Side anecdote: One weekday after I vacationed in Tahoe, I saw a BART (subway) ad for Tahoe, and I was like, "oh, great , probably because I just came back from ... wait, that's not possible!"
Any one of:
* OS-level confirmation (permission entitlements, cpu usage, etc)
* Packets resembling sound data being caught in flight.
* An internal leak of the method they'd be using to do so from one of three of the largest tech employers. Not even Apple can keep their secrets secret, and they're probably the most paranoid tech company in existence.
You know, literally anything concrete, rather than evidence-free accusations based on fallible memory.
So far, not one bit of these instances can't be explained by a combination of Baader-Meinhof and confirmation bias, with a mix of plenty of non-audio data that Facebook no doubt has. People are so willing to paint FB as this boogeyman that they're disregarding basic logic.
I understand that any one person’s anecdotes are weak evidence, but your comments are going much further and claiming that such tests can never be evidence, even though much scientific knowledge is similarly obtained.
Not even necessary. You could do keyword recognition on the device itself, pushing a list of keyword<->waveform maps, and sending an indicator when they're recognized.