VIRIAN
MENU
HomeOperating modelAgentic SOCSOAR comparisonDesign a proof

CONTROL HAS A LOCATION

Sovereign security operations.

Keep security evidence, operating authority and AI inference inside the boundary your organization requires, without changing the decision model.

13 minute readUpdated 15 July 2026Virian field note 04

01 / DEFINITION

Sovereignty is the ability to keep control.

Sovereign security operations keep sensitive security data, administrative authority and AI processing within approved legal, geographic and technical boundaries. The deployment may be on premises, in-country, private cloud or hybrid, provided control remains explicit and verifiable.

Data residency answers where data is stored. Sovereignty is wider. It asks who can administer the system, which laws and contracts apply, whether external services are required, where models run, how updates arrive and whether the operation can continue through a connectivity or geopolitical disruption.

For security operations, these questions are material. Incident evidence can contain credentials, network structure, employee identity, regulated records and details of defensive controls. The operating model should fit the boundary, not force the boundary to fit the software.

02 / THREE DIMENSIONS

Data, operations and models.

A credible sovereign design makes each dimension visible. Keeping storage in-country while support, keys or inference remain externally controlled may satisfy one requirement and fail another.

DIMENSION 01

Data sovereignty

Define where raw evidence, derived context, incident records, logs, backups and model prompts are stored and processed.

DIMENSION 02

Operational sovereignty

Control administrator identity, privileged access, keys, support paths, upgrades, monitoring and continuity from within the approved boundary.

DIMENSION 03

Model sovereignty

Choose where inference runs, which models can see each data class, how model artifacts arrive and whether the system can operate without an external API.

DIMENSION 04

Decision sovereignty

Keep policy, approval rights and action authority under the customer’s governance, regardless of infrastructure placement.

03 / PLACEMENT MODELS

One operating model. Four places to run it.

Scope, Hunt and Strike keep the same responsibilities while infrastructure, data movement and model routing are configured for the organization’s jurisdiction, assurance and availability requirements.

Security operators working in a modern customer-controlled facility
01 / CUSTOMER ESTATE

On premises

Virian, evidence stores, integrations and approved models run inside the organization’s facilities and administrative domain.

A secure modern data facility in an urban setting
02 / NATIONAL BOUNDARY

In-country colocation

A dedicated or approved facility provides local residency, connectivity and operating control without placing hardware on the customer campus.

Engineers operating local AI infrastructure in a secure facility
03 / ISOLATED TENANCY

Private cloud

Virian runs in a customer-controlled cloud environment with isolated networking, identity, keys, logging and model endpoints.

A controlled hybrid keeps sensitive evidence and policy local while routing only approved tasks or data classes to external services. The boundary is designed at field level, not described with a single “hybrid” label.

04 / LOCAL AI

Inference can stay with the evidence.

Local AI allows the investigation and decision engine to run inside the sovereign boundary. An organization may use locally hosted open-weight models, approved commercial models in a private environment, or a routed model set where each task and data class has an explicit destination.

This is not only a privacy choice. Local inference can support disconnected operations, predictable access control and continued service when an external provider is unavailable. It also creates responsibilities for capacity, model evaluation, patching, monitoring and lifecycle management.

MODE 01

Fully local

All inference runs on customer-controlled infrastructure. No incident content is sent to an external model service.

MODE 02

Approved private endpoint

Models run in an isolated private cloud or in-country service with customer identity, networking and data controls.

MODE 03

Policy-routed

Local models handle sensitive tasks. External models receive only approved data classes and explicitly permitted requests.

MODE 04

Disconnected ready

Core investigation and response continue without an external control plane, model API or routine internet dependency.

05 / PRODUCTION CONTROLS

The hard questions begin after location.

A deployment diagram does not prove sovereignty. The production design must cover the full operating lifecycle, including ordinary maintenance and exceptional access.

ControlDecision to makeEvidence to retain
Identity and privilegeWho can administer infrastructure, models, policy and integrations?Named identities, role assignments, approvals and privileged session logs
Keys and secretsWho generates, holds, rotates and recovers encryption keys and integration credentials?Key ownership, rotation history, recovery procedure and access trail
Network and egressWhich destinations are allowed, for which task and data class?Approved routes, firewall policy, proxy logs and denied connection events
Model lifecycleHow are model artifacts evaluated, signed, delivered, updated and rolled back?Model inventory, evaluation results, artifact provenance and change approval
Platform updatesHow do security fixes enter connected, intermittent and disconnected environments?Release provenance, signing, compatibility tests and installation history
Backup and recoveryWhich state must survive failure, and where may replicas exist?Backup location, restore tests, recovery objectives and failover exercises
Operations and supportCan support occur without unrestricted remote access or cross-border data movement?Access request, scoped session, commands performed and session closure

06 / DESIGN QUESTIONS

Turn “sovereign” into a testable architecture.

ASK 01

What must never leave?

Classify raw telemetry, credentials, identities, incident evidence, prompts, outputs, logs and support bundles separately.

ASK 02

Who can stop the system?

Define local authority to suspend actions, revoke credentials, isolate models and restore a known-good state.

ASK 03

What must work offline?

Specify which detection, investigation, approval, response and audit functions continue during disconnection.

ASK 04

How is control proven?

Require configuration, access, routing, update and recovery evidence that can be inspected by local assurance teams.

07 / PRIMARY SOURCES

Current sovereign cloud and local AI patterns.

DESIGN THE BOUNDARY FIRST

Place autonomy where you can govern it.

Map data classes, authority, model routing, continuity and support before infrastructure is selected.

Design the architecture