生成式人工智能的二维垄断风险和分类分级治理路径  

Two-dimensional Monopoly Risk and Category-and Class-based Governance Path of Generative Artificial Intelligence

作　　者：黄宏宇 HUANG Hongyu(Law School,Central South University,Changsha 410083,China)

机构地区：[1]中南大学法学院,湖南长沙410083

出　　处：《北京邮电大学学报（社会科学版）》2024年第5期8-18,共11页Journal of Beijing University of Posts and Telecommunications(Social Sciences Edition)

摘　　要：对于生成式人工智能这类具有自我学习、思考和决策能力的新兴技术,反垄断应从垄断行为规制转向垄断风险治理。因兼具消费产品和技术模型的双重技术性质,生成式人工智能既能够凭借自主意识垄断数据、构筑竞争壁垒、剥削消费者利益,还能够持续赋能作为生成式人工智能应用者的科技巨头巩固垄断地位、实施垄断行为,具有在技术和应用两个维度可预见的垄断风险。对此,反垄断应在明确维护和提高市场可竞争性的时代要义基础上,注意对两个维度垄断风险的场景化治理。对生成式人工智能技术本身,可以借鉴欧盟《人工智能法》的先发经验,根据生成式人工智能的运行机制分层分级界定垄断风险;对生成式人工智能应用者,可以根据事前合规、事中规范和事后规制三个阶段属性构建具有针对性的科学治理路径。总体而言,要提高反垄断对技术全链条和主体全方位的穿透式监管能力。For emerging technologies that have the ability of self-learning,self-thinking and self-decision-making such as generative artificial intelligence,anti-monopoly should shift from monopoly behavior regulation to monopoly risk governance.Because of the dual technical nature of both consuming products and technical models,generative artificial intelligence can not only monopolize data,build competition barriers and exploit consumers with its independent consciousness,but also continue to enable tech giants as controllers of generative artificial intelligence to consolidate monopoly status and implement monopoly behaviors.It has predictable monopoly risk on the dimensions of technology and application.In this regard,anti-monopoly should adhere to the scenario-based management of two-dimensional monopoly risk on the basis of maintaining and improving competitiveness of the market.For generative artificial intelligence itself,through learning from the EU Artificial Intelligence Act,monopoly risk can be categorized and classified according to the operational mechanism of generative artificial intelligence.For generative artificial intelligence controllers,targeted scientific governance paths should be built in the light of prior compliance governance,in-process governance and post-regulation.In general,it is necessary to strengthen the anti-monopoly penetrated regulation on the technology and the entity.

关 键 词：生成式人工智能 垄断风险 分类分级治理 穿透式监管 

分 类 号：D912.29[政治法律—民商法学]