a very different project from 400 corporate laptops spread across three offices.
The correctness of the enclave depends on one rule: no CUI leaves the enclave. That means no CUI on email, no CUI pasted into a corporate chat tool, no CUI dragged onto a shared drive, and no CUI printed to a shared printer. The Department of Defense's CMMC program makes this concrete β CMMC Level 2 assessment requires the contractor to demonstrate that CUI is handled only within assessed scope, and an unassessed system that receives CUI is a finding regardless of how secure the enclave itself is.
Contractors sometimes try to have it both ways by declaring a boundary around everything and then claiming only a subset of controls. That is not how the standard works. The boundary and the control set are one package. Choose the package that matches the work you actually do.
Network Segmentation Choices Compared for CUI Environments
Three segmentation patterns dominate real CUI programs, and the right pick depends on headcount, whether you share infrastructure with non-CUI work, and how much you can afford to duplicate identity and logging. The comparison below uses the same dimensions across all three so the trade-off is visible.
| Dimension | Physical air gap | Virtual enclave (hypervisor segmentation) | Flat network with VLANs | |
|---|---|---|---|---|
| Boundary clarity for the assessor | Highest β separate hardware, separate rack | High β separate VMs and virtual network, documented | Low β shared switching fabric and shared services | |
| Duplicate identity store required | Yes | Yes, unless directory trust is tightly scoped | No, but shared directory puts every domain member in scope | |
| Typical CUI endpoints supported | Up to about 10 without heavy cost | 10 to 200 comfortably | Any number, but scope grows with each one | |
| Shared file and print services | None β separate | Isolated virtual file server | Shared, which pulls those servers into scope | |
| Logging and monitoring | Separate stack | Separate log pipeline, often virtual | Shared SIEM, but the SIEM becomes in-scope | |
| Failure mode | Users copy files out because the gap is inconvenient | Misconfigured virtual switch bridges enclave to corporate | One flat VLAN mistake exposes all CUI at once | |
| Incremental admin burden | Highest hardware and patching cost | Moderate β hypervisor patching, duplicate directory | Lowest admin burden, highest assessment burden |
The virtual enclave wins for most organizations between roughly 10 and 200 CUI users. Its balance is real: you duplicate the directory, the logging, and the file service, but you do it as virtual machines, and you keep the corporate network clean. Above 200 users the duplication cost starts to argue for a purpose-built environment, and below 10 users a small physical enclave is easier to defend because there is less shared infrastructure to explain.
Skip the VLAN-only option if you have any shared services at all. A flat network with VLANs works only when the CUI VLAN is genuinely isolated β its own domain controller, its own file server, its own printers. The moment those services are shared, the VLAN stops being a boundary and becomes a diagram.
Identity and Endpoint Requirements Inside the Enclave
The identity store is where many enclave designs fail assessment. NIST SP 800-171 Rev. 3 requirement 03.01.01 calls for limiting system access to authorized users and processes, and the assessment procedures in 800-171A look for account management evidence: who approved the account, when it was last reviewed, and whether privileges match the role. A CUI enclave needs its own accounts, its own approval records, and its own periodic review.
Endpoint configuration follows the same logic. Requirement 03.04.01 covers baseline configurations, and the assessment procedure asks for documented configuration settings applied to in-scope components. If your enclave desktops inherit a corporate image with six months of drift, you'll be documenting exceptions instead of baselines. Build a separate gold image for enclave endpoints and keep it under configuration control.
Multifactor authentication is the requirement contractors ask about most. Under 800-171 Rev. 3, requirement 03.01.03 requires MFA for local and network access to privileged accounts and for network access to non-privileged accounts. In the enclave model that means every enclave login, plus the administrative path into the hypervisor and the enclave's own management interfaces.
The failure case worth planning for: an enclave endpoint that phones home to a corporate management tool β patch management, endpoint detection, asset inventory β creates a network path from the corporate network into scope. Either deploy enclave-side instances of those tools or document the connection as a boundary-protection asset with its own controls. Both work; the undocumented middle does not.
