他山之石：欧盟人工智能监管的经验与启示

Learning from Masters: Experience and Enlightenment of EU Artificial Intelligence Regulation

摘要:

目前学界关于人工智能监管的内涵尚未达成统一、监管体系构建研究缺乏理论分析框架；同时在实践中，中国现有监管体系已经滞后于中国突飞猛进的人工智能发展，而欧盟在人工智能监管领域处于全球领先地位。为此，在明确人工智能监管内涵基础上对监管主体、方式和边界进行初探，以技术-组织-环境（TOE）框架为基础构建分析框架，从技术、组织、环境等层面剖析欧盟人工智能监管措施以及这些措施之间的相互作用，并总结其对完善中国人工智能监管体系的经验与启示。研究显示，“人工智能治理”和“人工智能监管”是两个相关但又有区别的概念，前者的内涵更为广泛、后者的内涵则较为集中，前者注重制定适当的指导方针、后者更注重对人工智能系统进行监督和控制，建构人工智能监管体系将成为保障社会稳定和可持续发展的重要基石。借鉴欧盟成功经验，中国政府可以从以下方面进一步完善人工智能监管体系：在技术层面强化可解释性人工智能技术使用；在组织层面建构人工智能监管的利益相关者合作机制和设立央地协同的人工智能监管机构；在环境层面开展人工智能监管沙盒试点、完善基于风险分级的人工智能监管法律框架和加强人工智能监管的区域与国际合作。

关键词:人工智能监管;监管体系;监管措施;技术-组织-环境分析框架;欧盟

Abstract:

At present, the connotation of artificial intelligence regulation has not been unified, and the research on the construction of regulatory system lacks a theoretical analysis framework. At the same time, in practice, China's existing regulatory system has lagged behind China's rapid development of AI, while the EU is in the global leader in the field of AI regulation. Therefore, on the basis of clarifying the connotation of AI regulation, the preliminary exploration of the subject, mode and boundary of regulation, an analytical framework based on the technology-organization-environment(TOE) framework is constructed, the AI regulatory measures and the interaction between these measures of EU are analyzed from the perspectives of technology, organization and environment, and its experience and enlightenment to improve China's AI regulatory system are summarized. Research shows that "AI governance" and "AI regulation" are two related but different concepts, the connotation of the former is broader, while the connotation of the latter is more concentrated, the former focuses on developing appropriate guidelines, the latter focuses more on monitoring and control of AI systems, the construction of artificial intelligence regulatory system will become an important cornerstone to ensure social stability and sustainable development. Drawing on the EU's successful experience, the Chinese government can further improve the AI regulatory system in the following aspects: (1)At the technical level, strengthen the use of explainable artificial intelligence technology. (2)At the organizational level, construct a stakeholder cooperation mechanism for artificial intelligence supervision and establish a central-local coordination artificial intelligence supervision agency. (3)At the environmental level, carry out artificial intelligence regulatory sandbox pilots, improve the legal framework for artificial intelligence supervision based on risk classification, and strengthen regional and international cooperation in artificial intelligence supervision.

Key words:artificial intelligence regulation; regulatory system; regulatory measures; TOE framework; European Union