Google admits its new smart speaker was eavesdropping on users (2017) | Hacker News Reader

Google admits its new smart speaker was eavesdropping on users (2017)

money.cnn.com

25 pointsotp1249 comments

Hmmm yet again, I am inclined to not believe this. How is it that these bugs always end up invading our privacy, and not the other way round? ("Google/Facebook admits a bug in their system reduced the data they could collect"). Because those bugs are true bugs as far as these companies are concerned, and they're squashed ASAP.

This is like when Facebook had an oopsie about leaving the microphone on, "oh no sorry it was just a bug." Right, when you abused an iOS API to leave your app running all the time, a side effect was that you also recorded all the conversations on the phone. These companies are becoming a cancer on our privacy.

>How is it that these bugs always end up invading our privacy, and not the other way round? ("Google/Facebook admits a bug in their system reduced the data they could collect")

Because you and I both know that's not newsworthy and wouldn't make it out from inside the company where it happened?

> My Google Home Mini was inadvertently spying on me 24/7 due to a hardware flaw

Are we really supposed to believe this?

Charitably, Google does not take privacy seriously, because they release products that have gaping flaws they neglect to test or correct. It would have been trivial for a test engineer to see a unit is collecting data on conversations without the wake word.

Or uncharitably, Google does not take privacy seriously, because they are releasing spyware, and just lie when they get caught.

Edit: In this case, maybe some leeway can be given because they were test devices, but it still makes me wary of having these type of things in my home.

This case seemed a pretty cut and dry bug in pre-release software. They actually launched the device with the at-fault feature disabled while they fixed the bug, even. I get that Google is difficult to trust when it comes to privacy issues, but in this case it seemed like a bug much in the way a few weeks ago an Alexa picked up keywords that it understood as "Call this person" and the people on the other end of the call were able to hear a private conversation.

> This case seemed a pretty cut and dry bug in pre-release software.

In this case it was actually hardware—the button used to manually trigger the device (instead of saying “okay google”). Seems weird that Google managed to screw up a button, but i doubt it was malicious.

Originally, there was going to a be a feature where you could touch the top of the Mini to activate the assistant without saying the wake word.

The reporter put their test device in a bathroom. The sensor on the top reported false positive activations in this environment.

The touch-to-activate feature was disabled after the reporter's feedback.

The device flashes lights when it is recording and says it doesn't know how to answer the question when it gets input it doesn't understand, so the alternative hypothesis that it was spyware doesn't make any sense. The first hypothesis that it was bad testing is the only one that makes sense, but it was corrected before release, so the rest of your analysis of that hypothesis does not apply.

Well.. yes? Unless you think Google maliciously meant to record the owners of the test devices?

Hardware is hard, and Google's screwed it up more than once.

By this point, is the the only difference between breaches by Google (and its ilk) and Equifax is that in the case of the former we can opt out using it?

Reply on news.ycombinator.com