Where Network Paths Get Pulled Into CUI Scope
Traffic inspection is the sleeper issue. A next-generation firewall that performs deep packet inspection on CUI traffic logs URLs, file hashes, and sometimes payload fragments. Those logs are CUI-derived, which puts the firewall's log store and the syslog path to the SIEM inside the boundary. That's not a hypothetical: it's the single finding that most often forces a scope expansion late in an assessment.
The fix is architectural rather than procedural. If the firewall must inspect CUI traffic, treat its logs as CUI and store them in the enclave's logging pipeline. If the logs don't need to be that detailed, reduce the inspection on the CUI VLAN to flow-level metadata rather than full payload. Decide which path you want before the assessor asks.
Remote access is the second path. A VPN concentrator that terminates sessions from both corporate and enclave users is shared infrastructure, and the enclave session traverses it. Either give the enclave its own VPN termination point or accept that the concentrator is in scope. The cost difference between the two is real but smaller than the assessment cost of a late scope expansion.
Cloud services follow the same rule. A CUI workload in a cloud tenant needs its own account boundary, its own logging destination, and documented handling that matches the standard. The Federal Risk and Authorization Management Program, FedRAMP, authorizes cloud service offerings for federal use. If your CUI lives in a cloud service, the standard's requirement 03.17.01 governs the agreement, and the service's FedRAMP status is the evidence an assessor will look for.
CMMC Levels and What They Mean for the System You Build
CMMC Level 2 aligns with NIST SP 800-171 Rev. 2 and requires a third-party assessment for most contractors handling CUI, conducted by a Certified Third-Party Assessment Organization, known as a C3PAO. That alignment matters because Rev. 3 and Rev. 2 aren't interchangeable in the near term: the CMMC program's assessment criteria were written against Rev. 2, and Rev. 3 adds requirements such as organization-defined parameters that contractors need but which sit outside the current CMMC scope.
The DoD's phased implementation means the requirement appears in contracts over time rather than on a single date. Contractors see it as a contract clause, and the clause specifies the level. If your contract requires CMMC Level 2, your enclave must be assessable against the Level 2 criteria, which is a narrower and more prescriptive test than a plain 800-171 self-assessment.
CMMC Level 3, for the highest-priority programs, adds a subset of NIST SP 800-172 requirements on top of Level 2 and requires government-led assessment. Fewer contractors will need it, but the ones who do need to design for it up front: the 800-172 requirements include advanced persistent threat protections that are hard to retrofit onto an enclave designed only for Level 2.
The practical consequence: build your enclave to the standard first, then map to the level your contracts specify. An enclave designed against 800-171 gives you most of the Level 2 evidence and a defensible upgrade path. An enclave designed to a checklist gives you neither.
Documenting Scope: The System Security Plan That Survives an Assessment
The system security plan is the artifact that proves your scoping decision. It names the boundary, lists in-scope components, describes the connections between them, and documents how each of the 110 requirements is met β or why it is not applicable, with justification. NIST publishes a template for the plan, and assessors generally expect its structure.
Two sections of the plan do most of the scoping work. The system environment section describes the enclave, its users, its data flows, and its connections to external systems. The requirements section then walks each of the 110 requirements and names the implementation and the responsible party. If a requirement is marked "not applicable," the plan has to say why β usually because the capability isn't present in the enclave, not because you didn't get around to it.
Keep the boundary diagram in the plan current. Assessors will compare the diagram against what they see in the environment, and a diagram that's six months out of date is worse than no diagram because it suggests the boundary isn't actively managed. Update the diagram when assets move, when new services enter the enclave, and when connections change.
One documentation habit pays off repeatedly: maintain a scope decisions log. Every time you decide an asset is out of scope, write the date, the asset, the reason, and who approved it. That log turns a contentious assessment conversation into a reference lookup.
Determining the Required System Level: A Step-by-Step Scoping Sequence
The sequence below produces a documented boundary and a level decision you can defend. It assumes the reader has contract language specifying CUI handling and needs to determine what to build.
