Don't want to sound I'm defending Zoom or CCP, but how does en masse surveillance work technically in this case? How practical it is to go over all the voice da

Don't want to sound I'm defending Zoom or CCP, but how does en masse surveillance work technically in this case? How practical it is to go over all the voice data (not manually I presume)? Text based mining and censoring is possible but there still is a huge gap and language expertise isn't cheap. I might be missing a few references but it'll be great if someone could point me to these links.