
Knowledge Base Data Residency: What to Verify Before You Choose a Region
Knowledge Base Data Residency is the documented geographic boundary in which specifically identified knowledge-base data is stored, replicated, or contractually committed to remain.
That definition is narrower than many vendor claims suggest. Selecting an EU, US, UK, or other hosting region does not automatically tell you where search indexes are built, where backups are kept, where support personnel can access content, where AI inference runs, or where third-party applications process data.
A useful residency review therefore asks five separate questions:
- Where is each data type stored?
- Where is it processed?
- Where is it replicated or backed up?
- Who can access it, and from where?
- Which of those answers are contractually committed rather than merely documented or configured?
This guide explains how to map those boundaries, evaluate vendor claims, review search and AI data flows, assess contracts and subprocessors, compare deployment models, test a configured region, and maintain evidence after purchase.
Data residency is not the same as data location, sovereignty, or compliance
Vendors, regulators, lawyers, and architects sometimes use similar terms differently. Treat them as operational questions rather than interchangeable labels.
| Term | Practical meaning | What it does not prove |
|---|---|---|
| Data residency | Where defined data is stored or pinned under a technical and contractual commitment. | That all processing, support access, backups, and integrations remain in the same region. |
| Data location | The current physical or logical location of data or a service component. | That the location is durable or contractually guaranteed. |
| Processing location | Where data is viewed, transformed, indexed, analyzed, embedded, moderated, or used to generate an output. | That the data is stored there permanently. |
| Inference residency | Where an AI model performs inference on in-scope content. | Where source documents, embeddings, logs, or outputs are stored. |
| Data sovereignty | The legal and governmental authority that may apply to data, infrastructure, providers, or access. | That data is physically stored within one national border. |
| Data localization | A legal, regulatory, contractual, or internal requirement that certain data remain or be processed within a defined boundary. | That every data category is subject to the same requirement. |
| Cross-border transfer | Movement or legally recognized access to personal or regulated data across a relevant boundary. | That the data must always be physically copied to another country. |
| Regional hosting | A vendor offering that may apply only to specified products, plans, data types, or customers. | That the entire platform is region-bound. |
| Data isolation | Logical or physical separation among customers, tenants, or workloads. | That the isolated data stays in a particular country. |
| Encryption | Protection of data in transit or at rest. | Where the data is stored, processed, or accessed. |
Microsoft’s current Microsoft 365 data-residency documentation, for example, distinguishes data at rest from data in processing. Its comparison of residency offerings also shows that some commitments concern specified data stored at rest, while the EU Data Boundary includes broader storage and processing commitments for in-scope services.

The distinction matters because a knowledge platform can store article bodies in one region while processing search queries, support requests, analytics, or AI prompts somewhere else.
Define the requirement before evaluating vendors
“We need EU hosting” is not a complete requirement. It does not identify the data, the legal or contractual driver, the services in scope, or the exceptions the organization can accept.
Before contacting vendors, document:
| Requirement field | What to define |
|---|---|
| Business objective | The risk, customer commitment, policy, procurement condition, or operational need driving the requirement. |
| Jurisdiction | The country, group of countries, legal area, or contractual boundary that matters. |
| Data categories | The exact knowledge, identity, search, log, backup, integration, and AI data included. |
| Data subjects | Customers, employees, partners, patients, citizens, or other affected groups. |
| Storage boundary | Where primary data, replicas, and backups may be stored. |
| Processing boundary | Where indexing, analytics, AI, translation, and other processing may occur. |
| Access boundary | Where support, operations, professional-services, and subprocessor personnel may access data. |
| Exceptions | Emergency support, abuse monitoring, disaster recovery, or customer-enabled transfers that are permitted. |
| Required evidence | Documentation, DPA language, order-form terms, audit material, architecture records, and technical validation. |
| Owner | The person accountable for approving the requirement and its exceptions. |
| Review trigger | Feature launches, new integrations, contract renewal, region changes, or legal developments. |
Potential drivers include customer contracts, internal risk policy, public-sector procurement, sector-specific obligations, cross-border transfer assessments, latency, business continuity, or customer assurance.
Do not assume a legal requirement exists merely because a local region feels safer. Ask qualified legal, privacy, security, procurement, and architecture teams to identify the actual obligation.
