Germany-only data residency with Azure AI Foundry Regional Standard deployments

Houssem 0 Zuverlässigkeitspunkte
2026-07-21T11:32:22.73+00:00

Hello Microsoft team,

We are currently evaluating Azure AI Foundry for an enterprise AI knowledge platform with strict GDPR and German data residency requirements.

We have reviewed the documentation and understand that:

  • Global Standard deployments may route requests across multiple regions.
  • Microsoft recommends using Regional Standard deployments for data residency requirements.
  • Azure OpenAI does not use prompts, completions, embeddings, or training data to train foundation models.

However, we would like to obtain clarification on the following points.

1. If we deploy Azure AI Foundry in Germany West Central and use Regional Standard model deployments, are:

  • prompts,
  • responses,
  • inference processing,
  • temporary processing,
  • and embeddings

processed exclusively within Germany, or only within the Azure region "Germany West Central"?

2. Can Microsoft officially confirm that no inference requests leave Germany when using Regional Standard deployments in Germany West Central?

3. Which language models currently support Regional Standard deployments in Germany West Central?

4. Which embedding models currently support Regional Standard deployments in Germany West Central?

5. If the following services are all deployed in Germany West Central:

  • Azure AI Foundry
  • Azure AI Search (or Qdrant)
  • Azure Blob Storage
  • Azure Database for PostgreSQL
  • Azure Key Vault
  • Azure Monitor / Log Analytics

can Microsoft confirm that:

  • customer documents,
  • prompts,
  • responses,
  • embeddings,
  • vector indexes,
  • metadata,
  • and stored data

remain processed and stored exclusively within Germany?

We are looking for an official clarification because this information is required for customer compliance and GDPR documentation.

Thank you very much for your support.Hello Microsoft team,

We are currently evaluating Azure AI Foundry for an enterprise AI knowledge platform with strict GDPR and German data residency requirements.

We have reviewed the documentation and understand that:

  • Global Standard deployments may route requests across multiple regions.
  • Microsoft recommends using Regional Standard deployments for data residency requirements.
  • Azure OpenAI does not use prompts, completions, embeddings, or training data to train foundation models.

However, we would like to obtain clarification on the following points.

1. If we deploy Azure AI Foundry in Germany West Central and use Regional Standard model deployments, are:

  • prompts,
  • responses,
  • inference processing,
  • temporary processing,
  • and embeddings

processed exclusively within Germany, or only within the Azure region "Germany West Central"?

2. Can Microsoft officially confirm that no inference requests leave Germany when using Regional Standard deployments in Germany West Central?

3. Which language models currently support Regional Standard deployments in Germany West Central?

4. Which embedding models currently support Regional Standard deployments in Germany West Central?

5. If the following services are all deployed in Germany West Central:

  • Azure AI Foundry
  • Azure AI Search (or Qdrant)
  • Azure Blob Storage
  • Azure Database for PostgreSQL
  • Azure Key Vault
  • Azure Monitor / Log Analytics

can Microsoft confirm that:

  • customer documents,
  • prompts,
  • responses,
  • embeddings,
  • vector indexes,
  • metadata,
  • and stored data

remain processed and stored exclusively within Germany?

We are looking for an official clarification because this information is required for customer compliance and GDPR documentation.

Thank you very much for your support.

Azure OpenAI in Foundry-Modellen
Azure OpenAI in Foundry-Modellen

Ein Azure-Dienst, der Zugriff auf die GPT-3-Modelle von OpenAI ermöglicht und Unternehmensfunktionen bietet

0 Kommentare Keine Kommentare

Ihre Antwort

Antworten können von Fragestellenden als „Angenommen“ und von Moderierenden als „Empfohlen“ gekennzeichnet werden, wodurch Benutzende wissen, dass diese Antwort das Problem des Fragestellenden gelöst hat.