Artificial intelligence systems are not neutral; they encode institutional priorities, distribute risk, and reflect the power of their designers. Despite widespread ethical principles, a persistent gap remains between normative intent and technical implementation. Here, we show that responsibility in AI must be embedded in the architecture of design rather than applied retrospectively. Drawing on interdisciplinary research and governance analysis, this chapter introduces a framework structured around four design commitments: justification, consent and control, participation, and stewardship. Taken together, they connect ethical ambition to concrete engineering and institutional practice. It advances a model of accountability built on three reinforcing layers: internal governance, regulation that shapes design choices, and public oversight that sustains legitimacy. By reframing design as a mode of governance, the chapter argues that institutional responsibility and global equity depend on redistributing design agency and authority, establishing a foundation for AI systems that are both technically robust and publicly legitimate.