人工智能训练数据收集的合法性困境与制度建构  

Legitimacy Dilemma and Institutional Construction of Artificial Intelligence Training Data Collecti

作　　者：黄绍坤 Huang Shaokun

机构地区：[1]武汉大学法学院

出　　处：《荆楚法学》2024年第3期89-100,共12页Jingchu Law Review

基　　金：国家社会科学基金后期资助项目“国家数字法学视域下私权力研究”(项目编号:23FFXB034)的阶段性成果。

摘　　要：人工智能训练数据收集的合法性困境,严重阻滞了其发展。在数字时代,数据日渐具备要素化、公共属性化、使用规模化等特征。数据要素化要求释放数据效能,公共属性需要弱化私人控制,使用规模化表明个别协商无法进行。对此,制度建构应从传统语境转向数字语境,从个体权利保护思路转向数据要素利用思路。在进行制度建构时,应打破著作权、隐私权等权利领域的限制,基于各项权利内容数据化后的共性,进行一体化建构。在对场景、获取对象、获取方式和使用目的等进行限制的前提下,设立训练数据获取的许可豁免规则,设置知情权、拒绝权、删除权、算法解释权等权利,强化主体责任,并积极探索人工智能收益分配机制。The legality dilemma of the use of artificial intelligence training data has seriously hindered its development.In the digital age,data is increasingly characterized by elements,public attributes,and large-scale use.Data factionalization requires the release of data effi-ciency,public attributes need to weaken private control,and the use of scale indicates that in-dividual negotiation cannot be carried out.In order to adapt to the development of digital socie-ty,the construction of artificial intelligence data utilization rules should shift from traditional context to digital context,and from individual rights protection to data element utilization.In the construction of the system,we should break the restrictions in the fields of copyright,priva-cy and other rights,and carry out an integrated construction based on the commonness of each right content after data.Under the premise of restricting the scene,acquisition object,acquisi-tion method and purpose of use,the permission exemption rules for the acquisition of training data are set up,and the rights such as the right to know,the right to reject,the right to delete,and the right to interpret algorithms are set up,so as to strengthen the main responsibility of service providers,and actively explore the mechanism of artificial intelligence income distribu-tion.

关 键 词：人工智能 训练数据 著作权 合理使用 数据在先权利 

分 类 号：D922.17[政治法律—宪法学与行政法学] TP18[政治法律—法学]