​If AI ever becomes an existential threat to humans, it’s most likely because of bugs in the code that weaken the guardrails rather than AI becoming sentient and going rogue.