Inventory every knowledge-base data type
Article bodies are only one part of a knowledge system. A realistic residency review covers every copy and derivative that the platform creates.
| Data family | Examples |
|---|---|
| Core knowledge content | Articles, drafts, versions, comments, community posts, attachments, images, video, templates, translations, and exports. |
| Identity and access | User profiles, email addresses, groups, roles, SSO attributes, SCIM data, sessions, authentication logs, and access rules. |
| Search data | Indexes, query logs, click events, synonyms, ranking features, zero-result data, and cached snippets. |
| Operational data | Audit logs, application logs, errors, telemetry, notifications, usage analytics, and support tickets sent to the vendor. |
| Backup and recovery | Operational backups, customer-created backups, snapshots, replicas, disaster-recovery copies, archives, and restore metadata. |
| Integration data | API and webhook payloads, connector caches, imported content, exported files, marketplace-app data, and credential references. |
| AI data | Parsed text, chunks, embeddings, vector indexes, reranking inputs, prompts, retrieved passages, generated answers, conversation history, evaluation datasets, logs, and feedback. |

For each data type, record whether it is stored, replicated, processed, transferred, cached, logged, backed up, accessible to support, sent to a subprocessor, used by AI, included in deletion, and included in export.
Atlassian’s current data-residency scope illustrates why this level of detail matters. Confluence page content, attachments, comments, metadata, whiteboards, databases, and search data can be pinned, while permission and restriction configuration data and space keys are listed outside the pinned scope. The answer is therefore not simply “Confluence is in the EU”; it depends on the data category.
Build a data-flow and residency matrix
A vendor’s regional-hosting statement becomes useful only after it is mapped to the organization’s actual data flows.
| Matrix field | What to capture |
|---|---|
| Data type | The specific article, attachment, log, index, prompt, backup, or identity record. |
| Source | The system or user action that creates the data. |
| Destination | The platform, subprocessor, integration, or storage service that receives it. |
| Primary storage region | The committed location of the main stored copy. |
| Replica region | Availability-zone, regional, or cross-region replicas. |
| Backup region | Operational, customer-created, and disaster-recovery backup locations. |
| Processing region | Where transformation, indexing, analytics, or moderation occurs. |
| AI inference region | Where model inference is performed. |
| Support-access region | Where authorized personnel may access the data. |
| Subprocessor | The external provider involved in storage, processing, support, or delivery. |
| Transfer mechanism | The contractual or technical basis for a relevant cross-border transfer. |
| Retention and deletion | How long copies remain and how deletion propagates. |
| Contract commitment | The agreement, DPA, order form, or service term that establishes the obligation. |
| Evidence | The official page, contract clause, audit artifact, or technical result supporting the entry. |
| Exception | Documented circumstances that allow data to leave the preferred boundary. |
| Owner | The internal role responsible for the answer and follow-up. |
Do not fill gaps with assumptions. Mark each answer as:
- Contractual.
- Vendor-documented.
- Customer-configured.
- Technically observed.
- Inferred.
- Unknown.
- Requires vendor confirmation.
An unknown is a valid result. Hiding uncertainty produces a more dangerous assessment than recording it explicitly.
Separate storage, processing, access, replication, and transfer
A regional deployment can have several different geographic boundaries at the same time.
A provider may:
- Store article bodies in one region.
- Keep availability replicas in another zone within that region.
- Run analytics or search ranking in a different geography.
- Allow global support staff to access content under controlled conditions.
- Create AI embeddings or generate answers through another service.
- Send selected records to a customer-enabled CRM, webhook, or marketplace application.
Salesforce’s current Hyperforce data-residency documentation makes this distinction explicit. Local data storage means customer data is stored at rest in the country, but data may still be transferred out for processing, support, service operation, or customer-initiated actions. Salesforce’s separate Hyperforce Operating Zone offering provides broader regional storage, processing, access, and support commitments for specified regions and covered products.
Therefore, “stored in-region” should never be used as evidence that all processing and access also remain in-region.
Evaluate backups, replicas, and disaster recovery
Backups are often the least understood part of a residency commitment. They may be covered, excluded, or governed by separate rules.
Ask:
- Are operational backups in scope?
- Do replicas remain inside the selected boundary?
- Are disaster-recovery copies stored in another country?
- Do old backups move after a regional migration?
- How long do old copies remain?
- How are deleted articles removed from backups?
- Where is a restore performed?
- What happens during regional failover?
- Are backup configuration and reports covered by the same commitment?
Atlassian states that automated backups for pinned in-scope app data are stored in the selected location, but its scope table separately distinguishes backup contents from backup policies, reports, configuration, and metadata. Review both the stored data and the operational metadata around it.

