Untitled Page

Author	Mr KAMRAN AHMADLI ILHAM
Organisation/Institution	Tongji University, Shanghai International College of Intellectual Property
Country	CHINA
Panel	Intellectual Property Rights
Title	Text and Data Mining Exceptions for AI Model Training under Chinese Copyright Law: An EU-Comparative Approach
Abstract	The development of the AI industry in China depends on large-scale text and data mining (TDM), yet the legality of such practices remains unclear. Thus, the lack of an explicit TDM exception in Article 24 of the Chinese Copyright Law creates persistent grey zones and exposes stakeholders to ongoing risk. Moreover, the 2023 Interim Measures on AI add another layer of pressure by requiring developers to use lawful training data and respect IP rights, without addressing TDM-based training. The article, therefore, examines how Chinese law can accommodate TDM for AI model training to give developers legal certainty while preserving rightsholders’ interests. To map this ambiguity, the paper first examines Chinese judicial and policy developments, including the 2011 guidance on reasonable use under the three-step test and the recent “Ultraman” case, where training is accepted as fair use as long as it avoids market substitution and does not disclose expressive elements. Though it points to an evolving judicial stance in China, the decision still does not supply nationwide certainty. Against this background, the article uses the European Union’s DSM Directive as a structured template and analyses mandatory and general TDM exceptions under Articles 3 and 4, respectively, the latter being for any purpose with lawful access, subject to rightsholders’ opt-out. Read together, this architecture provides an operational pathway for AI training while preserving control over copyrights. The paper then argues for realistic reforms in China, referring to statutory clarification as a long-term solution to introduce a dedicated “TDM exception” as a central measure. However, in the short term, a more realistic and urgent path is the affirmation and consolidation of adaptive judicial interpretations, combined with soft-law guidance, to prepare the ground for more sophisticated TDM exceptions.
Biography	Kamran Ahmadli is a doctoral student at the Shanghai International College of Intellectual Property (SICIP), Tongji University, where he focuses on copyright law and emerging technologies. He holds an LL.M. in Intellectual Property Law from Tongji University and an LL.B. (Law) from Baku State University. Before starting his PhD, Kamran worked as an expert lawyer at “Əsaslı Tikinti və Təchizat” LLC (under the Ministry of Culture of the Republic of Azerbaijan), and previously as a junior lawyer at Dentons and Azerbaijan Railways. He also gained practical experience through leadership roles in student-run societies and internships with local law firms and public authorities, working on commercial, contract, and administrative matters. His research interests lie in copyright and other areas of intellectual property law, AI and data governance, and space law, with a particular focus on how legal systems adapt to new technologies. In his current work, including this paper, he examines how closed-list copyright systems can respond to large-scale text and data mining for AI while maintaining a fair balance between innovation and rightsholders’ interests.