I thought about something similar in the context of "dangerous" AI. In a hypothetical scenario where super-smart AI got control of the internet and all the devices, would it be able to start controlling people?