Keep these concepts distinct:
- Availability replica.
- Operational backup.
- Customer export.
- Disaster-recovery copy.
- Archive.
- Legal hold.
An export is not a complete backup, and deleting a live article does not necessarily delete every backup, log, replica, or legal-hold copy immediately.
Check search residency separately from content storage
A searchable knowledge base creates new data derived from the articles it indexes.
Review:
- Where the full-text index is stored.
- Whether the index contains article text, metadata, permissions, or snippets.
- Where search queries and click events are logged.
- Where ranking and analytics run.
- Whether search caches cross regional boundaries.
- How permission changes and deletions propagate to the index.
- Whether federated search retrieves content at query time or creates a local copy.
- Whether connected enterprise-search systems preserve source permissions.
Do not assume the search service follows the same residency commitment as the article database.
Search queries may also contain names, account details, symptoms, error messages, or sensitive business terms. Treat query logs as data requiring their own classification and retention rules.
Evaluate AI, embeddings, vector stores, and inference separately
AI-assisted search and support create a second residency architecture layered on top of the knowledge base.
Map every stage:
- Document extraction and OCR.
- Parsing and chunking.
- Embedding generation.
- Vector storage.
- Reranking.
- Prompt construction.
- Model inference.
- Safety and abuse monitoring.
- Conversation logging.
- Evaluation and feedback.
Distinguish:
- Source-content residency.
- Embedding residency.
- Vector-index residency.
- Inference residency.
- Prompt and response retention.
- Model-training policy.
- Abuse-monitoring retention.
- External model providers.
OpenAI’s current data-residency and inference-residency guidance is one clear example of this separation. Product-specific commitments may differ between storage, inference, logs, features, and customer tiers.

OpenAI’s current ChatGPT Business documentation also states that selecting a non-US storage region does not determine inference location and that safety and abuse-monitoring data remains in the United States.
Ask whether:
- AI can be disabled.
- Only approved sources are indexed.
- In-region inference is available.
- All modalities receive the same commitment.
- Preview or beta features are covered.
- Generated answers and chat history stay in-region.
- Third-party AI subprocessors are involved.
- Customer content is used for model improvement.
- Deleted sources are removed from retrieval indexes.
- Permissions are enforced before retrieval.
A knowledge base is not fully region-resident merely because its article database is regional when its AI pipeline is not.
Review third-party apps, connectors, and subprocessors
A platform’s residency commitment may stop at the platform boundary.
Separate services may process data through:
- Marketplace applications.
- Browser extensions.
- Analytics providers.
- Translation services.
- CRM and ticketing integrations.
- Search and AI providers.
- Monitoring and email services.
- Customer-configured webhooks and API destinations.
Inspect the vendor’s subprocessor list, DPA, marketplace terms, app permissions, hosting locations, retention, support access, transfer mechanisms, and deletion behavior.
Atlassian notes that eligible Forge apps may move with supported host apps in certain configurations, while other applications may remain outside the pinned location. Zendesk’s regional policy likewise distinguishes covered functionality from products, secondary data, beta features, and third-party services that are not covered by the same hosting commitment.
Enabling an integration can intentionally move data outside the selected region. Document that as a customer-approved transfer rather than assuming the original platform controls the destination.
Evaluate support and operational access
Data may remain stored in one region while authorized staff access it from another.
Ask:
- Where support personnel are located.
- Whether they can access article content, attachments, logs, or backups.
- Whether access requires customer or managerial approval.
- Whether privileged access is time-limited and logged.
- Whether regional support is available.
- Whether emergency access creates an exception.
- Whether subcontractors provide support or migration services.
- Whether diagnostic bundles leave the selected region.
Salesforce’s Hyperforce Operating Zone documentation is a useful example of a broader commitment: it describes regional storage and processing controls together with restrictions on Salesforce-initiated transfers and regional support access for covered deployments.
Do not treat human-access location as a data-center question only. It is a separate control and contract issue.
Distinguish a region setting from a contractual commitment
A region selector is a technical setting. It may not create a legally binding promise.
Review the evidence in this order:
- Service-specific contract or order-form commitment.
- Data Processing Addendum or regional-hosting policy.
- Service-specific terms and security exhibits.
- Subprocessor list and architecture documentation.
- Official product documentation.
- Trust-center material and audit reports.
- Vendor-questionnaire responses.
- Marketing pages and sales presentations.
Marketing pages can identify possible offerings, but critical residency requirements should be reflected in binding or service-specific documentation.
