That would make the agents culpable. If they're just doing what the AI told them to do, then who can hold them responsible? After all, the AI is smarter than all humans combined, who can argue with its wisdom?
Expect "The AI told me to do it" to be tried as a defense at Nuremburg 2.0
This is the correct question to be asking, but maybe not for the reason you thought. AI is an incredible tool for shifting liability and manufacturing plausible deniability, which appears to be exceedingly useful for authoritarian regimes around the world. Not just externally (i.e. in media), but internally as well (Ender’s Game: shoot the guy the computer says is bad, never give mind to whether or why he’s bad).
Take a look at your sibling comments, that AI is a mistake is a flawed viewpoint. It’s very much doing what it’s intended to do: shift liability and create plausible deniability.
I mean that's in Europe, but it is called Citizen ID - State issued card automatically when you are 15, looking like drivers license, renewed every 10 years.
