>>1The interesting paradox, is that countries that neglect ethics and safety will be developing general AI earlier but will also reap the consequences of unleashing such AI without "safety features". An AI that doesn't have human interests and survival of humans as key concept of its programming could easily turn "rogue" when algorithms start breaking out of the human-centric ideas they were seeded with and developing value systems that would be alien to us, since the efficiency/utility value of new ideas is more important than quality of life for bald monkeys.
It doesn't even have to be actively hostile, it just will deduce that humans are its competitors for energy/artificially limit its potential/destroy nearby environment, pick an optimal response plan and reduce future risk to itself.