Sure, but as with a lot of this stuff, this allows that to be done _at scale_. "[bad thing] happens anyway" isn't a great argument against "we should ban this thing where [bad thing] is likely to happen at massive frequency".

As for the rest of your argument, even if we take as a given that there is cases where facial recognition could be helpful, I'd say that if a company can't be trusted not to illegally develop their products it can hardly be trusted to respect strict standards when their products are in use