How are you applying a filter or circumventing anything if you are simply training a brand new model with a bunch of "cloaked" images?

Oh, i was thinking the scenario was more a specialized tool to remove the "glaze".

Hmm. That's a good point and now i am less sure.

I suspect it would come down to intentionality.

I looked at the definition in the law and it was unclear to me whether this technique met the definition. Among other things there's no "authorized deGlazing" to recover the original going on during human viewings. But there seemed to be enough leeway that maybe a judge could find a fit