A usable commitment should identify:
- Covered products and services.
- Covered data types.
- The geographic boundary.
- Whether it covers storage, processing, access, backups, or support.
- Customer actions that can move data.
- Exceptions and failover behavior.
- Change-notification procedures.
- End-of-service export and deletion behavior.
Zendesk provides a current product-specific example. Its Regional Data Hosting Policy limits commitments to entitled customers and covered functionality. It separately lists service-data types, supported hosting regions, exceptions, secondary data, and product-specific transfer limitations.
Data residency is not automatic legal compliance
This section is general information, not legal advice.
GDPR and international transfers
GDPR does not create a universal rule that all personal data must remain inside the EU or EEA. The European Commission’s international-transfer guidance describes mechanisms such as adequacy decisions, Standard Contractual Clauses, Binding Corporate Rules, codes of conduct, certifications, and specific derogations.
A residency decision may support a privacy strategy, but organizations still need to assess applicable transfers, providers, purposes, subprocessors, contractual safeguards, and other legal obligations.
HIPAA and cloud hosting
HIPAA does not universally require ePHI to be stored only in the United States. HHS states that covered entities and business associates may use cloud providers, including providers that store ePHI outside the US, when they satisfy the applicable HIPAA requirements, enter into an appropriate BAA, and perform the required risk analysis and risk management.
Geographic location can still affect risk, enforceability, contracts, and organizational policy. The absence of a universal US-only requirement does not make every overseas deployment appropriate.
FedRAMP and authorization boundaries
Federal buyers should examine the authorized cloud-service boundary, in-scope services, integrations, data flows, and supporting systems rather than relying on a vendor-wide logo. FedRAMP’s published RFC on boundary policy is explicitly draft guidance, not a finalized policy, but it illustrates the importance of identifying services that handle federal information or directly affect its confidentiality, integrity, or availability.
Customer contracts and internal policies may impose stricter conditions than a general law. Review the exact agency, jurisdiction, data category, authorization, and contract with qualified specialists.
Compare deployment models by responsibility and scope
| Deployment model | Residency advantage | Key questions |
|---|---|---|
| Multi-tenant SaaS | Vendor-managed regional offerings and contractual commitments may reduce operational burden. | Which services, plans, data types, backups, support access, AI, and subprocessors are covered? |
| Dedicated managed cloud | May provide greater isolation and customized regional controls. | Which layers are dedicated, and do processing, support, and subprocessors follow the same boundary? |
| Self-managed public cloud | The customer can choose cloud regions and design the application architecture. | Who manages the application, database, search, backup, monitoring, support, and AI services? |
| Private cloud | Can provide dedicated infrastructure and tighter network control. | Who operates it, where are backups and support services, and which external SaaS dependencies remain? |
| On-premises | Provides direct infrastructure-location control. | Do off-site backups, remote support, email, analytics, integrations, and AI create external flows? |
| Hybrid | Can separate public, private, and regulated knowledge workloads. | How are identity, search, synchronization, deletion, monitoring, and source authority coordinated? |
No deployment model is automatically superior. Choose the least complex architecture that satisfies verified storage, processing, access, backup, and contractual requirements.
Encryption and key control do not replace residency analysis
Encryption can reduce exposure, but it does not change the geographic location of the encrypted data or the system processing it.
Evaluate:
- Encryption in transit and at rest.
- Provider-managed and customer-managed keys.
- Bring Your Own Key and external key-control options.
- Key storage and key-administrator locations.
- Key rotation and revocation.
- Backup and search-index encryption.
Customer-managed keys may improve control, but they do not eliminate all support-access, sovereignty, transfer, or processing questions.
Verify regional availability and migration behavior
Residency availability may depend on product, plan, add-on, customer status, account creation date, or deployment region.
Confirm:
- Which regions are generally available now.
- Which products and features work in each region.
- Which plans, add-ons, or contracts are required.
- Whether the setting applies at the tenant, app, site, user, or resource level.
- Whether existing data can be migrated.
- Whether migration creates downtime.
- Which data is excluded.
- What happens to old backups and replicas.
- Whether sandboxes and test environments follow the same boundary.
- Which integrations must be reconnected.
- Whether rollback is available.
- How future region changes are handled.
Do not treat roadmap statements as purchasing commitments. Salesforce’s current Hyperforce documentation, for example, explicitly warns customers not to rely on planned regions when making purchasing decisions.
