Microsoft is updating the way we ask for permission to use your voice data to improve Microsoft products that use our speech recognition technology. As a result, in addition to your voice settings, you’ll also see some changes in how we manage your voice data.
Here's a brief overview of what this means:
With your permission, your voice clips—audio recordings of what you say when you use your voice to interact with Microsoft products and services—will occasionally be sampled and listened to for the purposes of improving our speech recognition technology and training our speech recognition models. Your voice clips can only be listened to by Microsoft employees or contractors, and will only be used to help improve our speech recognition technology.
Even if you don’t opt in to contributing your voice clips, you can still use our voice-enabled products and services.
We’re also changing the way we manage voice data, which will impact the voice data you see on the privacy dashboard. Voice clips will be de-identified as they are stored—they won’t be associated with your Microsoft account or any other Microsoft IDs that could tie them back to you. New voice data will no longer show up in your Microsoft account privacy dashboard.
To support these changes, new voice settings will roll out on a product-by-product basis. To make sure we apply these changes uniformly across all our products, we stopped logging any voice data for product improvements beginning on October 30, 2020. We won’t log voice data to sample and be listened to by people for product improvement going forward unless you agree to contribute voice clips specifically for that purpose.
These changes will not impact any voice data that Microsoft previously collected and has retained from its products or features prior to October 30, 2020.
To learn more about Microsoft and your privacy, please see the information in the Microsoft Privacy Statement.
What are voice clips?
Voice clips are audio recordings of what you say when you use your voice to interact with Microsoft products and services. When you use your voice to interact with our products, our speech recognition technology automatically generates the audio of what you say into words so that the Microsoft services can respond—for example, when you use Cortana to set a reminder, speech recognition technology is used to translate your command into text so that Cortana can set it for you. Our new opt-in voice settings allow you to give Microsoft permission to sample and listen to these voice clips to improve our speech recognition technology.
Why does Microsoft ask me to contribute my voice clips?
We use voice clips to help train our speech recognition technology to be better, more accurate, and more precise for you and everyone who speaks your language. For example, your everyday use of our voice-enabled products helps our speech recognition models to learn and better recognize complex and nuanced aspects of how people talk—like accents or regional dialects, and how sentences are structured in different languages. Sampling voice clips also helps us make our technology better at understanding speech in different acoustic settings—like when there’s a lot of ambient noise versus when things are quiet.
These improvements allow us to build better voice-enabled capabilities that benefit users across all our products and services.
What does Microsoft do to help protect my privacy?
Our automated software takes steps to de-identify your voice clips. This means we remove any association to your Microsoft account or any device identifiers that could be traced back to you. We also automatically remove long strings of numbers that could be phone numbers, social security numbers, or credit card numbers—and we remove anything that sounds like an email address.
Keep in mind that if you allow people other than yourself to interact with a voice-enabled product or service from Microsoft that you gave us permission to use, you’re responsible for ensuring that they understand and consent to Microsoft using their voice clips as described here. Also note that if other sounds are audible when a voice clip is recorded, those sounds might be recorded as well.
Who has access to my voice clips?
Microsoft employees and vendors working on behalf of Microsoft will be able to review your voice clips in order to build and improve our speech recognition technologies. Prior to employee or vendor review of voice data, we protect users’ privacy by taking steps to de-identify the data, requiring non-disclosure agreements with relevant vendors and their employees, and requiring that employees and vendors meet high privacy standards.
Where are my voice clips stored, and for how long?
After your voice clips are de-identified, voice clips are stored in a secure, encrypted server. We keep voice clips that have been contributed to be sampled and listened to for up to two years, during which time we may sample them. If we do sample a voice clip, we may keep it for longer than two years so that we can continue to train and improve our speech recognition models.
Which products do these changes affect?
These new voice settings will be made available on a product-by-product basis. Various Microsoft products will be updating their voice data controls and settings over the coming months, including Microsoft Translator, SwiftKey, Windows, Cortana, HoloLens, Mixed Reality, and Skype voice translation. In the interim, without your permission, your voice clips will not be sampled to be listened to for product improvement.
This change does not impact all Microsoft products that integrate speech recognition technology. For example, Xbox, while it has voice features, does not currently plan to sample and listen to customer voice clips for the purposes of product improvement and will not be updating its voice settings.
Further, Office apps and services, while they have voice features, do not currently sample and listen to customer voice clips.
These changes also do not affect other audio recordings stored by other services, such as Microsoft Teams meeting recordings.
Why does voice data no longer show up in the privacy dashboard on account.microsoft.com?
To make sure we apply these changes uniformly across all our products, we stopped logging any voice data for product improvements beginning on October 30, 2020.
Any voice clips you agree to contribute after October 30, 2020 will not be associated with your Microsoft account. Because of this, new voice data will no longer show up on your privacy dashboard. Microsoft will display voice data previously collected and associated with your Microsoft account on the privacy dashboard as long as we retain a copy.