This is interesting discussion.
http://www.theverge.com/2014/1/6/528056 ... my-and-you
The gist is, companies can pay AT&T to allow their data through to your phone without affecting your data cap. It's just another way that the carriers can differentiate, though it adds a layer of complexity. It also re-opens the debate that caps are stupid. Or are they?
You could make the argument that you buy a pipe (a speed tier) and whatever data you can get, is yours. You're paying for access to equipment and access to faster gear is more expensive, for the most part.
Except the thing the cap does is limit your use of the network. So perhaps if we had no caps, they would have to throttle instead. Would I rather have 3Mbps unlimited or 10-20Mbps (up to the limit of the network and general RF characteristics) with a 2-5GB monthly cap? I'd rather have the monthly cap, based on my own observations. For my computer, I think I'd want the opposite. Throttle me and let me buy more or less bandwidth, but don't cap me. And this is basically what we have today. The 3Mbps of unlimited is Sprint, 10-20Mbps capped is everyone else (10-20Mbps being the limit of the network not a throttle), and the bandwidth tier of unlimited is Cox. Sounds similar to Sprint but the network has actual performance...
All that said, one person suggested we have some oversight/regulation to ensure that, say, Vimeo has the opportunity to bid as a Data Sponsor just as freely as Youtube can. I agree with this. Otherwise, I don't necessarily see the harm in this approach. LIke someone else in the comments said, what if Youtube and Netflix were major data drivers? And what if Youtube and/or Netflix had a Sponsored Data agreement with AT&T so that you stream as much as you want? My response would be, if Netflix subscriptions don't increase, and Youtube remains free, then it seems like a value add. As a bonus, if AT&T service got cheaper as a result (doubtful but economically feasible), even better.
It just means the consumer has to weigh more things but weighed properly, the result could be a better value than a simpler option.