AI won't kill ALL humans, but will be used to kill a large number of a very specific kind of humans, the humans that billionaires don't like.

What AI labs are doing is constructing a narrative that AI could be dangerous, so when they actually press the button to mass kill humans, the remaining humans will blame the AI, not the people who pressed the button.