However, as you said, they don't need to. Localization and usage metrics alone can tell you an incredible amount of extremely detailed information.
For foreground, I suppose it'd be fairly trivial to tell if the app was making calls to AVAudioRecorder.
The basic thing you need to make people comprehend: to learn a fact X, you don't have to actually learn it - it's enough for you to learn about such Y that P(X|Y) >> P(X). And you don't have to learn Y either, you have to learn V and W such that P(Y|W) >> P(Y) and P(Y|Z) >> P(Y). Etc. This method applies recursively.
And once they comprehend the causality graph this forms, you need to make them understand just how much information we radiate all the time, and how humans are still very inefficient in doing the calculations mentioned above with all that data. Facebook is only the tip of an iceberg; it's only going to get worse from here, because modern technology keeps letting us explore the causality graph even further and faster.
Unfortunately, I don't think they would. They'd get a slap on the wrist fine, maybe have one or two senior executives resign, they'd post a mea culpa, saying how they now it was wrong and promise to do better, and then everyone would forget about it a few weeks later when some new thing started to dominate the news cycle.