The rapid advancement of generative artificial intelligence (AI) has brought not only new possibilities for creativity, automation, and cross-domain collaboration, but also profound ethical challenges. This paper proposes a structured ethical reconstruction framework that addresses the systemic tensions emerging from the widespread integration of Generative AI into public knowledge, institutional judgment, and individual agency. First, we construct a three-dimensional model of ethical tensions across cognition, behavior, and governance, revealing how misalignment among these domains can amplify risks such as misinformation, algorithmic bias, and responsibility vacuums. Second, drawing on the theory of embedded governance, we present a multi-layered responsibility framework that integrates national regulation, platform mechanisms, organizational ethics, and user accountability. Finally, the paper affirms the irreplaceable value of human expression and judgment through three institutional mechanisms: the right to interrupt, the right to contest output, and the right to human oversight. These mechanisms are embedded into Generative AI systems via interface design, content labeling, and regulatory instruments. By combining philosophical reflection with implementable governance pathways, this study contributes to the evolving discourse on trustworthy and inclusive Generative AI, with practical implications for policy, platform design, and user ethics.