It's also noteworthy that listening to audio via phone microphones is terrible. Speech to text works like shit, and the expectation is that people need to speak as plainly as possible, and over a long period of manual adjustments will it get to a point where it's halfway usable.

Ever gotten a pocket dial from someone? Can you hear anything that even resembles speech over the rustling of fabric? Seems like a wild leap to assume that corpos are listening in on random audio, when the software designed around people specifically speaking plainly and clearly to their phone barely works at all.

Plenty of things to be concerned about with info privacy, but it's important to recognize the limitations of hardware.