Fabric Data Agents - residency

[ptmk25]

I tested and ontology could only be created with the 2 tenant settings for processing and storing outside of our region being enabled. Assuming the same is true for data agents. I would like an official answer that this is the case. Unless you are in the US region, data sent to azure OpenAI can be stored outside you capacity's geographic region, compliance boundary, or national cloud instance.

Copilot admin settings - Microsoft Fabric | Microsoft Learn

Thank you.

[tayloramy · Accepted solution]

Hi @ptmk25, 

I had the exact same question about Fabric IQ and Data Agents last month, so I set up a call with Microsoft to get an official answer. 
This is the answer they gave me in writing after the call: 

Following up on our call this morning: cross‑region data movement happens only when you enable the cross‑region feature required for Copilot or Data Agent, specifically when the OpenAI model is not available in your target region (which is currently the case in Canada). In those situations, the referred data are only prompts, conversation history, and metadata (schema information) that are sent across regions. Your stored data in tables, databases, or OneLake never moves across regions unless a user explicitly includes that data in a prompt. Any data included in a prompt is explicitly & intentionally provided by the user and is not pulled automatically by the AI agent from your table, database, or OneLake. Refer to documentation or screenshot below. 

Data storage follows the same rules as data processing. The documentation we reviewed during the call covers only conversation history, not the data in your tables, databases, or OneLake. Conversation history is the portion sent across regions so that AI agents can maintain context to be able to provide accurate & contextual responses. As mentioned above, if a user explicitly provides data in a prompt, then the data will be included in the conversation history as well. Users can delete this history at any time by clearing the chat, otherwise it is automatically removed after 28 days.

[arabalca]

Hi @ptmk25 ,

this is officially confirmed in Microsoft's documentation. The situation with Fabric Data Agents is analogous to what you observed with the Ontology feature, and the docs make it very clear.

Official source: Configure Fabric data agent tenant settings – Microsoft Learn

The Data Agents tenant settings page describes three separate settings that must be enabled if your capacity is outside the EU data boundary and the US:

Processing outside region – data sent to Azure OpenAI can be processed outside your capacity's geographic region, compliance boundary, or national cloud instance.

Storage outside region – data can be stored outside those same boundaries.

Conversation history outside region – session history (retained for up to 28 days unless manually deleted) can also be stored outside the region.

The underlying reason lies in the Azure OpenAI architecture: the Data Agent is powered by Azure OpenAI LLMs, which are deployed to a select set of data centers. By enabling cross-geo processing, data may be processed in a region where Azure OpenAI Service is available, which could be outside the user's geographic region, compliance boundary, or national cloud instance.

In other words: if your Fabric capacity is in, say, West Europe, Spain, or Latin America, and you are not within the EU Data Boundary or the US, you need to enable all three settings to use Data Agents — and by doing so, you accept that both processing and storage (including conversation history) may leave your region.

This is consistent with what you describe regarding the Ontology feature. I understand that logically this makes sense: both features rely on the same Azure OpenAI backend, so they share the same regional availability constraints. It wouldn't be reasonable to expect different behavior between them.

If my comment helped solve your question,  it would be great if you could mark it as the accepted solution. It helps others with the same issue and it also motivates me to keep contributing.

Thanks a lot. I really appreciate it