Microsoft contractors are listening to personal conversations from Skype’s translation service, as well as Cortana queries.
According to internal documents, screenshots and audio recordings that a contractor provided to Motherboard, humans are analyzing Skype call audio that goes through the program’s translation feature. The audio clips contained conversations such as people talking intimately to loved ones, discussing personal issues like weight loss and relationship problems, and more sensitive and private data.
The audio clips are usually short, reportedly about five to 10 seconds long. However, the contractor said recordings could be longer.
Skype’s ‘Translator’ service launched in 2015 and brought near real-time audio translations to phone and video calls. It relies primarily on artificial intelligence (AI) to perform the translations, but humans facilitate some of the work to improve the algorithms and ultimately the translation.
A ‘FAQ’ for Skype Translator notes that it “collects and uses your conversations to help improve Microsoft products and services. To help the translation and speech recognition technology learn and grow, sentences and automatic transcripts are analyzed, and any corrections are entered into our system, to build more performant services.”
Elsewhere, the FAQ says: “To help the technology learn and grow, we verify the automatic translations and feed any corrections back into the system, to build more performant services.”
A Microsoft spokesperson told Motherboard that the company “collects voice data to provide and improve voice-enabled services like search, voice commands, dictation or translation services.”
“We strive to be transparent about our collection and use of voice data to ensure customers can make informed choices about when and how their voice data is used. Microsoft gets customers’ permission before collecting and using their voice data,” the spokesperson said.
They also said that the company has several procedures in place to protect users’ privacy before sharing data with contractors. This includes removing identifying information from data, requiring non-disclosure agreements from contractors and employees, and also requiring contractors to meet the high privacy standards set out in European law.
However, the Microsoft contractor who shared the files with Motherboard said: “The fact that I can even share some of this with you shows how lax things are in terms of protecting user data.”
How Skype Translator shares data with contractors
When contractors receive audio to transcribe, it comes with translations generated by Skype. Contractors then need to select the most accurate translation or provide their own. The audio is treated as confidential Microsoft information.
The data is only available through a secure online portal and doesn’t include identifying information like user or device identification numbers.
However, the employees that view the information are, at least in part, work-from-home contractors. Motherboard found online job listings for Microsoft’s contractors that said employees could work from home.
The contractor who came forward had heard conversations that were clearly phone sex, heard people enter full addresses into Cortana and seen Cortana queries for porn.
“I generally feel like that while we do not have access to user identifiable information, that if Microsoft users were aware that random people sitting at home in their pyjamas who could be joking online with friends about the stuff they just heard that they wouldn’t like that,” the contractor told Motherboard.
Microsoft is the latest tech company embroiled in the human revision controversy. After a contractor revealed that Apple contractors heard people having sex, drug deals and other sensitive content in recordings, the Cupertino-based company globally paused human reviews of Siri recordings. Google did the same with its Assistant platform, but only in Europe after German pressure about the review process. Amazon simply adjusted the language of its Alexa policies and made it easier to opt-out of human reviews.