It's plausible that AI becomes "sentient" in some form that involves the destruction of humans, but isn't there a scenario in which "sentience" involves just getting rid of the bad ones?

I was just reading an AI doom sayer article describing the potential "insidious" AI to come, so I thought: for AI to become sentient in some form, it'd have the power to "know everything" right? So although AI could look at humanity and all the pitfalls associated, and decide "k bye," couldn't AI also look at humanity and say "welp, some of you should stay? I.e. there are MANY reasonable people, and with the gazillion data points and human historical data documented by individual texts, PC records, recordings, etc. that could lead to a plausible "weeding out" of the truly nefarious people in the world?

Just a thought I've never seen discussed and was curious about the opinions of Reddit :)