<p>Artificial intelligence (AI), as the core driver of the new wave of technological revolution and industrial transformation, is comprehensively embedding itself into the domain of social governance in the posture of a "creative subject," propelling a deep paradigm shift from "experience-driven" to "data-enabled" governance. However, AI governance models originating from Western technological contexts generate institutional friction with the differentiated institutional environments, cultural habitus, and power structures of various countries during the process of global diffusion, triggering tensions in the localization of governance knowledge. Drawing on Karl Polanyi’s theory of "embeddedness" as its foundational framework, and integrating Clifford Geertz’s epistemology of situated knowledge with Doreen Massey’s "progressive sense of place," this paper constructs a three-stage analytical model of "global diffusion–local embedding–dynamic reconstruction" to systematically examine the mechanisms of knowledge construction, encoding transformation pathways, and localization reconstruction logic of AI technology in the field of social governance. The research finds that the local embedding of AI governance knowledge manifests through three distinct pathways: "technological external forces transforming governance schemas," "pluralistic construction of subject identity," and "localized practice of emergent governance forms." Simultaneously, algorithmic opacity, monopoly of data power, and the crisis of governance subjectivity constitute the core alienation risks constraining the modernization of AI-driven social governance. On this basis, the paper proposes a practical framework of "contextual adaptation, subject restoration, and pluralistic synergy," providing theoretical grounding and policy reference for charting a path of AI social governance with Chinese characteristics.</p>