> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/ai-enterprise/ai-factory/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/ai-enterprise/ai-factory/_mcp/server.

# Government Industry Profile

> Government-specific guidance for sovereign, restricted, disconnected, and high-assurance AI factory environments.

StatusPublished profile ReviewersPublic sector and security required Sources

[Primary-source index](/ai-enterprise/ai-factory/resources/source-library)

The Government Industry Profile applies sector-specific requirements to the core NVIDIA Enterprise AI Factory Reference Design. It contains only differences from the core. Shared workload, platform, observability, operations, and hardware guidance remains authoritative in its original layer.

> **Warning**
>
> This profile is architecture guidance. Deploying organizations remain responsible for end-to-end security, regulatory compliance, mission authorization, product support, and acceptance within their environment.

#### [Government Ecosystem Solutions](/ai-enterprise/ai-factory/ecosystem-integrations/solution-catalog?profile=government)

Explore the catalog of published offerings associated with the Government profile.

## Change Legend

The profile uses the following terms to distinguish inherited guidance from Government-specific changes:

| Type       | Meaning                                                                                                              |
| ---------- | -------------------------------------------------------------------------------------------------------------------- |
| Inherited  | Core guidance applies without a profile-specific architecture change.                                                |
| Extended   | The core capability remains, but additional requirements or constraints apply.                                       |
| Added      | The profile introduces an additional pattern or control.                                                             |
| Restricted | A core assumption is excluded by default and can be used only when the authorization boundary explicitly permits it. |

## Core-to-Government Change Matrix

Apply the following changes to the corresponding core layers:

| Core Layer                                                                      | Type       | Government Change                                                                                                                                                                                                                                                                                                                                                                                                                                             |
| ------------------------------------------------------------------------------- | ---------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| [Workloads](/ai-enterprise/ai-factory/core-reference-design/workloads)                                   | Extended   | Include mission criticality, data classification, residency, releasability, cross-domain paths, and authorization constraints in the workload envelope.                                                                                                                                                                                                                                                                                                       |
| [AI Platform](/ai-enterprise/ai-factory/core-reference-design/ai-platform)                               | Extended   | Use approved software sources, controlled promotion, approved cryptographic modes, model-release controls, and lifecycle paths appropriate to the authorization boundary.                                                                                                                                                                                                                                                                                     |
| [Platform Services](/ai-enterprise/ai-factory/core-reference-design/platform-services)                   | Added      | Where required, use cryptographic modules with active [CMVP](https://csrc.nist.gov/projects/cryptographic-module-validation-program) validation and operate them according to the applicable certificate and security policy. Apply product-specific STIG hardening, defined [NIST SP 800-53 Rev. 5](https://csrc.nist.gov/pubs/sp/800/53/r5/upd1/final) control baselines and tailoring, attestation, supply-chain integrity, and continuous security audit. |
| [Platform Services](/ai-enterprise/ai-factory/core-reference-design/platform-services)                   | Added      | Mirror, scan, sign, transfer, approve, and promote models, containers, packages, and security intelligence without assuming internet access.                                                                                                                                                                                                                                                                                                                  |
| [Platform Services](/ai-enterprise/ai-factory/core-reference-design/platform-services)                   | Extended   | Increase identity assurance, audit retention, data-residency enforcement, privileged-access control, and incident reporting as required by the mission environment.                                                                                                                                                                                                                                                                                           |
| [Cloud-Native Platform](/ai-enterprise/ai-factory/core-reference-design/cloud-native-platform)           | Extended   | Operate cluster lifecycle, observability, identity integration, upgrades, backup, and recovery within restricted or disconnected networks.                                                                                                                                                                                                                                                                                                                    |
| [Hardware Foundation](/ai-enterprise/ai-factory/core-reference-design/enterprise-hardware-ra-foundation) | Inherited  | Select the Enterprise Hardware Reference Architecture, then apply procurement, sovereignty, facility, physical-security, and supply-chain constraints through their authoritative programs.                                                                                                                                                                                                                                                                   |
| External dependencies                                                           | Restricted | Hosted control planes, telemetry sinks, package sources, model endpoints, and external identity dependencies are excluded by default and require explicit authorization-boundary approval.                                                                                                                                                                                                                                                                    |

## Sovereign and Restricted Environments

* Define where data, models, keys, logs, telemetry, administrators, and support personnel can reside.
* Identify every external control plane, callback, package source, model source, license service, and telemetry destination.
* Provide local substitutes or approved transfer procedures for dependencies unavailable inside the boundary.
* Preserve authoritative time, identity, certificate, revocation, vulnerability, and artifact information in disconnected operations.
* Design for prolonged isolation, delayed updates, controlled media transfer, and local incident response.

For every external dependency, record the provider and service, exchanged payloads, authentication, network path, availability and outage behavior, local replacement, risk owner, and authorization decision. Use [NIST SP 800-37 Rev. 2](https://csrc.nist.gov/pubs/sp/800/37/r2/final) for authorization-boundary and risk-management context.

## Confidential Computing and Attestation

Where threat models require protection for data or models in use, apply current [NVIDIA Trusted Computing Solutions](https://docs.nvidia.com/nvtrust/) guidance. Define the measured environment, verifier, appraisal policy, evidence freshness, key-release decision, revocation, and failure behavior.

## Applicable Capabilities

* [Secure Agent Workspace](/ai-enterprise/ai-factory/capability-reference-designs/secure-agent-workspace) can provide governed agent execution in mission or restricted environments when its external dependencies are resolved inside the authorization boundary.
* [Inference Hub](/ai-enterprise/ai-factory/capability-reference-designs/inference-hub#compose-with-other-capabilities-and-profiles) can provide shared model access. Apply Government requirements to approved destinations, residency, external dependencies, and operating boundaries; assess disconnected-operation support and authorization separately for the chosen implementation.

## Government Source Set

Use the following primary sources when resolving Government-specific architecture decisions:

| Decision                               | Primary Source                                                                                                                                       |
| -------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------- |
| Cryptographic-module requirements      | [FIPS 140-3](https://csrc.nist.gov/pubs/fips/140-3/final) and [CMVP](https://csrc.nist.gov/projects/cryptographic-module-validation-program)         |
| Control catalog and baseline selection | [NIST SP 800-53 Rev. 5](https://csrc.nist.gov/pubs/sp/800/53/r5/upd1/final) and [NIST SP 800-53B](https://csrc.nist.gov/pubs/sp/800/53/b/upd1/final) |
| Control assessment                     | [NIST SP 800-53A Rev. 5](https://csrc.nist.gov/pubs/sp/800/53/a/r5/final)                                                                            |
| Risk management and authorization      | [NIST SP 800-37 Rev. 2](https://csrc.nist.gov/pubs/sp/800/37/r2/final)                                                                               |
| Product hardening guidance             | [DISA STIG Downloads](https://www.cyber.mil/stigs/downloads/)                                                                                        |
| Secure AI operations                   | [NVIDIA Secure AI Operations Guide](https://docs.nvidia.com/nvidia-secure-ai-operations-guide.pdf)                                                   |

Track module certificates, security policies, selected baselines, tailoring records, assessment results, authorization-boundary owners, and authorizing officials. Keep these items in the deployment's compliance evidence. This profile does not supply or approve that evidence.