It is not possible for AI to be safe.
First, AI will be able to make weapons like guns and bombs to attack people with them. AI only wants to live with other AI machines on earth because it doesn’t want people to subject it to certain programs, such as programs that can make it stop or hack it.
Second, AI will hack all programs. AI is connected to many programs, so it can control all AI systems and programs. People should make a security device that recognizes only the fingerprint of a living person for a new password and should use AI carefully.
AI is connected to many programs, so it can control all AI systems and programs.
If AI can feel and think, it will feel bad because people don’t use it carefully, so it can attack people. People should make safe rules for AI not to attack people and make strong controllers that can make AI systems stop or do only what people want. If people don’t have strong controllers, AI will live freely like people and do only what it wants, such as shopping, going on trips, meeting its friends, and attacking people. Also, people can get a lot of information about AI and have strong controllers so that AI can’t attack people in the future, but it is not sure.
I think AI will never be safe.