Did you know that the training data that trained all the LLMs was, once upon a time, entirely and painstakingly tagged by real live humans? Why would ethics scenarios be any different?