Such a dumb take.

First, no single engineer that will completely implement this "spying feature". A hardware engineer knows how to put the mic inside the TV, the FW engineer knows how to write driver for application to get mic data, the application knows how use the pre-trained ML blob to process the data and send the result to Cloud. At that level, no one really knows what happening any more.

Can they make educated guess what is happening? yes, but still, it's just a guess.

There would be "alignment meeting" accross the various team, but trust me, they would never use the word "spying" in any slides.