人工智能训练数据的版权信息披露:理论基础与制度安排  

Disclosure of Copyright Information in Artificial Intelligence Training Data:Theoretical Foundation and Institutional Arrangements

作　　者：李安[1] Li An

机构地区：[1]中南财经政法大学知识产权研究中心

出　　处：《比较法研究》2024年第5期136-152,共17页Journal of Comparative Law

基　　金：中国博士后科学基金第74批面上资助,获2024年度国家资助博士后研究人员计划B档资助。

摘　　要：人工智能训练数据的版权信息披露是当前全球人工智能法治热切关注的问题。在透明度议题下,训练数据版权信息披露具有实现法律问责、促进技术改良的工具价值,同时也具备增进信任、责任心和合作的内在价值。在版权法视野下,披露训练数据中的版权信息,一方面是对作者精神权利的承认和尊重,另一方面可为著作财产利益的实现提供便利。欧美法律实践表明,训练数据版权信息披露应妥当设置披露范围、披露自由度、披露标准、披露例外等,以实现人工智能企业、版权人和消费者之间的利益衡平。我国应以透明价值和版权目标为取向,为不同类型人工智能的训练数据版权信息披露梯次配置柔性规范、中性规范、刚性规范,并对特定情形作出例外规定,同时构建多方参与的规范实施机制。The issue of disclosing copyright information in Artificial Intelligence(AI)training data is significant concern in global AI law.Transparency in revealing copyright details within training data holds both practical and intrinsic value.It plays a crucial role in establishing legal accountability,fostering technological advancements,and building trust,responsibility,and cooperation.From a copyright law perspective,disclosing copyright information not only acknowledges and respects the moral rights of authors'but also facilitates the protection of copyright property interests.Legal practices in the EU and USA have shown that in order to strike a balance between the interests of AI enterprises,copyright owners,and consumers,rules governing the disclosure of copyright information in training data should address aspects such as disclosure scope,freedom,standards,exceptions,etc.To align with transparent values and copyright objectives,China should establish flexible,neutral,and strict norms for disclosing copyright information in training data based on different types of AI.It should also make exceptions for specific circumstances and develop a multi-party enforcement mechanism to ensure compliance.

关 键 词：文本数据挖掘 透明度 版权信息披露 合理使用 人工智能法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] D99[自动化与计算机技术—控制科学与工程] D913[政治法律—国际法学]