Classify
Classify every data class the workload touches and the sensitivity that governs it.
Done when: No workload proceeds with an unclassified data class in scope.
In-country Deployment Framework (KSA-READY)
In-country Deployment Framework
Residency, hosting, key custody and exit, decided per data class before the first workload is deployed.
The business problem
The AI is approved. Where it runs is not.
Hosting, residency, key custody and exit are decided late, so launches stall in security review.
What you hear inside the organisation
Can this data leave Saudi Arabia?
Who holds the encryption keys?
What if we need to change provider?
Security wants an architecture, not a promise.
Who it serves and what changes
Your teams, before and after the In-country Deployment Framework
Client teams and what they need
CISO and security
Residency and keys under control
IT infrastructure
A hosting choice per data class
Procurement and legal
Exit terms before signing
AI teams
Approval without redesign
The outcome you get
- Hosting decided per data class
- Keys held where policy requires
- A tested exit plan
- Security approval at the first review
A worked example
A government entity adopting an AI document assistant
The problem we saw
The use case was approved, but security rejected the hosting design twice because residency and key custody were unclear.
How the framework solved it
We decided hosting per data class, kept keys inside the entity, placed sensitive work on in-country infrastructure, and wrote the exit plan up front.
The business result
Security approved the design at the next review, and the same pattern now serves every new AI workload.
Illustrative example based on common client situations.
In short
What is the In-country Deployment Framework?
In-country Deployment Framework (KSA-READY): Residency, hosting, key custody and exit, decided per data class before the first workload is deployed.
Sovereignty questions surface late, usually at the security review, and stall a working system.
- How it moves
- Each step closes with a check before the next begins
- Owned by
- Run by architecture with security and legal; signed off by the risk committee.
- You keep
- Residency matrix, Key custody design, Rehearsed exit runbook
How it works
How the In-country Deployment Framework works
Sovereignty is an architecture decision with a test, not a hosting label. If the exit has never been rehearsed, it does not exist.
Place
Decide where each class is processed and stored: in-country, regional, or a controlled hybrid.
Done when: Every class has a residency ruling with the authority it answers to.
Custody
Decide key custody, administrative access and the logging that proves both.
Done when: The organisation can demonstrate who could technically read the data.
Model routing
Route each request class to a model tier that satisfies its residency and sensitivity.
Done when: No sensitive class can be routed outside its permitted boundary, by configuration.
Exit
Rehearse the exit: export the data, the models, the prompts and the evaluation sets.
Done when: The exit has been performed once, not merely written into a contract.
Is it right for you?
Is the In-country Deployment Framework right for you?
A strong fit if
- You handle classified or regulated data
- Launches stall in security review
- You are choosing a cloud or AI provider
Probably not yet if
- All your data is public
- Hosting is already fixed and approved
Free assessment
Discuss the In-country Deployment Framework with our team
A working session with our senior team. We review what fits your situation, what does not, and where to start.
A focused conversation
We listen to your situation and goals.
Fit review
Which parts of this framework apply to you, and which do not.
A written view
Clear next steps, whether or not you work with us.
Go deeper
Applying the In-country Deployment Framework in practice
Open only the part you need: how it starts, what your team keeps, maturity levels, sector rules and the evidence a reviewer will ask for.
Sovereignty questions surface late, usually at the security review, and stall a working system. KSA-READY answers them at design time: where each data class may live, who holds the keys, what may cross a border and under which condition, and how the organisation would leave.
Version
v1.0
Last reviewed
August 2026
Family
Saudi Data & AI Standards
How this was developed: Built from delivery practice on Saudi and international engagements, then reconciled clause by clause against the published standards below. Revised when a standard is updated or a gate proves unworkable in the field.
Residency matrix
A ruling per data class with the authority behind it.
Key custody design
Who holds the keys, who can read what, and how that is logged.
Rehearsed exit runbook
An export that has actually been performed and timed.
Banking & finance
Outsourcing and cloud expectations require the residency ruling before the security review, not during it.
- Who reviews it
- SAMA, plus PDPL and NCA controls
- Systems in scope
- Core banking, cards, KYC and AML systems, credit bureau feeds, the general ledger
- Where it lands first
- KYC and AML review, credit decisioning support, complaint and dispute handling
1
Classify
- Data class map
- Sensitivity ruling
2
Place
- Residency matrix
- Hosting decision record
3
Custody
- Key custody design
- Privileged access log
4
Model routing
- Routing policy
- Boundary test results
5
Exit
- Exit runbook
- Rehearsal report
Sovereignty is an architecture decision with a test, not a hosting label. If the exit has never been rehearsed, it does not exist.
