Does Translator text translation via a resource-specific endpoint stay within US datacenters?

Luisina Peralta 0 Reputation points
2026-10-23T23:31:44.6+00:00

We use Azure AI Translator for text translation through a third-party CAT tool plugin. Our Translator resource is in West US, and the plugin only supports the resource-specific endpoint (<resource-name>.cognitiveservices.azure.com), not api-nam.

The region support documentation confirms that api-nam processes requests in East US 2 or West US 2, but it doesn’t state where requests to a resource-specific endpoint are processed for US resources.

Are text translation requests sent to the resource-specific endpoint of a US-region resource processed only in US datacenters, including during a datacenter failure? We need this for a HIPAA data residency review.

Azure Translator in Foundry Tools
0 comments

1 answer

Sort by: Most helpful
Most helpful Newest Oldest
  1. Hassan Nazeer 0 Reputation points
    2026-10-24T04:09:08.92+00:00

    Short answer: for a US resource, the public docs don't say. The current Region support for Azure Translator page only gives a processing location for the resource-specific endpoint in one case, Switzerland. For a West US resource on <resource-name>.cognitiveservices.azure.com, no processing location is written down, so I wouldn't put "US only" in a HIPAA review based on the docs alone.

    What the page does document for standard NMT text translation:

    Endpoint Request processing location (as documented)
    Global api.cognitive.microsofttranslator.com Closest available datacenter. During a datacenter failure the request might be processed outside the originating geography
    Americas api-nam.cognitive.microsofttranslator.com East US 2 or West US 2
    Switzerland, resource-specific endpoint Switzerland North or Switzerland West

    The same page also says that "a resource location and a request processing location aren't always the same", and that Translator doesn't persist customer data submitted for text translation. That second point usually helps a residency review, but it's a separate question from where processing happens.

    What I'd do:

    1. Get it in writing. Open a support request, or go through your Microsoft account team, and ask exactly this: "For an Azure Translator resource in West US called through its resource-specific endpoint, which datacenters process text translation requests, including during failover?" An auditor will want that answer, not a forum post.
    2. If the CAT plugin can't use api-nam, one option is a small proxy that you host in a US region. The plugin calls the proxy and the proxy calls api-nam.cognitive.microsofttranslator.com. That gets you onto the endpoint with a documented US processing location. Keep in mind the proxy then sees the text and becomes part of your HIPAA scope.
    3. Check the HIPAA side separately. Whether Translator is covered under your BAA is answered by Microsoft's compliance documentation and your agreement, not by the region page.

    Limits of this answer: it covers standard NMT text translation only. LLM-based translation in Foundry follows the model deployment type (global, data zone or regional) instead, and Document Translation has its own table, where the resource region decides processing. I haven't traced the network path myself.

    Source: Region support for Azure Translator

    Was this answer helpful?

    0 comments

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.