Shian-Loong Bernard Lew Thank you for an interesting question. I am always worried about the "blind" rule followers. Zygmunt Bauman explains many challenges with the rule follower. A machine is hard to understand as ethical more than that it is an excellent rule follower. That means that the machine might have a sort of "moral" strategy, but would be harder to understand as having an ethical reflective capability. Maybe the difference is both marginal and irrelevant in many cases.
Since ethics and morals are related to human behavior a manual audit is the only way to ensure compliance to standards and ethics. Machines do not understand nuances and can be misled by programmers to hide questionable ethics scenarios.