Lets say we've got to the point where we've developed a true AI and that AI had its own free will. But we've programmed in various traits, for example, they can get annoyed, feel anger and various other emotions. Certain desires. Certain dependencies (for example, that acquisition of energy to power them). We lead them out into the world with only a certain set of rules and to obey our laws, but rather than to program these rules into them, allow them to choose.
I cannot see the difference between this and the creation of a child. Thus the common answers to the OP questions would be the same as the common answers for a child. Those answers have been hashed for so long they're pretty boring.
If we remove the illusion of free will from both the AIs and humans, the answers get more interesting.