After becoming a mostly touch-based society thanks to smartphones and tablets, tech giants are trying to turn us back towards using our literal voices to drive our interactions with technology. AI assistants, smart speakers, and speech recognition have become staples of many products and services but, unlike touch, voice data also carries more privacy implications. After being criticized and scrutinized over such privacy invasions like its peers, Microsoft is now announcing a bigger change to its voice data collection practices that it says gives users more control.
The scandal over voice clips privacy exploded in 2019 when Amazon was revealed to be exposing audio clips recorded by Alexa to third-parties, mostly contractors, without users knowing about it. It wasn’t the only one though and, soon enough, Google and Microsoft were also dragged into the spotlight. Microsoft since then promised to change the way it handles speech data and its latest announcement makes what is perhaps the biggest change.
According to Redmond, users will soon have to opt into voice data collection which will then be used for the usual improvement of services. This implies that the default will be to disable data collection, at least once the update to apps and services have been rolled out. Microsoft says that users who do opt out of it will still be able to use those Microsoft products as normal.
There is, however, one catch to that change. It also says that Microsoft may still continue to collect information related to voice activity even if it doesn’t collect the voice clip itself. That includes, for example, the translations or transcripts that are produced when users interact with its speech recognition AI. Either way, Microsoft also promises that they will be scrubbed of any personally identifiable information before they are processed.
The company stopped storing voice clips back in October and this update is pretty much its announcement that it will resume the process but is making it optional. It says that users who opt into the collection could have their audio clips retained on servers for up to two years. However, Microsoft makes the rather confusing qualification that data could be retained far beyond that if the audio clip is sampled for transcription by employees or contractors.