>It's just a statement of how all voice recognition currently works.

Amazon has said on numerous occasions that no data transfer occurs without a trigger word hitting the mic -- a feature that was a main point when discussing the safety of having an always-on internet-connected mic in the house.

As for whether or not they're telling the truth, I don't know; but trigger-words have always been a feature that Amazon loved mentioning from a security/privacy standpoint.

> As for whether or not they're telling the truth, I don't know;

Luckily it's possible to check [0]! Although it gets a bit more complicated and can change, my understanding is that currently most people observe it increase it's network usage after it's trigger phrase, but not at other times (it uses the network for other stuff too, but audio data is typically rather large in comparison).

[0] https://www.iot-tests.org/2017/06/careless-whisper-does-amaz...

[1] 10.1007/s00779-018-1174-x <- Might want to use sci-hub

The pessimist in me think that a determined actor could simply capture non-trigger voice data offline, and bundle it with the rest of the traffic whenever the next trigger word occurs. But I am talking out my ass and have in no way verified any of this

It does seem to me like Amazon worded that statement very carefully, and in a way that allowed for that kind of behavior to occur.

Conversely, if they were not doing that kind of thing, they could have removed that ambiguity from what they said.

This was my thought too; there doesn't seem to be a way to verify this isn't happening.

If data is being buffered and only sent after the trigger words wouldn't the data transmitted vary depending on how much was said before the trigger word?

Maybe. All uploads could be padded with the maximum buffer size so you can't tell the difference. The buffer could flush only small amounts at a time. Some compression algorithm could be used that becomes more efficient with larger recordings.

What you should be asking with any "smart" device is "can I prove this device will do no harm to me".

Honestly I have never understood the value proposition of any smart device. Why would I want any of that functionality? Never once in my life have I ever wanted to talk to my TV. I'm beginning to (again) question the wisdom of carrying a smartphone.

Also see this Amazon patent related to intentionally removing PII from speech transmitted to the cloud.

http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=H...

In case the link doesn't work, it's US Patent 9922646.

Did you link the right patent? That claim is only for locating a user in a room based on multi-room voice input.

Grrr - wrong link:

http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO2&Sect2=H...

Yes of course. I don't get why that affects the privacy policy though. There are occasional false positives and you still might say something sensitive after the trigger word.

I’ve heard that the Echo’s mute button is a hardware off switch for the mic, though for obvious reasons you probably wouldn’t leave it on most of the time.

iOS can do voice recognition processing entirely locally for many languages. I’m sure the same is true for Android.

Google has made strides in local-only translation. I can't seem to find the Hacker News article but believe it was wired into the Google keyboard iirc.