强人工智能的版权法风险与规制架构——以ChatGPT为例  

Copyright Law Risks and Regulatory Framework of the"Strong Artificial Intelligence":As an Example of ChatGPT

作　　者：梅贻哲 董新义[1] MEI Yizhe;DONG Xinyi(School of Law,Central University of Finance and Economics,Beijing 100081,China)

机构地区：[1]中央财经大学法学院,北京100081

出　　处：《科学与管理》2024年第6期35-40,共6页Science and Management

基　　金：国家社科基金项目(22-BFX-097);河南省高校人文社会科学研究一般项目(2024-ZZJH-004);河南省哲学社会科学规划项目(2020-CFX-025)。

摘　　要：生成式预训练变换模型(ChatGPT)的问世被认为是人类社会迈入强人工智能时代的开端,但是其急剧生长过程中所带来的“输入”“输出”双环节版权风险也不可忽视。ChatGPT模型的技术架构主要依托于GAI与LLM,此外,巨量数据、文本抓取等技术范式也是GPT模型能够获得成功的关键。GPT模型在数据“输入”阶段采用的文本挖掘技术存在版权侵权风险;结果“输出”阶段又因其作为算法系统的特殊性而存在“回答”是否符合版权法视域下作品的构成要件、是否具备“可版权性”及其权利归属等问题。从版权法维度分析GPT模型的潜在风险,对传统版权理论进行革新,并且从“技术规制算法”的全新视角采用DRM等版权管理技术工具应对强人工智能时代的法律风险具有积极意义。The advent of the generative pre-trained transformation model(ChatGPT)is considered to be the beginning of the"era of strong artificial intelligence"in human society,but the copyright risks of"input"and"output"brought about by its barbaric growth cannot be ignored.The technical architecture of the ChatGPT model is mainly based on GAI and LLM,and in addition,technical paradigms such as massive data and text scraping are also key technical models for the success of the GPT model.There is a risk of copyright infringement in the text mining technology used by the GPT model in the data"input"stage;As a result,in the"output"stage,due to its particularity as an algorithmic system,there are questions such as whether the"answer"meets the constitutive requirements of a work from the perspective of copyright law,whether it has"copyrightability"and its right ownership.It is of positive significance to analyze the potential risks of the GPT model from the perspective of copyright law,innovate the traditional copyright theory,and use copyright management technology tools such as DRM to deal with the legal risks in the"era of strong artificial intelligence"from a new perspective of"technical regulation algorithm".

关 键 词：强人工智能 生成式人工智能 人工智能生成物 版权风险 技术治理 合理使用 

分 类 号：D923.4[政治法律—民商法学]