If the system flags the wrong person and a witch hunt ensues, it will validate every right-wing persecution complex there is.
Could you explain your worry further? To me, the problem with AI facial recognition is that a government or company using it has all the power. If they get a false positive, the wrong person gets hurt with no recourse. Civilians can't do anything detrimental with a person's identity that's not already illegal. Cops have been identifiable since they were catching slaves and outside of organized efforts (which this is not) there's been no issue with thatt.
Also, it seems trivial to add a step after the system returns an identification that checks if that person is actually employed by ICE. If not, oops it got it wrong; no harm, no foul. Even if it's wrong after that step, then what? I've seen no evidence that ICE agents are receiving anything beyond verbal harassment in the first place (outside of protests, where any hypothetical harm is random and not based on identity)
They don't need "powerful ammo". They're already rounding ppl up and illegally imprisoning them. And they plan to keep going further; ICE is on a hiring spree right now and just got a huge budget increase. They don't need valid excuses to be jackboots because they'll just do it anyway.
Refusing to use available tools to fight back is ridiculous. There is no "what if" here. They're already doing it and they've made it clear that want to do more