Mass Facial Recognition and the First Amendment

I think this draft is a valiant effort, but I think there are still some serious gaps in the argument that need to be filled.

"Facial recognition" can mean either the real-time identification of individuals from opportunistic or intentional acquisition, or image searching on previously acquired images. You are correct that Clearview AI is simply making a commodity service based on familiar technology, and is doing so with an unclean set of training data and a searching database that cannot pass at least some legal standards. But how training data or the search dataset are acquired is separate in all respects from whether it is constitutionally problematic for the state to search for faces in databases. Showing crime victims booking mugshots of criminals is not unconstitutional, and whether the program in use was trained on Fair Face and is searching dreivers' license photos is mere technical detail.

Faces are evidence, whether seen by witnesses or by cameras. But there are secondary consequences of, e.g. the collection of every identity that can be determined from photographs at peaceful demonstrations rather than crime scenes. It might be that the best route to improvement is to deal at the level of detail rather than constitutional theory for more of the 1,000 words.