1: Pull the contract clause. Identify the CMMC level, the NIST SP 800-171 revision referenced, and the CUI categories in the statement of work. This is the requirement you are scoping against. 2: List every business process that touches CUI. Interview the program managers, not just IT. CUI arrives by email, by portal download, by customer handoff, and by subcontractor transmission. 3: Trace each process to its assets. Workstation, application, file store, database, print path, and any external service in the flow. Record the CUI contact type for each asset: direct, protective, or none. 4: Mark the protective ring. Identity, logging, backup, vulnerability scanning, and the network devices carrying CUI traffic. These enter scope even though they hold no CUI at rest. 5: Test each candidate for shared use. An asset used for both CUI and non-CUI work is in scope until it's separated. Record the separation plan or accept the scope. 6: Choose the segmentation pattern using the comparison table above, sized to your endpoint count. 7: Draft the boundary diagram and the scope decisions log. Circulate both to the program and security leads for sign-off. 8: Map the boundary to the assessment level from step 1, then build the system security plan against the requirements in that standard. 9: Re-test the boundary at each contract renewal and whenever a shared service is introduced. New tools are the most common source of scope creep.
If a shared service turns out to be unavoidable β a corporate patch manager that must reach enclave endpoints, for example β document it as a boundary-protection asset and give it the controls the standard requires for that role. The failure handling is documentation plus compensating control, never silence. When an assessor finds an undocumented connection, the finding is about the documentation as much as the connection.
Frequently Asked Questions About CUI System and Network Levels
Does a single CUI file on a shared drive make the whole corporate network in scope?
No, but it makes the shared drive and the systems that reach it in scope until the file is removed and the exposure is documented. The practical response is to quarantine the drive, delete the file per your CUI handling procedure, and then assess whether the path that put it there is still open. A one-time exposure is a finding you remediate; a recurring one means your boundary is wrong.
Is CMMC Level 2 the same as NIST SP 800-171 Rev. 3?
Not exactly. CMMC Level 2 assessment criteria were written against NIST SP 800-171 Rev. 2, while Rev. 3 was published in May 2024 and adds organization-defined parameters and reorganized requirements. A contractor assessed at Level 2 meets the Rev. 2 control set, and the Rev. 3 gap needs separate treatment in the system security plan. Confirm the revision your specific contract clause cites.
Do cloud services hosting CUI need their own FedRAMP authorization?
The standard requires that the service provider's CUI handling be governed by an agreement and documented, and assessors look for evidence of the provider's authorization status. FedRAMP authorization is the recognized evidence for federal-facing cloud services. If a provider isn't authorized, the assessment burden shifts to your documentation of how the service meets the requirements, which is a much harder case to make.
Can a small business meet CUI requirements without a dedicated enclave?
Yes, if the business handles CUI on very few endpoints and can keep the corporate network genuinely separate. Below roughly 10 CUI endpoints, a small physical enclave is often cheaper than duplicating identity and logging across a virtual environment. The constraint is shared services: the moment a CUI endpoint uses a corporate file server or printer, the enclave argument weakens.
How many assessment objectives does an assessor actually test?
NIST SP 800-171A walks the 110 requirements as roughly 320 individual assessment objectives, and each one needs evidence. That count is why scope discipline matters so much: each additional in-scope component multiplies the evidence you must produce across that objective set, and a 400-endpoint scope is a fundamentally different project from a 25-endpoint one.
What happens if CUI is stored on an out-of-scope system during an assessment?
The assessment finding targets the unassessed system receiving CUI, regardless of how well the enclave itself is controlled. Under the CMMC program, CUI must be handled only within assessed scope. Remediation usually means bringing the system into scope or removing CUI from it and documenting the correction; assessors generally don't accept a promise to fix it later as evidence.
Does encrypting CUI at rest take the storage system out of scope?
No for the storage system itself β encryption is a control, not a scoping exemption. The standard's scoping guidance keeps components that store CUI in scope even when the data is encrypted. Encryption may reduce the severity of a loss, but the asset still needs the applicable requirements, including key management under the cryptographic protection requirements.
This article is general educational information about CUI scoping and network architecture, not a legal opinion, compliance certification, or contractual interpretation. CUI obligations flow from your specific contract and the standards it cites, so confirm requirements with your contracting officer and a qualified compliance professional in your jurisdiction. If a security incident involving CUI has occurred, follow your contract's incident reporting timeline and contact the relevant authorities.