生成式人工智能数据训练合理使用规则的法理逻辑与实现路径OA

While generative artificial intelligence data training enhances the supply of high-quality works,it also induces risks of copyright infringement.The core controversy lies in whether the fair use rule can cover the process of accessing and utilizing works in data training.Based on the legal interpretation of market failure and transformative use theory,as well as the practical consideration of reducing biases in generative artificial intelligence technology,embedding the fair use rule from copyright law into the scenario of generative artificial intelligence data training is proved to be both feasible and necessary.However,challenges remain in the application of the fair use rule,including heightened conflicts of rights and interests,a narrow range of eligible subjects,and overly broad application scope.Therefore,it is necessary to expand the scope of fair use subjects to include commercial entities under the guidance of the principles of interest balancing and proportionality,recognizing commercial use under specific conditions.Meanwhile,the scope of works used under fair use in generative AI data training should be limited to the two stages of data input and model training,while in terms of behavior,it should be confined to acts of reproduction.