Newsroom News Breaking Comics Tags RSS
News hardware & software 16 August 2019, 14:00

author: Conrad Hazi

Microsoft Shared Skype and Cortana Recordings With Third Parties

Microsoft has recently updated its privacy policy and clarified how Cortana and Skype recordings are used. As it turns out, the company from Redmond also uses third parties to assess the quality of the speech recognition mechanism and provides them with fragments of recordings.

Hakuna matata! You wouldn't be asking your assistants anything obscene, now would you?

Earlier this month, Apple suspended the rating program of its digital assistans Siri due to concerns about privacy breaches. The company was caught sending voice recordings of unconscious users to its contract staff to assess how Siri handles speech recognition. It turned out, however, that employees also received fragments of private conversations, which caused a lot of commotion among users. Meanwhile, it seems that Microsoft is using similar methods in the development of Cortana and Skype.

Is anybody even surprized...?

As Motherboard notes, Microsoft's new privacy policy, which has recently been published, reports that ithe company is using people to evaluate the quality of Cortana's hints and how it deals with speech recognition. Therefore, fragments of voice recordings must be presented to third parties. Looking at the number of fragments of private content listed by the members of the Siri testing group, we can safely suspect that the situation has been similar for a long time both for Skype and for Windows 10's Cortana application.

Microsoft strives to be transparent about the issue and "takes steps to anonymise" sound recordings, which are used to improve the speech recognition mechanism and the translator. In addition, the Redmond giant stresses that it is taking all necessary measures to ensure 'high privacy standards set by European law'. Wonder if that's really the case...

So far, it has been known that Amazon, Google, Facebook, Apple and Microsoft have used a combination of manual and automatic methods to improve speech recognition mechanisms and text suggestions.