Stage
Exit gate
Artefacts
1
Classify
Classify every data class the workload touches and the sensitivity that governs it.
Exit gate
No workload proceeds with an unclassified data class in scope.
Artefacts
- Data class map
- Sensitivity ruling
2
Place
Decide where each class is processed and stored: in-country, regional, or a controlled hybrid.
Exit gate
Every class has a residency ruling with the authority it answers to.
Artefacts
- Residency matrix
- Hosting decision record
3
Custody
Decide key custody, administrative access and the logging that proves both.
Exit gate
The organisation can demonstrate who could technically read the data.
Artefacts
- Key custody design
- Privileged access log
4
Model routing
Route each request class to a model tier that satisfies its residency and sensitivity.
Exit gate
No sensitive class can be routed outside its permitted boundary, by configuration.
Artefacts
- Routing policy
- Boundary test results
5
Exit
Rehearse the exit: export the data, the models, the prompts and the evaluation sets.
Exit gate
The exit has been performed once, not merely written into a contract.
Artefacts
- Exit runbook
- Rehearsal report
Standard
What this framework satisfies
NCA
Cloud cybersecurity controls, privileged access and key management.
PDPL
Transfer conditions and processing location evidenced per data class.
SDAIA
National AI principles applied to deployment and model selection.
NDMO
Data storage, sharing and interoperability specifications.
ISO/IEC 42001
Supplier and lifecycle controls covering third-party model providers.
Sovereignty as a label
A regional data centre with foreign key custody answers the marketing question, not the risk one.
One boundary for all data
Treating every class as top secret makes the programme unaffordable. Rule per class.
Untested exit clauses
A contractual right to leave that has never been executed is an assumption, not a control.
Built on
- 1
NCA Essential and Cloud Cybersecurity Controls
National Cybersecurity Authority control catalogues.
- 2
Personal Data Protection Law (PDPL) and its implementing regulations
Lawful basis, subject rights, transfer and retention.
- 3
NDMO data management and personal data protection standards
National Data Management Office specifications and domains.
- 4
SDAIA AI Ethics Principles
Saudi Data & AI Authority, national principles for AI.
- 5
ISO/IEC 42001: AI management systems
International Organization for Standardization, management system requirements for AI.
Citations describe the parts of each standard the framework answers. Bilytica is not an accreditation or certification body.
In Saudi Arabia
Written for Saudi Arabia, not adapted to it
This family exists because the Saudi control set, Arabic quality and the in-country deployment ruling are not add-ons to a global method. They decide whether a system is allowed to run at all.
AuthorityWhat it requiresWhat we buildWhat you keep
SDAIA, PDPL
National AI ethics principles applied as testable controls.
Each principle becomes a check with a pass condition and a named reviewer.
Ethics review record per release, in Arabic.
Arabic quality standard
Saudi Arabic output has to be correct in register, terminology and numerals, in front of a customer.
Arabic evaluation sets built from Saudi content, scored before release.
Arabic quality score with the failing cases attached.
NCA ECC
Deployment location and network boundaries are a documented decision.
A deployment ruling per data class states what may leave Saudi Arabia and what may not.
Signed residency ruling referenced by the architecture record.
Where it runs
The default is in-country, in your environment. Any exception is written down with the data class, the reason and an expiry date.
Arabic, as it is spoken here
Arabic is the governing version of every artefact in this family, with English supplied alongside.
FAQ
Frequently Asked Questions
Does in-country mean no external models at all?
No. It means each request class is routed to a tier its classification permits, and the boundary is enforced by configuration and tested.
When should this run?
Before the first workload is deployed. Retrofitting residency after go-live is the most expensive rework in an AI programme.
How does it relate to the Sovereign AI Framework?
KSA-READY produces the deployment rulings; the Sovereign AI Framework operates the architecture those rulings define.
Where this connects
Rest of the library
DFFData Foundation FrameworkAIRMAI Readiness ModelVALUEAI Value FrameworkTRUSTAI Assurance FrameworkDELIVERAI Delivery FrameworkAGENTAgent Operations FrameworkSOVEREIGNSovereign AI FrameworkDMMData Management Maturity ModelDPAData Platform Architecture FrameworkDQFData Quality & Observability FrameworkDECIDEAnalytics & Decision FrameworkDGOVData Governance Operating ModelSDCFSaudi Data Compliance FrameworkAQSArabic Quality Standard
Topic cluster
Explore the topic cluster
Curated links across our Saudi hubs, sector pages and long-form blog research.