The privacy fears regarding Facebook and its apps have once again come to the forefront after a Bloomberg report highlighted that the company paid outside contractors to transcribe audio clips from users of its Messenger services. Facebook was getting audio clips transcribed, which it got under the Voice to Text feature on the Messenger app.
This report comes after Facebook recently settled a privacy lawsuit with the US Federal Trade Commission (FTC) and agreed to pay a $5 billion fine. Facebook is not the only company to face criticism over how it handles users’ audio content. Apple, Amazon and Google have also faced heat over employing human contractors to listen to audio clippings of users, who were relying on the voice-assistants from these companies.
In Facebook’s case, Bloomberg has also revealed that the employees who were transcribing the audio clips did not know where it was recorded or the manner in which it was obtained. The report adds that many of these employees, who wished to remain anonymous for fear of losing their jobs, said that some of the content was vulgar, included user conversations, and they found it disturbing.
On its part, Facebook admitted that it was transcribing users’ audio and said it was suspending the program. “Much like Apple and Google, we paused human review of audio more than a week ago,” Facebook said in a statement, according to the report.
What is the Facebook Voice to Text feature?
Facebook was transcribing data from audio clips which it received in the ‘Voice to Text’ option for audio messages sent on Messenger. Facebook did not roll out this feature to all countries and locations. For instance in India, we could not see the option for auto-transcription for voice messages.
Facebook Voice to Text feature in Messenger would display a text for the voice clips sent or received. Users would have to turn on the option when they sent a voice message to their friend via the Messenger app.
Facebook’s page for Voice to Text explains that even if the user turns it off, but if someone else in the chat still has this feature turned on, that person will still see transcriptions. However, the page does not mention that third-parties will be analysing these audio clips for transcriptions and double checking how the artificial intelligence software fared.
A user had to send a voice clip in a chat and after that they could tap the Transcribe option, which would turn on Voice to Text. The audio message would then show the transcribed text at the bottom of the audio recording.
Did Facebook inform users that it would transcribe these audio clips?
When users turned on the transcription feature, Facebook did not mention or prompt users that data from this could be shared with potential third-party companies for further analysis. It only asked them for a yes, or no, with regard to turning on the Voice to Text feature.
Which company was reviewing these Facebook audio clips?
According to Bloomberg, one of the companies reviewing the user conversations is TaskUs Inc. Facebook is their largest client, but employees can only use the code name to refer to the social media company, which is “Prism.”
While TaskUs also reviews content for Facebook, it was also looking at the transcription work. The company confirmed that Facebook asked them to pause this particular task over a week ago.