Exploring Legal Pathways for Fair Use of Copyright for AI Data Training
Data training profoundly influences the functionality and effectiveness of artificial intelligence(AI).Its involvement in the use of copyrighted works has triggered numerous copyright disputes both domestically and internationally,with the fair use principle frequently invoked as a defense in such cases.The legal rules and application of fair use in China have long been subject to various judicial challenges,which are also reflected in determining fair use in the context of AI data training.Therefore,there is an urgent need to consider the legal rules of fair use for AI data training and shape a Chinese approach.There are divergent views on copyright infringement in AI data training and its regulatory pathways in academia and practice.Regulatory pathways for copyright infringement in AI data training can be categorized into four main models:acknowledging the existence of copyright infringement,acknowledging copyright infringement but claiming statutory licensing,acknowledging copyright infringement but claiming fair use,and denying the existence of copyright infringement.After analysis,it is concluded that fair use is the optimal solution for regulating copyright infringement in data training.However,China's current copyright fair use rules have not yet been adapted to address AI data training issues.The transformative use test in the U.S.,as a general rule for expanding fair use scenarios,is not suitable for direct transplantation into Chinese law to determine the new types of uses,including AI data training,due to its theoretical and practical flaws.Based on that the legalization of AI data training aligns with China's social policy development needs,helps to enhance the new productive forces,and fosters the fair growth of the domestic and international AI industry,China should establish specific AI data training legal provisions for fair use.However,it should be noted that even with the establishment of specific legal provisions for fair use in AI data training,the substantial interests of copyright holders must not be ignored.This article conducts in-depth research and comparative analysis of the current mainstream regulatory pathways for copyright infringement in AI data training.After concluding that copyright infringement in AI data training should be regulated through fair use,it further clarifies and justifies the establishment of specific legal provisions for fair use in AI data training.Through comparative analysis with other relevant statutory proposals and by analyzing the objectives and implementation principles of specific legal provisions for fair use in AI data training,this article creatively puts forward detailed design suggestions for such provisions.In the context of AI data training,concise fair use legal provisions are proposed to comprehensively and effectively achieve the purposes and balancing spirit of copyright law,aiming to minimize frictions and maximize benefits for all the stakeholders.
artificial intelligencecopyright infringementdata trainingfair usetransformative use
万方数据
维普
