Google moves tab in the controversies that began in early summer around how they allowed third parties to listen to our recordings in Google Assistant, a controversy that has also splashed other companies doing similar practices such as Apple or Amazon. It has done so by ensuring that it will reduce the number of audio recordings that are collected and that users will be better informed about how the data is treated.
They have explained it in a blog post, in which they have insisted that the assistant has audio recording options disabled by default. They have also promised to reduce the amount of audio stored by each user, promising to erase the vast majority of the recordings of users who have decided to contribute with them to “improving the service,” which is why they insist on what it stores.
“By default, we don’t store your audio recordings,” they have defended in their blog post. “To store your audio data, you can opt for Voice and Audio Activity (VAA) settings when you configure the wizard.” This is one of those options that almost invite you to activate by promising improvements as better recognition of your voice.
Therefore, this practice will not change, but what Google has said it will do is reduce the number of recordings stored automatically and without users having to do anything.
For all those who have activated the VAA option, we will soon begin to automatically erase the vast majority of the audio data associated with your account that has more than a few months – Google blog.
“This new policy will come to VAA later this year.”
The search engine company also says that it will change the texts of its configuration to make it clear that if you activate the VAA your audios can be heard by human reviewers to improve recognition technology.
We remember that the absence of this clarification by large companies as part of the controversy. In addition, if you are a user who already has the option activated, you will be asked to review the configuration to confirm that you want to continue with it.
Other changes promised by Google
In the blog post, Google has also ensured that its assistant will automatically erase the audio it picks up when it detects that it has been activated in error by some sound similar to the activation command. They ensure that they will focus on this area by improving the way to detect these unintended activations and exclude recordings from the human analysis process.
Google ensures that “soon” a method will also be added to configure the sensitivity of the wizard to reduce accidental activations or, in case you prefer to maximize sensitivity, make sure you can activate it in noisy environments.
There will also be changed in the transcription process involving human persons who access the audio. According to the search engine company, the samples sent to them are never associated with a user account. They promise that linguistic experts only access 0.2% of the recordings obtained from users, and always only when they activate the VAA. New security layers will be added to this process.
All in all, what Google has done in its statement is to ensure that the process is transparent since users always have to activate that option of Voice and Audio Activity. But of course, we all know that sometimes this activation becomes confusing, so it is good for Google to say that it will make it clear that if you activate it your audios will be reviewed by human employees.
They also promise that you will soon be able to adjust the sensitivity of the wizard to avoid activations by mistake and that it will erase these erroneous activations when detected. But of course, at the moment everything is in promises, and we will have to wait to see how they implement everything. They have promised transparency, so we should see how they keep their word.