The ultimate test for AI ethics will be AI vs AI. Like, if an AI can detect and shut down another AI going rogue, that's a safety net. Game them against each other, survival of the ethically fittest!