Data scope
Identify what may be processed and where.
PRIVATE AI
Private AI with an agreed data boundary
Choose an architecture for sensitive work based on actual processing and operational requirements. Private-cloud and dedicated options retain their stated readiness limits.
Evidence snapshot: 2026-10-22 · review by 2026-11-22 · partial_read_only
THE PRACTICAL TASK
What to agree for your team
Architecture
Compare eligible local, hybrid and dedicated configurations.
Acceptance
Verify the chosen environment rather than assuming isolation from its name.
From requirements to a pilot
- 01
Classify the data
- 02
Choose the architecture
- 03
Verify the boundary
- 04
Agree the operating model
Controls
Model policyEarly Access
Server-side policy constrains providers, models, external routes, and fallback.
Production policy configuration pending.Technical evidence
- Scope
- workspace
- Evidence
- docs/business/model-policy.md
Deployment profiles
RU private cloudPlanned
A planned private-cloud deployment in Russia.
No customer infrastructure acceptance.Technical evidence
- Region
- ru_contract_specific
- Evidence
- docs/enterprise/deployment-profiles.md
Customer-dedicatedPlanned
A customer-dedicated deployment after technical assessment.
No dedicated production build evidence.Technical evidence
- Region
- contract_specific
- Evidence
- docs/enterprise/deployment-profiles.md
On-PremBeta
Customer-environment packaging with outbound access denied by default.
Signed artifacts and a customer restore drill are still required.Technical evidence
- Region
- customer_environment
- Evidence
- docs/enterprise/deployment-profiles.md;deploy/onprem
Evidence snapshot: 2026-10-22 · review by 2026-11-22 · partial_read_only