Google “spies” on your voice commands to Google’s Assistant. This was discovered when Belgian broadcaster VRT NWS made an in-depth investigative report, detailing how Google employs staff globally to transcribe voice commands that people issue to the AI assistant. 

Some weeks ago, Amazon was found guilty of spying on users of its AI digital assistant, Alexa. Now, according to a report from Dutch-language news channel VRT NWS, Google was found to be hiring contractors to listen to and transcribe voice commands that users give to their Google Assistant. Google “spied” on commands made via smart speakers and smartphones.

The “Spying”

The news report says that one contractor supplied over 1,000 recordings for the investigative piece, with many of the conversations containing users’ personal details. About 153 of the recordings were “accidentally-recorded” conversations, since these weren’t started with the “Ok Google” or “Hey Google” command. The news report claims that although Google has informed its customers in its privacy policy that they’ve agreed to let Google record their commands, Google fails to mention that other people will be listening to the recordings.

In the video report, one transcriber who requested anonymity, “Peter”, shared that he was hired to transcribe Dutch recordings from Google Assistant users in Belgium and Holland. “Peter” reveals that he and his colleagues log onto a Google platform where they can listen to audio recordings, which they transcribe in perfect detail. Apart from writing down the conversations verbatim, transcribers must also identify whether it’s a man’s, woman’s or child’s voice. “Peter” and other transcribers handle about 1,000 recordings a week. “Peter” and his co-workers are sent “anonymous” user audio files, but none of the voices are altered or disguised, and in some cases the recordings include information that can be traced back and easily identify the users.

You can watch VRT NWS’ English-subtitled report here:

Privacy issues with AI

Google isn’t the only company nor is the field of AI assistants struggling with privacy issues. Privacy is one of the biggest obstacles that the AI industry faces today, and the results aren’t exactly encouraging. In the case of healthcare in America, a poll recently found that only 1 in 5 adults would trust AI-generated advice. As for AI in general, another poll revealed that only 25% of respondents said they found AI’s decisions “trustworthy” and a mere 12% claim they’ve interacted with an “empathic” machine.  It doesn’t help the cause of AI that some years back, even Smart TVs were found to have been listening in on people’s private conversations. Maybe with the help of large media companies, an information drive like this can better inform consumers and give a more human face to AI.

“Shhh it could be listening” (

Google Speaks Up

In a blog post, Google acknowledged that “one of these language reviewers has violated our data security policies by leaking confidential Dutch audio data. Our Security and Privacy Response teams have been activated on this issue, are investigating, and we will take action. We are conducting a full review of our safeguards in this space to prevent misconduct like this from happening again.”

The blog post goes further to clarify that their “language experts only review around 0.2% of all audio snippets” and that “reviewers are directed not to transcribe background conversations or other noises, and only to transcribe snippets that are directed to Google.” The blog post notes that Google Assistant may have committed a “false accept” of some noise or words that the smart speaker or phone’s software mistook for the activation words like “Ok Google”. Google claims safeguards are there to prevent these “false accepts.” Finally, as an added safety measure, the blog post suggests a way to review or change user device settings and delete all the activity recorded by devices on their accounts.

What this means for marketers

As with past cases involving possible privacy breaches by Samsung Smart TVs and Amazon Alexa, Google Assistant is the latest probable “offender”. For marketers, this case shows just how sensitive ordinary consumers are and can be with the handling of their data and their privacy. Cases like these also serve as basis for ordinary consumers’ prevailing distrust of AI. Apart from an information drive, marketers should advocate for continued transparency and the ethical use of AI tech must remain top priority.

To get the full details on this story, read it here:

Is this a serious privacy breach on Google’s part, or simply one media outlet’s piece of sensationalized news? Would you still use AI assistants like Google Assistant knowing “someone” somewhere could be listening? Let us know in the comments!

Leave a Reply