SAN FRANCISCO – Google admitted on Thursday that human language experts hired by the firm listen to approximately 0.2 percent of the conversations that users have with its virtual assistant.
The widespread assumption – which is often promoted by companies using this technology such as Amazon, Samsung, Apple and Google – was that conversations between a user and their virtual assistant are entirely private, and that the interaction occurs exclusively through artificial intelligence; that is, the only ones who listen to the users are robots.
However, Google’s admission on Thursday that 0.2 percent of these conversations are heard by humans in order to allegedly improve the quality of service and develop speech technology for more languages sheds light on a practice that companies generally avoid advertising, although it is known within the industry to be more or less commonplace.
“As part of our work to develop speech technology for more languages, we partner with language experts around the world who understand the nuances and accents of a specific language,” Google said in a statement.
Audio recordings in the Dutch language were provided to Belgian public broadcaster VRT NWS by one of the experts hired by Google in that country to listen to segments of the conversations.
The company has already announced it would take action against the leak, which it called a violation of its data security policies.
But Google admitted to employing experts around the world whose role is to “review and transcribe a small set of queries to help us better understand those languages.”
The company, which is based in Mountain View, California, claimed that “a wide range of safeguards” were used to protect user privacy and that only 0.2 percent of all audio was reviewed.
“Audio snippets are not associated with user accounts as part of the review process, and reviewers are directed not to transcribe background conversations or other noises, and only to transcribe snippets that are directed to Google,” it added.
However, the Belgian broadcaster was able to identify postal addresses and other sensitive information in the recordings, which allowed them to get in touch with those whose voice was recorded and confirm that it was really them.