大模型数据训练中的著作权合理使用规则研究

Research on the Rules of Fair Use of Copyright in Large Model Data Training

摘要: 随着数字传播技术的不断演进，著作权法原本旨在通过权利设计激发创作、推动传播的立法宗旨，在实践过程中逐渐遭遇适用难题。权利人与使用者的利益平衡发生倾斜，相关方在现行法律体系下展开了激烈的博弈。然而，当前司法实践在新型技术使用行为前面临判定僵化问题，因此应当在现有法律框架内通过灵活解释著作权法条款以容纳合理的数据训练行为的可能性，也考虑通过修订《著作权法实施条例》以增设机器合理使用条款，结合国内外产业实践和版权法上的相关规定，完善合理使用的判定方法，并规定大模型服务提供者的相关义务。

Abstract: With the continuous evolution of digital communication technologies, the legislative purpose of copyright law, which was originally intended to stimulate creation and promote dissemination through the design of rights, has gradually encountered difficulties in application in practice. The balance of interests between rights holders and users has tilted, and the relevant parties have engaged in intense debates within the current legal system. However, current judicial practice faces rigid judgment issues in the face of new technological uses. Therefore, it is necessary to accommodate the possibility of reasonable data training within the existing legal framework through flexible interpretation of copyright provisions, and also consider revising the “Implementation Regulations of the Copyright Law” to include provisions for the fair use of machines, combining domestic and international industry practices and relevant copyright regulations, improving methods for determining fair use, and stipulating the relevant obligations of large model service providers.