> It is still up to the detective to investigate further and decide whether to pursue charges. And then it is up to court to issue the warrant.

This is how it should work, but I still think it is important to discuss these failures in the context of AI risks.

One of the largest real-world dangers of AI (as we define that now) is that it is often confidently wrong and this is a terrible situation when it comes to human factors.

A lot of people are wired in such a way that perceived confidence hacks right through their amygdala and they immediately default to trust, no matter how unwarranted.