It seems to me that if you but a voice activated TV then you would expect it to listen to what you say. The issue here is where and how the translation of voice into command is done. I suspect the TV is too dumb to do accurate voice recognition on it's own so a sound bite is sent to a server somewhere. The server does the conversion and then either sends the command back to the TV or communicates with another server to stream the requested content. There has to be a certain amount of anonymity because the source of the sound bite isn't (I hope) tagged with the name and address of the source but some numerical identifier. I also suspect/hope that the sound is translated by machine and once that is done it is immediately discarded, also I would also think that anything that doesn't fall into the limited vocabulary that the server understands is immediately discarded and a "huh what was that?" returned to the user. Any other method of doing it makes no sense because Samsung expect to sell millions of the things. Recording millions of conversations is something that costs money and Samsung is in the business of spending money.
I suspect that someone at Samsung got a lawyer involved who said you have to disclose that the thing records sound and sends it to a third party. In the perfect storm where the sound is kept, and a human has access to it, and the same human can figure out who the speaker is, and the human cares, then it's an issue.
A lawyer would write "knives are sharp and can cause personal injury even death" but that doesn't mean I'm going to clear them out of the kitchen.