Vendor due-diligence checklist
| Area | Questions to ask |
|---|---|
| Scope | Which exact products, modules, features, plans, and environments are covered? |
| Data types | Which content, attachments, identities, logs, indexes, backups, and AI data are in or out of scope? |
| Storage | Where are primary data, replicas, backups, and archives stored? |
| Processing | Where do search, analytics, translation, OCR, moderation, and other processing occur? |
| AI | Where are embeddings, vector indexes, prompts, inference, outputs, logs, and safety data handled? |
| Support | Where are support and operations personnel located, and what content can they access? |
| Subprocessors | Which providers receive content, and what are their locations, retention, and transfer mechanisms? |
| Integrations | Do marketplace apps and customer-enabled connectors follow the same boundary? |
| Backup and recovery | What happens during failover, restore, region migration, and live-data deletion? |
| Contracts | Which document contains the commitment, and what exceptions and notice requirements apply? |
| Migration | Can current data move, how long does it take, and what stays behind? |
| Exit | What can be exported, what is deleted, and how long do residual copies remain? |
Require evidence, not yes-or-no answers
For every vendor response, record:
- The claim.
- The exact scope.
- The evidence URL or document.
- The document owner and revision date.
- Whether the answer is contractual or informational.
- The applicable product, plan, region, and customer type.
- The exception.
- The internal reviewer.
- The follow-up question.
- The final decision.
An undocumented sales answer is not equivalent to a contractual commitment. A certification logo is not proof that every component of the knowledge platform follows the required regional boundary.
Test the configured residency
Documentary and contractual review should be supplemented by authorized technical validation.
Possible checks include:
- Verify the tenant’s configured region.
- Review vendor-provided instance and region details.
- Inspect audit, integration, and support-access logs.
- Test backup and restore behavior where supported.
- Validate the results of a regional migration.
- Review AI endpoint and model-region configuration.
- Confirm connected applications and data destinations.
- Test deletion from article, search, and AI indexes.
- Confirm that excluded users cannot retrieve restricted content.
- Update the data-flow matrix after enabling a new connector.
Do not perform unauthorized scanning or testing that violates vendor terms. An IP address or network route alone may not prove the contractual location of every data copy.
Operate residency as an ongoing control
Residency can change after purchase as vendors add features, subprocessors, AI services, regions, or new infrastructure.
Monitor changes to:
- Product architecture.
- Hosting regions.
- AI providers and models.
- Subprocessors.
- Regional-hosting policies.
- DPAs and service terms.
- Backup and disaster-recovery services.
- Integrations and marketplace apps.
- Support models.
- Plan packaging.
- Acquisitions and product consolidation.
- Beta and preview features.
Assign owners for subprocessor notices, document review, contract renewal, integrations, legal review, technical validation, exception management, and customer commitments.
Trigger a review when the organization enables a new AI feature, adds a connector, migrates region, enters another country, adds a new data category, or receives a vendor-policy change.
Implementation roadmap
| Phase | Primary work | Evidence before proceeding |
|---|---|---|
| 1. Define requirements | Business drivers, jurisdictions, data types, boundaries, exceptions, evidence, and owners. | The requirement is specific enough to test and contract. |
| 2. Map current flows | Repositories, search, AI, backups, integrations, support access, subprocessors, and transfers. | Unknown and unapproved flows are visible. |
| 3. Evaluate options | Vendor documentation, products, plans, regions, contracts, deployment models, and exceptions. | Each option has comparable scope and evidence. |
| 4. Design controls | Configuration, permissions, encryption, integration restrictions, AI source controls, monitoring, and exception handling. | Owners approve the intended controls and residual risk. |
| 5. Contract and configure | DPA, order form, regional terms, migration, technical setup, and ownership. | The contractual and configured scopes agree. |
| 6. Validate | Region, data flows, backups, search, AI, integrations, deletion, permissions, and recovery. | Tests and documentary evidence support the requirement. |
| 7. Operate and review | Evidence refresh, vendor changes, subprocessors, architecture review, incidents, and renewal. | Changes are identified and reassessed before they create an unapproved flow. |
Worked example: a fictional B2B knowledge platform review
This example is fictional. Northstar Systems operates a public customer help center and a private employee knowledge base. It serves EU and US customers, stores diagnostic attachments, uses an enterprise-search connector, and plans to add an AI support assistant.
