Ethical AI is no longer discussed only in philosophy, policy, or public debate—it is increasingly being designed and built into technologies, standards, and engineering practices. This talk explores what happens when ideas such as values, rights, and responsibility move from public discussion into the everyday work of designing and deploying AI systems. Drawing on empirical investigations based in science and technology studies (STS), the talk examines the rise of “ethical AI by design” as a key approach to governing AI. I identify four main ways in which ethics is envisioned as designed into AI: through machine ethics, through transformative AI (Artificial General Intelligence or Superintelligence), through education of engineers, and through technical standards. I call these ‘design articulations’, emphasising how each imagines different roles for humans and machines, knowledge and ethical theories, and different ways of building and managing ethical responsibility. Some approaches focus on highly autonomous AI systems and future potential, while others emphasise human decision-making, professional practices, and standardisation. Across all of them, ethics becomes closely linked to risk management, technical standards, and market-driven innovation. This shift moves moral and political authority away from traditional public institutions and toward technical and corporate actors. I finish the talk by situating these results in relation to the broader goals of our project, Acting on AI: Digital Constitutionalism in the age of artificial intelligence.