TBH I expect all security agencies are using or trialling this sort of technology. It automates a very labour intensive task and helps them track bad guys with a lot less effort. Although of course the problem is they say that as they don't know who all the bad guys are then it's best to track everyone just in case. Most governments don't allow that (at least they say that they don't).
The technology has not been that great atm and gets it wrong as much as 20% of the time, but as the algorithms develop and the accuracy improves there will be more pressure to use it. It's just another way that people will be replaced by robots and computers.
They've been learning: The "I am not a robot" tests you see on websites are filled with pictures that the computers got wrong, but they've learned with our help and now that challenge is regularly hacked by the computers of spammers - and that's just a simple example without the super computing power of a national agency.