Several EU customer contracts require specified customer-content data to remain stored within the EU. The contracts allow documented cross-border support access and approved processors when appropriate safeguards are in place.
| Evaluation step | Northstar’s illustrative action |
|---|---|
| Define the requirement | The team identifies article content, customer attachments, identity data, search indexes, and backups as in-scope stored data. |
| Inventory data | It discovers that search queries, support bundles, AI prompts, embeddings, and analytics are separate data categories. |
| Map flows | The team documents article storage, backup regions, search indexing, AI processing, CRM exports, and global support access. |
| Review vendor scope | The preferred vendor offers EU storage for article content but treats some analytics, support, and third-party apps separately. |
| Review contracts | Procurement records the covered service, in-scope data, support exceptions, subprocessors, notice process, and deletion terms. |
| Configure controls | Northstar enables EU hosting, restricts unapproved connectors, excludes sensitive attachments from AI, and limits support access. |
| Validate migration | The team verifies the tenant region, checks migrated content, reviews old-backup handling, and tests deletion from search. |
| Monitor | Security owns subprocessor review, RevOps owns connectors, the knowledge team owns content classification, and procurement reviews terms at renewal. |
Northstar does not claim that the platform is “fully EU-only.” It records the approved storage commitment, the separate processing and support exceptions, and the features excluded from AI until their data flows are verified.
Common Knowledge Base Data Residency failures
| Failure | Corrective action |
|---|---|
| Asking only where the primary database is hosted | Inventory content, attachments, identity, search, logs, backups, integrations, and AI separately. |
| Treating a region selector as compliance | Map the setting to the actual requirement, legal review, and contract scope. |
| Ignoring processing location | Document indexing, analytics, moderation, translation, and AI processing. |
| Ignoring old backups | Ask whether historical copies migrate, expire, or remain in the original region. |
| Ignoring support access | Review personnel location, approval, logging, emergency access, and contractors. |
| Ignoring marketplace apps | Review each application’s contract, subprocessors, permissions, location, and deletion. |
| Relying on sales slides | Require official documentation and binding service-specific terms. |
| Assuming all modules share the same scope | Evaluate each product, feature, service, and AI component separately. |
| Treating encryption as residency | Record encryption as a separate security control alongside location and access. |
| Assuming on-premises means no cross-border flows | Map remote support, SaaS integrations, backups, email, monitoring, and AI. |
| Using roadmap regions as commitments | Base purchasing decisions on generally available and contractually supported capabilities. |
| Failing to document exceptions | Record emergency, support, failover, monitoring, and customer-enabled transfer exceptions explicitly. |
Reusable data-residency worksheet
| Requirement field | What to record |
|---|---|
| Business requirement | The contract, policy, risk, or operational driver. |
| Jurisdiction | The relevant country, region, or legal boundary. |
| Data subject | Customer, employee, partner, citizen, patient, or another group. |
| Data classification | Public, internal, confidential, personal, regulated, or highly sensitive. |
| Knowledge source | Help center, employee KB, CRM, ticketing system, file repository, or community. |
| Data type | Article, attachment, log, index, backup, prompt, embedding, or output. |
| Flow field | What to record |
|---|---|
| Source and destination | The system creating and receiving the data. |
| Storage region | Primary stored copy. |
| Replica and backup region | Availability, backup, and DR locations. |
| Processing region | Indexing, analytics, transformation, or moderation. |
| AI inference region | Where the model executes inference. |
| Support-access region | Where authorized people can access the data. |
| Subprocessor | The external provider involved. |
| Transfer mechanism | The relevant contractual or legal safeguard. |
| Control field | What to record |
|---|---|
| Retention and deletion | Live, backup, log, index, and subprocessor behavior. |
| Export | Included data, format, and portability limitations. |
| Encryption and keys | At-rest, in-transit, key ownership, and key location. |
| Contract document | DPA, order form, regional policy, or service terms. |
| Plan or add-on | The entitlement required for the commitment. |
| Exception | Support, emergency, monitoring, failover, or customer-enabled transfer. |
| Evidence | Official link, clause, report, or test result. |
| Owner and review date | The accountable reviewer and next review trigger. |
| Decision | Approved, conditionally approved, rejected, or unresolved. |
The operating principle
Knowledge Base Data Residency is a data-flow and contract question, not a map-pin question. The meaningful unit of analysis is not the vendor’s logo or primary database region. It is each data type, each copy, each processor, each human-access path, each AI stage, and each contractual exception.
Define the requirement precisely, map the complete flow, verify the scope in current official documentation and contracts, test the configured environment, and repeat the review whenever the product or architecture changes.
Sources verified: November 9, 2026.



