Google Home smart speakers, the company dryly warns in a note buried deep on a support page, can “incorrectly” record their users even when they haven’t first said an activating wake phrase like “hey, Google.” It just so happens that, at least for a brief period of time this summer, those microphone-enabled devices were doing exactly that.
The company admitted Monday, following a report by Protocol, that it had updated an unspecified number of Google Assistant-enabled devices to respond to auditory cues beyond the user-specified wake phrase. Google told Protocol this was a mistake that was quickly fixed, but did not appear to address the larger privacy concerns that such a mistake signifies. After all, how are users supposed to trust a live microphone in their home if someone can remotely update it to be even more invasive without their knowledge?
This is not the first time that Google Assistant-enabled products have acted contrary to whatever stated privacy protections they had. In July 2019, we learned that not only were Google Homes recording users, unprompted, but that some of those recordings were then sent to real humans to listen to. Amazon’s Alexa-enabled devices were also doing this, as was Facebook’s Portal. (Following a backlash, Google announced users could opt in to something called “Voice & Audio Activity,” which permits people to listen to your Google Assistant recordings.)
Understandably, then, any change to a Google device that makes it more prone to activate and possibly record users is worth scrutinizing — a difficult task when the update is made behind closed doors.
The development was flagged by a Redditor, who noticed their Google Assistant-enabled device was listening for a smoke alarm, and, after hearing it, had a pre-programed set of actions ready to go.
“Burned something in the kitchen and the cheap $10 smoke detector went off,” wrote Reddit user Brazedowl, who noticed the change and shared a screenshot of the notification. “Then I got a notification on my phone that Google heard the smoke detector going off.”
While it’s not 100 percent clear, it’s possible someone at Google jumped the gun ahead of a new partnership with security company ADT. Enabling Google devices to pick up on smoke alarms sounds like just the thing such a partnership would herald. And, as Protocol mentions, Google Nest speakers do actually listen for “critical sounds” like alarms or glass breaking, as part of the Nest Aware subscription service.
“A recent software update enabled these alerts on some of our speakers that didn’t have a [Nest Aware] subscription,” a spokesperson told Protocol, “but we’ve since rolled that back.”
We reached out to Google with a host of questions about this supposedly mistaken update. Specifically, when did this update go into effect, and how many devices were altered? What user data, if any, was collected as a result of this and what was done with it?
Unfortunately, we received no immediate response. In a world where approximately 24 percent of American adults have smart speakers, that leaves a lot of people with unanswered questions.
It’s important, after all, to know if Google intends to make similar unannounced changes to its products in the future. Because even if a company spokesperson swears this was a one-off mistake, knowing what Google does with customer information gathered through accidental invasive recordings is invaluable.
After all, when it comes to technology pushing further and further into the home, mistakes are bound to happen again. And again.