MAI-Transcribe-2 via fast transcription: Preview or GA under the DPA, and does data stay in the region?

Patryk Wikłacz 0 Reputation points
2026-11-07T16:16:49.3633333+00:00

I run a desktop dictation app for Windows. A user holds a key and speaks, and the app sends the recording to our API. Our API sends it to Azure Speech and pastes the resulting text back into the user's application. Users can dictate anything, so the audio and text may contain personal data, sometimes sensitive.

We plan to use MAI-Transcribe-2 in production, through Azure Speech fast transcription (enhancedMode), on a resource in northeurope. So far we have only run test calls.

  1. Is MAI-Transcribe-2 through this API generally available, or a Preview under the Microsoft Products and Services Data Protection Addendum? The MAI-Transcribe page no longer mentions preview, but the regions page still marks mai-transcribe-2 as preview for Voice Live.
  2. If it is a Preview, does it allow Personal Data processing, so that the DPA's personal data and GDPR terms apply?
  3. Does the Location of Customer Data commitment apply, so that audio and transcripts are processed only in the resource region?
  4. Is audio or text stored after the request completes? The Speech to text privacy page says fast transcription keeps no data, but does not mention MAI-Transcribe.

Please point me to the document that states each answer.

Azure Speech in Foundry Tools
0 comments

1 answer

Sort by: Most helpful
Most helpful Newest Oldest
  1. Gulnaz Mushtaq 490 Reputation points MVP
    2026-11-07T19:43:21.6433333+00:00

    For Azure Speech fast transcription (enhancedMode) in northeurope:

    The MAI-Transcribe documentation  no longer labels fast transcription Preview. The Preview label on the regions page  refers to Voice Live. Get Microsoft to confirm GA status contractually.

    If GA, normal DPA Personal Data and GDPR provisions apply. If Preview, Personal Data processing is permitted only where Microsoft explicitly allows it. Microsoft's regions documentation  says Azure Speech processes data in the resource region. However, the DPA's Location of Customer Data commitment has exceptions for eligible Previews.

    Speech data privacy documentation  says fast transcription does not retain or store customer audio or text after processing. It does not explicitly name MAI-Transcribe-2.

    Before production, obtain written Microsoft confirmation of three things:

    • MAI-Transcribe-2 enhanced-mode fast transcription is GA.
    • Personal Data, including sensitive data, is permitted under the DPA.
    • Audio and transcripts remain in North Europe, with no post-request retention.

    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.