I have complicated and rapidly evolving feelings about AI existential risk but I find this argument extremely unconvincing – we achieved this ability in the mid-1950s and have come very close to using it several times. (And many important nuclear alarm-raisers were religious.)
It takes an incredibly impoverished, materialistic, hubristic, godless view of man to assume that we are so powerful as to be able to wipe ourselves out. There are real security concerns coming from AI, but we can address those without resorting to science fiction.