Google still using humans to transcribe Google Assistant enquiries, sparking privacy concerns

Google has been no stranger to privacy concerns in recent years and with their always listening smart speakers it seems that there is a very fine line between helpful and creepy. Now an investigation by VRT NWS has uncovered some interesting practices as Google attempts to improve Assistant in non-English-speaking countries.

The report is claiming that Google employees are not just listening to a small percentage of enquiries (0.2%) but is then using subcontractors to transcribe them in a bid to improve their local voice assistant technology. A whistleblower provided VRT NWS with more than 1000 audio recordings which they had been asked to listen to and transcribe. Some of these recordings included sensitive information such as home addresses, pornography and medical questions — most of which you would expect people want to remain private.

VRT NWS actually reached out to some of the recorded folks who confirmed that it was their voices and how shocked they were to discover their recording had been made available to Google subcontractors.

The idea behind the recordings is to make the speech recognition more accurate. The subcontractor logs into the secure Google site and compares the automatically generated script to the recording for accuracy and completeness. The recordings are disconnected from the users’ information with the user name deleted and replaced with an anonymous serial number but the person can be relatively easily identified by listening to the recordings provided.

Of the recordings VRT NWS listened to 153 were recorded without the OK Google hotword being spoken which is striking. Anything that sounds remotely like the hotword triggers the beginning of recording. This is how sexual relations, conversations between parents and children, fights and professional phone calls with sensitive information were also recorded.

There were instances of domestic violence being recorded which raises the ethical question of what is Google’s responsibility here. There are no guidelines for the subbies on what they should do when this occurred but you can bet they suffered over it.

One of our three independent sources says he had to describe a recording where he heard a woman who was in definite distress. What are employees supposed to do with such information? We are told that there are no clear guidelines regarding such cases. It is, however, an important ethical matter.

In response Google admits to working with ‘language experts worldwide’ to help improve their speech recognition. Google Belgium said that the recordings are not linked to any personal or identifiable information and that:

This work is of crucial importance to develop technologies sustaining products such as the Google Assistant….. [They listen to] about 0.2 percent of all audio fragments

If you want to see the entire investigation check out the YouTube clip below:

Google is already under fire in Europe for many things and this just adds fuel to the fire. It will be interesting to see where this investigation leads for Google — it isn’t the first time they’ve been found to be doing this and they are not the first company to do so but with the spotlight already on them the privacy police in Europe may have a question or two to ask them.