Google workers will listen to audio snippets of people speaking to its digital voice assistant to help improve the product's quality - if users give the company permission to do so.
Alphabet Inc.'s Google paused all human review of assistant audio in July after a Dutch contractor leaked some clips to a journalist, who was then able to identify specific people on the recordings. On Monday, Google said it's bringing back human reviewers, but adding a new set of precautions to protect customers' privacy.
Google and other companies like Amazon.com Inc. use human transcription to check machine translators and make them smarter. The practice is widespread, but has made some users nervous that giant corporations are monitoring them. The companies maintain that audio snippets aren't linked to personally identifiable information.
Under the new Google policy, the company will tell users that their audio may be listened to if they opt in to a feature that also improves audio quality, Nino Tasca, a senior product manager on Google's Assistant team, said in a blog post. The company also is trying to more accurately recognize audio that was captured accidentally. Usually, the assistant only listens when a person says, "Hey, Google," but occasionally the computer might misinterpret something else as that "wake word" and begin listening when the customer isn't aware.
"We believe in putting you in control of your data, and we always work to keep it safe. We're committed to being transparent about how our settings work so you can decide what works best for you," the company said in a blog post.
Google also said users can view their past interactions with the voice assistant and delete any of them at any time.