Data Residency

**Data residency** (also called data localization) requires that data be physically stored and processed within a defined territory. The requirement may arise from a specific regulation—GDPR mandates that EU personal data not leave the EEA without adequate protections, India's PDPB imposes localization requirements for sensitive categories, and China's PIPL prohibits cross-border transfer without explicit clearance—or from enterprise procurement rules that require data to remain in a particular cloud region.

For AI applications, data residency creates concrete architectural constraints. When an application calls a cloud-hosted language model, user prompts and tool results travel across network boundaries. If those prompts contain personal or regulated data, that transit may breach residency requirements unless the model endpoint is deployed in the approved region. Vector databases that store embeddings of personal data are subject to the same rules as the original records. [[rag]] pipelines that pull from regulated document stores must confirm that the retrieval index itself is co-located appropriately.

The common misconception is that using a major cloud provider automatically handles residency. It does not. Region selection, data-transfer agreements, and logging configurations must all be verified independently for each tier of the stack—model inference, embedding store, context cache, [[audit-trail]], and backup. A logging pipeline that ships raw prompts to a centralized observability cluster in another region can invalidate the whole architecture. [[confidentiality]] and data residency are related but distinct: an encrypted request sent to the wrong jurisdiction still violates residency even if it is never decrypted.

MCP connectors offer one architectural path forward: the model queries data through a local [[mcp]] server that runs inside the approved jurisdiction, so only query results—not bulk data—leave the region. See [[gemini-enterprise-agents]] for how enterprise platforms address residency through dedicated regional deployments and data-classification policies.

## Related Terms

- [[glossary/confidentiality|Confidentiality]] — the information-security property that data is accessible only to authorised parties - [[glossary/privilege|Attorney-Client Privilege]] — the confidentiality protection that restricts access to certain communications - [[glossary/rbac|RBAC (Role-Based Access Control)]] — the access-control model that grants permissions based on role rather than individual identity - [[glossary/audit-trail|Audit Trail]] — the immutable chronological record of every action taken, enabling forensic review - [[courses/gemini-enterprise-agents|Course: Gemini Enterprise Agents]] — hands-on practice with the concepts covered in this entry