The cat is already out of the bag. Facebook does mass facial recognition.

Other facial recognition code is already available in open source. The difference is the current open code is not designed to work with a massive database.

What is really stopping other agencies and LEOs from deploying facial recognition is they don't have the implementation budget or the skill to properly scope the project.

Edit: I'm not sure if I should be more worried about blowing $1bn on a failed project, or more worried if they actually succeed. See http://en.wikipedia.org/wiki/Virtual_Case_File