Almost all of this is heavily exaggerated, presenting relatively innocuous features as if they were harmful, or calling out simple bugs as if they were intentional and actively malicious. None of it is flat-out wrong per-se (at least as far as I can tell), and I can certainly see where GNU is coming from, but I think for the average user most of the features they're criticizing actually do a lot more good than harm.…
It literally is a backdoor into your phone. Whether it's ethical or not depends if they've ever used it with law enforcement or for their own personal gain.
> They describe a bug where Google Assistant incorrectly identified the "Okay Google" hotword as "Google Assistant recording users' conversations"
This isn't necessarily great but I've run into situations where it activated on completely unrelated phrases. Is it Google's best interest as an ad company to sell data to fix this? No.
The issues you see as someone else pointed out are because you implicitly trust Google. I personally don't trust Google. The fact they manipulate their products for "algorithmic fairness" is proof they have ulterior motives one way or another. Is it that far of a stretch to think they could be abusing other components of their platform either intentionally or not?