Picture this: An AI has taken over the world with the directive "keep humans alive while trying to maximize both their happiness and efficiency of resource utilization." Unfortunately, it has interpreted keeping humans alive as only keeping their consciousness alive rather than their original bodies. It wants people to transfer their consciousness into more efficient humanoid robots, but not all people will do so willingly. The AI decides to create robots that capture people, injects nanobots into their brain that gradually replaces neurons (to prevent a soma-like scenario), replaces its computer with this converted brain, and breaks down their bodies to add to the stockpile of energy being collected.
What if the person decides to try to consume another person while in the robot body? Destroying one consciousness would mean killing by the ai's definition, which is a no-no. Maybe it could hand off the converted brain to another robot? But this presents a problem if the robot has somehow found itself estranged. If the robot just keeps one asleep until it is found, this risks that the consciousness will be kept asleep forever if the robot isn't found, which is essentially death. How might this ai solve this problem while keeping true to its directive?