Risk Drivers and Regulatory Pathways for Cross-Border Flows of Generative AI Training Data

Abstract

As a representative form of new quality productive forces, generative artificial intelligence (AI) depends fundamentally on the global circulation of data, creating a deeply coupled and mutually reinforcing relationship between model development and cross-border data flows. The scale of training data, the transnational spillover of associated risks, and the difficulty of effective oversight have nevertheless generated multidimensional security concerns involving national security, corporate interests, and personal privacy. Using a comparative analysis of the regulatory approaches adopted by the European Union and the United States, this article examines the structural sources of these risks and evaluates the practical limitations of existing governance arrangements. Grounded in China's domestic conditions and the holistic approach to national security, it proposes a tiered regulatory architecture based on three elements: differentiated preconditions for data protection and circulation, an independent regulator supported by multilevel coordination, and the international articulation of a Chinese approach to data governance. This framework seeks to balance security and development while contributing to a more coherent global regime for cross-border flows of generative AI training data.