I think they already have such systems via bots, but they don't block or undo things automatically because I assume it's hard to know programmatically the licence of things that are posted on the Internet.
It it is hard, that they should not run wikipedia.
That is crazy. Even systems like ContentID are full of flaws.
Kinda hard like the lawers are still after years trying to figure out the issue between Oracle and Google, Apple and Samsung... It's a really hard legal issue, expecting it to be easy for a computer to solve. It's naive.