In response to a report by Belgian public broadcaster VRT NWS showing that Google lets human workers listen to audio captured by Google Assistant software, the tech giant acknowledges that its language experts review 0.2 percent of all audio snippets recorded by its virtual assistant. Google justifies this practicing by arguing that “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.”
But the VRT NWS report shows that these safeguards aren’t sufficient to protect user privacy. In fact, the broadcaster obtained 1,000 audio snippets leaked by a reviewer. Of these, 150 seem to have been recorded by the assistant without user consent, as they lacked an explicit command by the user (such as “OK Google.”)
Google admitted that one of their “language reviewers has violated our data security policies by leaking confidential Dutch audio data,” adding that the company is “conducting a full review of our safeguards in this space to prevent misconduct like this from happening again.”
Read more: Google defends letting human workers listen to Assistant voice conversations