大型体育赛事适用人工智能技术的规制路径——以智能视频监控为例  

Regulatory Paths for Applying Artificial Intelligence Technology to Large-scale Sports Events

作　　者：黎浩田 李茵晖 LI Haotian;LI Yinhui(School of Law of Wuhan University,Wuhan Hubei 430064;Department of Physical Education of Zhongnan University of Economics and Law,Wuhan Hubei 430073)

机构地区：[1]武汉大学法学院,湖北武汉430064 [2]中南财经政法大学体育部,湖北武汉430073

出　　处：《成都体育学院学报》2025年第1期25-33,共9页Journal of Chengdu Sport University

基　　金：国家社会科学基金一般项目“算法营销侵蚀消费者自主选择权的法律规制研究”(24BFX052);湖北省体育局决策咨询研究项目“《体育法》实施背景下湖北省反兴奋剂教育的制度协同研究”(2024B002)。

摘　　要：人工智能技术在奥运会、世界杯、亚运会等大型体育赛事中应用于安保监控引发了人们对数据隐私、算法歧视等风险的广泛担忧。针对现行数字领域立法规制此类行为的局限性,有必要反思当下基于风险预防和基于分层治理的人工智能技术规制路径。对于体育赛事场馆中的智能视频监控,应当将规制焦点从预先定义风险转移至系统监测风险上来,将规制重心从模型分层治理推进至要素分阶规制的路径上来。针对大型体育赛事智能视频监控的全生命周期构设“预训练-微调及部署”的分阶规制机制:在预训练中的数据处理阶段,坚持标识数据和去标识数据二分法,动态完善关涉国家、企业和个人权益的类型清单,基于数据重要程度和去标识化程度对重新识别概率进行动态评估和信息披露。在预训练中的算法设计和模型开发阶段,构建与管理风险、监测严重事件、执行模型评估和对抗性测试相关的义务。在微调及部署阶段,对智能监控系统进行自动分类分流、自动预警预测、自动取证固证的功能区分,并构设相应的算法审计、算法备案、公开和解释机制。The application of AI technology for security monitoring in large-scale sports events such as the Olympic Games,the World Cup and the Asian Games has triggered widespread concerns about data privacy,algorithmic discrimination and other risks.Given the limitations of the existing legislation in the digital domain to regulate such behaviors,it is necessary to reflect on the current path of regulating AI technology application based on risk prevention and based on layered governance.For intelligent video surveillance in sports event venues,the focus of regulation should be shifted from pre-defined risks to systematic monitoring of risks,and the center of regulation should be pushed forward from model-based hierarchical governance to the path of elemental hierarchical regulation.For the whole life cycle of intelligent video surveillance of large-scale sports events,a hierarchical regulation mechanism of“pre-training-fine-tuning and deployment”should be set up:in the data processing phase of pre-training,the dichotomy of labelled data and de-labelled data is maintained to dynamically improve the list of types of data related to the rights and interests of the state,enterprises and individuals,and the dynamic assessment of re-identification probability is carried out based on the degree of data significance and the degree of de-labelling.Based on the degree of data importance and de-identification,the probability of re-identification is dynamically valuated and information is disclosed.During the algorithm design and model development phases of pre-training,obligations related to managing risk,monitoring critical incidents,performing model evaluation and adversarial testing are constructed.In the fine-tuning and deployment phase,differentiate the functions of automatic classification and triage,automatic warning and prediction,and automatic forensics and evidence fixing for the intelligent monitoring system,and construct corresponding algorithmic auditing,algorithmic filing,disclosure,and interpretation mechanisms.

关 键 词：2024年巴黎奥运会 大型体育赛事 智能视频监控 分阶规制 

分 类 号：G80-052[文化科学—运动人体科学]