How is the OP right? Huge claims require huge evidence; this trope has been disproven over and over again. Security researchers look at exactly this kind of thing, and nevermind this community is full of the people who would actually build such a thing. A massive dragnet isn’t actually as valuable as you think it would be.
It has not been disproven. A simple Google search "is my phone listening to me" provides a resounding yes.
These things need to hear your prompt at the very least, which entails (in many cases, if not most) listening at all times.
Security researchers DO look at this and they, along with everyone else, just shrug because you technically DID give consent when you accepted the thousand line policy you didn't read.
Hot word detection to activate an assistant occurs locally on your device for the purpose of activating said assistant. That is a far cry from “Pretty sure Apple and Google already do this, just to all phones, in all homes, and not just for music, but your entire life!” which suggests that both Apple and Google are deploying a dragnet uploading a 24/7 recording to their servers for nefarious purposes. That is just simply not true. That at minimum would leave a constant trail of bytes being sent over the network (which isn’t the case), and would massively drain battery life. It’s also highly illegal.
You and I can go back and forth on how the other is wrong by providing various internet links. I'll start:
"Mobile devices, the researchers conclude, listen to conversations through microphones and create personalized ads based on what the person wants or has done." [1]
"This passive listening ensures the virtual assistants are ready to help you with a task when needed. However, depending on the developer, voice tech apps may also use your conversation data to recommend ads and content. For instance, Google uses Assistant conversation data to personalize ad and content recommendations. Others, like Apple’s Siri, claim not to use conversation data to build marketing profiles or curate ads." [2]
But this exercise will actually produce less fruitful results because it's possible to prove anything via the online "research" nowadays. So let's try a different tack- thinking for ourselves.
FAANG are extremely notorious brokers of data. Everything about you and your browsing behavior is collected. I hope we can agree on this. Then why on earth would your conclusion be "they don't broker or process our audio data"? You'll have to have something better than it would consume battery life or would leave an identifiable trail of bytes, both of which could be mitigated by some clever programming.
Much better to have the hypothesis (hence "pretty sure") that they do, and then scrutinize. Until something definitive comes out that they absolutely do not, it's much more solid ground than a conclusion that you can simply trust these large corporations.
