机器学习的著作权侵权判定:超越非表达性使用理论  

On the Determination of Copyright Infringement in Machine Learning:Beyond the Non-Expressive Use Theory

作　　者：涂藤 TU Teng

机构地区：[1]广东财经大学人工智能法研究中心,广东广州510320

出　　处：《政治与法律》2024年第10期162-176,共15页Political Science and Law

摘　　要：针对人工智能机器学习的著作权侵权判定难题,近期引人注目的非表达性使用理论根据“表达性机器学习”和“非表达性机器学习”的类型化方法划分侵权责任,并提倡禁止人工智能模仿特定作者的个人创作风格。然而,复制权的目的解释、历史解释和判例分析表明,非表达性使用理论未能走出长久以来“实施复制即侵权”的理论误区,面临逻辑、法理和现实层面的三重困境。对此,应当对非表达性使用理论进行扬弃,重构机器学习的著作权侵权判定标准,以公众接触原作品表达的高度盖然性取代“实施复制即侵权”的形式主义理念。Regarding the difficult issue of determining copyright infringement in the machine learning of artificial intelligence(AI),the theory of the non-expressive use which has recently attracted significant attention delineates the infringement liability based on the method of categorizing machine learning into"expressive one"and"non-expressive one",and advocates for prohibiting AI from imitating the personal creative style of specific authors.However,the teleological interpretation,historical interpretation and case-based analysis of the reproduction right reveal that the theory of non-expressive use fails to overcome the long-standing misconception that"carrying out the reproduction equals infringement",facing three-faceted predicaments at the logical,jurisprudential,and practical levels.Thus,it is necessary to critically assimilate the theory of non-expressive use,re-establish a new standard for determining copyright infringement in machine learning,replace the formalistic notion of"carrying out reproduction equals infringement"with the high probability of public access to the expression of the original work.

关 键 词：机器学习 人工智能 侵权判定 非表达性使用 高度盖然性 

分 类 号：DF523[政治法律—民商法学]