Source: Claude API data residency docs

Anthropic's per-request inference_geo (global or us, at 1.1x price for US-only) and immutable workspace geo for storage at rest.

Contents
  • What: official API documentation
  • Publisher: Anthropic
  • Date: accessed 2026-11-14
  • Original: Data residency

Summary

Separates where inference runs from where data is stored, and shows how to enforce both per workspace.

Key claims

  • Inference geo: inference_geo on POST /v1/messages (or a workspace default): "global" (default, any geography) or "us" (US infrastructure only). The response usage.inference_geo reports where inference actually ran.
  • Workspace geo: where data is stored at rest and where endpoint-side processing (image transcoding, code execution) runs; set at workspace creation, immutable; only "us" exists today.
  • Workspace policy: allowed_inference_geos (requests outside it error) and default_inference_geo, settable in Console or Admin API (data_residency). Legacy US-only opt-outs were migrated to allowed_inference_geos: ["us"].
  • Supported on Claude 4.6 and later; earlier models return 400 if the parameter is sent. Supported in the Batch API per request.
  • Pricing: US-only inference is 1.1x on all token categories for Claude 4.6+ models; it also burns Priority Tier capacity at 1.1x.
  • Also available on Claude Platform on AWS and Anthropic-managed Claude on Google Cloud (workspace geo not configurable there). Not applicable on Bedrock or Google-managed Vertex (region set by endpoint or inference profile), nor on Microsoft Foundry (use US Data Zone Standard), nor the OpenAI-compatible endpoint.
  • Rate limits are shared across geos. Eligible for ZDR (except Covered Models) and HIPAA readiness.

Feeds

PII and data privacy

Suggest changes →