In cases where it may be possible for a hypothetical AI to seriously harm people via a network connection (regardless of whether it involves highly technical ex

In cases where it may be possible for a hypothetical AI to seriously harm people via a network connection (regardless of whether it involves highly technical exploits or just social engineering) we should probably be much more worried about humans doing it first, perhaps even right now. Because there’s a lot of malicious humans out there already. And our society is already dangerously dependent on fragile technology.