The only way that this could ever be plausible is if a self improving algorythm, the likes of which already exist had an error which meant that the AI didnt only change code in irder to improve itself at its desired function, allowed other things to be changed, and developing a hate for humans.

the plausibaility of this is still ridiculous because the complexity of making a robot hate and want to overrun humans is ridiculous and it is not viable that this would be coded in by pure chance (no dont you fucking dare turn this religious)

Also the likelyhood of someone thats clever enough to, code a self developing algorithm and be doing a job that needs it, being dumb enough to allow such a huge lapse in quality of coding that the AI develops itself, without the coder noticing and completely out of line with objectives is ridiculous