Where is data stored when using the fast transcription API?

Question

Dear Microsoft,

for a project, we are using the fast transcription API from Azure as a component to transcribe text from audio. We are working with sensitive data and would like to be sure to have control or at least insight into where our data is stored and processed. I had trouble finding a conclusive answer and I was hoping you could give me some insight into those uncertainties.

I am transcribing using the locale 'nl-NL' and using an AI Foundry resource in West Europe.

Thanks in advance,

C

Answer 1

Hi Lenting, Coen

For real-time speech-to-text services, Microsoft does not retain or store the data provided by customers.

In the case of batch transcription, customers can specify their own storage locations for the audio input. If the generated transcription text is stored in Microsoft storage, customers have the option to delete the data either through a deletion API or by setting a time-to-live parameter for automatic deletion.

For sensitive data, it is crucial to ensure that you specify your storage locations appropriately and utilize the deletion options available to maintain control over your data.

You shall be responsible for securing and maintaining the equipment and infrastructure required to operate speech containers located on your premises, such as your edge device and network.

To learn more about Microsoft's privacy and security commitments visit the Microsoft Trust Center.

You can secure your storage and speech resource with virtual network and modify default retention policy on storage side.

You can additionally set ACL (Access control less) for certain folders for certain user group.

Reference used -

https://learn.microsoft.com/en-us/azure/ai-foundry/responsible-ai/speech-service/speech-to-text/data-privacy-security?source=recommendations#how-does-speech-to-text-process-data

Upload audio file to Blob storage

Data protection on Storage side (Soft delete enabling, Data retention etc)

Life cycle management on Storage side

Virtual network for Azure Speech and other services

Hope it helps

Thank you.