Facebook has become the latest technology firm to admit listening to and analysing some audio recordings from users.
The social network said it had reviewed some small segments of audio to help improve the artificial intelligence of the transcription feature within its Messenger app.
The feature enables users to dictate a message with their voice before Facebook’s software transcribes it into text.
According to a report by Bloomberg, Facebook used contractors to carry out human review of some clips from the process.
But the social networking firm said it has now ended the practice.
“Much like Apple and Google, we paused human review of audio more than a week ago,” a Facebook spokeswoman said.
In recent months Amazon, Apple, Google and Microsoft have come under scrutiny after admitting they had listened to audio requests made to their respective virtual assistants.
The companies each said they used the practice to grade and analyse the quality of their voice and language recognition services, with the aim of improving their performance.
Some of the firms said they have since ended the practice, while Amazon said it will let users opt out of having their audio reviewed by humans.
Facebook said the audio used in its programme was de-identified and masked to protect user privacy, and that the company never listened to people’s microphones without device permission and explicit activation from someone.
The social network has previously been the subject of claims that it listens in to user audio in order to target them with advertising.
Facebook boss Mark Zuckerberg labelled the idea a “conspiracy theory” when he appeared in front of the US Congress last year, adding “we don’t do that”.
Last month, the social media company also agreed a £4 billion fine with the US Federal Trade Commission over privacy violations linked to the Cambridge Analytica data scandal.