How do we tell whether a human is safe? Incrementally granted trust with ongoing oversight is probably the best bet. Anyway, the first mailicious AGI would prob

How do we tell whether a human is safe? Incrementally granted trust with ongoing oversight is probably the best bet. Anyway, the first mailicious AGI would probably act like a toddler script-kiddie not some superhuman social engineering mastermind