If somebody someday far far far into the future manages to invent AI which both approximates an exceeds human capability, which is honestly doubtful because somebody capable of doing so would likely abstain from it, then at that point the AI would likely view humanity as something useful for coexistence rather than some animalistic threat.

Of course the absolute worst way to go about creating such a thing would be the physical and digital reconstruction of human and animal brains and experiences such that it would feel things like fear, self-preservation, and anger, but such attempts probably wouldn't provide any real value to its creators or financers in the first place.