Facebook has been paying hundreds of independent contractors to listen and transcribe audio content collected from the users of its Messenger app without the users’ direct consent.
Bloomberg was the first to report the story on Tuesday after it obtained information from workers tasked with transcribing the audio snippets. The outsourced workers say they were not given any details on how and where the audio material had been recorded. They were only assigned with transcribing the sometimes vulgar content without receiving further instructions.
The transcribers who came forward to Bloomberg after realizing the origin of the material they had been working on asked to remain anonymous for fear of losing their jobs.
Facebook confirmed it had been using transcription services to improve the quality of its artificial intelligence software. It said contractors had been reviewing the transcription work carried out by the AI, only using the data gathered from users who chose the option to have their Messenger voice chats transcribed.
Other tech giants developing their own AI don’t shy away from such practices either. Amazon, Apple, and Google have all been under public scrutiny for collecting audio clips from users, transcribing them using AI technology, and later subjecting those clips to human review.
In April, the news of Amazon employing thousands of workers in various countries to listen to Alexa voice requests broke. The company justified the act using the same explanation Facebook gave now – the human review was necessary for improving the software.
Soon after, Google was put under the spotlight when some of its Dutch language audio snippets were leaked. Belgian broadcaster VRT NWS got hold of more than 1,000 recordings collected by the search engine company through its Google voice assistant.
Some of the leaked audio clips contained personal information that could be used to identify the person speaking. Google said it has put this practice on pause worldwide while it investigates the Dutch leaks.
Apple also adopted the practice of subjecting sensitive user information to human review without the users’ knowledge, to develop its digital assistant Siri. The iPhone maker said it has stopped the practice for now but plans to reintroduce it after asking for explicit permission from users.
Critics of this, apparently, widespread practice warn that once human beings process the information, there’s no way of ensuring it stays contained.
“We feel we have some control over machines,” Jamie Winterton, director of strategy at Arizona State University’s Global Security Initiative told ABC News. “You have no control over humans that way. There’s no way once a human knows something to drag that piece of data to the recycling bin.”
The Irish Data Protection Commission, the regulatory authority that oversees Facebook in Europe, said it was “seeking detailed information from Facebook on the processing in question.”
In the light of yet another user privacy violation, Facebook shares dropped 2.2% at 10:02 a.m. in New York trading.