Facebook’s help center simply mentioned in a note that the Voice-to-Text feature uses machine learning; there is no mention of the fact that actual people may be listening to the voice recordings.
Now, the lead privacy regulator of Europe is seeking to know the in-depth operation of the voice-to-text feature in the Messenger and how it complies with the EU privacy laws. If human transcribers were used by the tech giant, then it might imply that actual humans are listening to all the recordings of people sent through Messenger.
“Further to our ongoing engagement with Google, Apple, and Microsoft in relation to the processing of personal data in the context of the manual transcription of audio recordings, we are now seeking detailed information from Facebook on the processing in question and how Facebook believes that such processing of data is compliant with their GDPR obligations,” said a spokesperson for Irish Data Protection Commission, as reported by TechCrunch.
The AI voice assistants offered by Google, Microsoft, Amazon, and Apple have attracted the attention of the EU regulators over the last few weeks. After yesterday’s report, Facebook said that it had already suspended human transcriptions earlier this month. The company further added that all the audio clips sent to its contractors were de-identified and were only collected for improving the transcription capability of the AI.
Apple and Google too have halted doing manual reviews of audio clips for their assistants. Amazon simply gave an option to out of manual reviews in Alexa’s settings.
The fact that human transcribers are used to translate voice-to-text under the guise of ‘machine learning’ is a troublesome revelation. Nearly a decade ago, a UK based startup known as Spinvox claimed to use advanced AI to convert voice-to-text. It was later reported that this startup was heavily reliant on call centers in third-world countries where actual humans went through the audio clips to transcribe them.