作　者：陈镜先
关键词：人工智能；税收征管；智慧税务；税收大数据；算法歧视；纳税人权利；税法规制
摘　要：

近年来，人工智能在税收征管中应用的深度和广度不断拓展。税务机关运用人工智能打击避税行为，这在提升征管效率、防范税收风险的同时，也对纳税人的平等权、知情权、信息权、救济权带来新的挑战。鉴于信息法、人工智能法目前的通用规制框架难以精准契合税收征管场景的特殊性，税法有必要在与相关法律保持协调衔接的基础上，基于纳税人中心主义的价值取向，采取问题导向的渐进式模式对税收征管运用人工智能进行规制。在制度设计上，税法规制应遵循“两端并举”的实践路径：在税务机关端，构建“事前授权评估—事中禁止完全自动化决策—事后明确责任归属”的全流程权力约束机制，从源头规范人工智能的应用边界与运行逻辑；在纳税人端，通过界定算法歧视认定标准、确立分层化算法解释义务、规范涉税数据收集使用规则、完善技术适配型司法审查机制等，实现纳税人平等权、知情权、信息权与救济权的体系化保障，最终推动税收征管效能提升与纳税人权利保护的有机统一。

 

The Challenges to Taxpayer Rights Posed by the Application Of Artificial Intelligence in Tax Collection and Administration And Regulatory Responses Under Tax Law
Chen Jingxian（Wuhan University）
Abstract In recent years, artificial intelligence (AI) has been applied in tax collection and administration with growing depth and breadth. While the deployment of AI by tax authorities to combat tax avoidance serves to enhance administrative efficiency and prevent tax risks, it also poses new challenges concerning tax‐ payers' rights to equality, information, data privacy, and judicial remedy. Given that the general regulatory frameworks of current information and AI-related laws are ill-suited to address the particularities of tax ad‐ ministration scenarios, it is necessary for tax law to regulate AI applications in this field. Such regulation should adopt a problem-oriented and gradual approach, maintain coordination and consistency with relevant legal regimes, and remain grounded, in a taxpayer-centered value orientation. In terms of institutional design, tax law regulation should follow a "dual-track" practical approach. As for tax authorities, a comprehensive power-constraining mechanism should be established across the entire process -- exante authorization and assessment, a prohibition on fully automated decision-making during the process, and ex post clarification of responsibility -- to define the boundaries and operational logic of AI applications at the source. As regards taxpayers, systematic protection of their right to equality, information, data privacy, and judicial remedy should be achieved by defining standards for identifying algorithmic discrimination, establishing tiered obligations for algorithmic explanations, regulating the collection and use of tax-related data, and improving technology- adaptive mechanisms of judicial review. Ultimately, this dual-track approach aims to promote the organic integration of enhanced tax administrative efficiency with robust protection of taxpayers' rights.
Key words artificial intelligence; tax collection and administration; smart taxation; taxation big data; algorithmic discrimination; taxpayer rights; tax law regulation

 

■ 作者简介 陈镜先，武汉大学国际法研究所讲师，湖北 武汉 430072。

[PDF]（下载数：）