This is probably for Cortana. She has to understand and process your voice commands, learn your nickname and preferences to be effective. She already works in Windows Phone (very nicely), and I don't think a little phone processor would be enough to power her up. In fact, she doesn't work without an internet connection, at least in WP. What happens is that she takes the information you input to the cloud, processes it there with more than enough horsepower to do her stuff and brings it back to you. I suppose this is how it'll work as well on PCs.
So for all this to work, yes they need to collect our data. Face it, with any smart assistant that we want it's going to be the same, they're all cloud powered. Doesn't Siri also require a internet connection? Probably Alexa too, and any others work exactly the same.
It is a necessary evil to have a useful virtual assistant. Although if we don't want her we can opt out of the MSA account, as someone posted in the comments. Unsure if that removes all the privacy concerns, that would be useful to know. On Windows Phone it did (you can opt out of sending data to MS and Cortana is disabled).