Google Responds to Device 'Eavesdropping' Claims

Google is making several changes to the way it handles audio collected from phones and smart gadgets such as the Google Home speaker. It follows revelations that humans listened to recordings collected by several tech giants.

Amazon, Apple and Google were all hit by claims that they'd not done enough to make clear that staff or contractors listened to recordings to check the accuracy of automatic speech recognition. In several cases, human engineers heard sensitive material which appeared to have been collected unintentionally, rather than the user meaning to make a spoken query or request.

The Google changes are to what it calls the Voice & Audio Activity setting. That's a program that automatically analyzes recordings to get better at recognizing a particular user's voice. Some of those recordings are then reviewed by human staff.

Users Must Opt In Again

The program was already opt-in, but Google is now going to effectively reset every account to its original opted-out status. Users can opt-in again in the settings menu of the relevant apps, but will see an express warning about the possibility of having their audio listened to by human engineers in order to improve the service.

The BBC notes that Apple switched to an opt-in system earlier this year, while Amazon and Facebook (which has lesser-known home gadgets with speech recognition) both have similar features switched on by default and users have to actively opt out. (Source: bbc.co.uk)

Google says it is also increasing privacy protections in the human review process. It notes staff currently listen to around 0.2 percent of all recordings; they only listen to recordings where the user is opted-in to the Voice & Audio Activity program; and, staff can't see details of the user who the audio clip comes from.

Old Recordings Will Be Deleted

Another change is to reduce the amount of audio that is stored on Google's servers. Google will "soon" delete anything that's more than "a few months" old. It says it will improve the technology to better recognize and delete recordings made accidentally.

It will also let users adjust the sensitivity with which devices listen out for a trigger phrase such as "OK Google." That could reduce the risks of unintended recordings, but increase the possibility that the device might not hear an intentional request if there's a lot of background noise. (Source: blog.google)

What's Your Opinion?

Do the changes sound reasonable to you? Do you trust tech companies really want to protect user privacy? Do you think the privacy risks are significantly limiting the audience for such devices?