Jiakang Yu
Generative artificial intelligence now produces expressive works—images, music, code, even legal analysis—that resemble human creativity with increasing sophistication. Yet copyright systems around the world still assume that authorship is exclusively human. As states scramble to update their intellectual property regimes, the question of whether AI-generated works deserve legal protection has become one of the central challenges in international copyright law.
This blog post summarizes how three major jurisdictions—the United States, the European Union, and China—define authorship in the age of generative AI. Their divergent approaches illustrate the growing possibility of global regulatory fragmentation and the need for harmonized international standards.
I. United States: Human Authorship and Fair Use Flexibility
American copyright law remains firmly anchored in the principle that authors must be human. Courts and agencies have repeatedly emphasized that creativity requires human intellectual conception. In Thaler v. Perlmutter, the U.S. Court of Appeals for the District of Columbia Circuit affirmed that AI-generated works cannot be copyrighted absent meaningful human input.[1] The U.S. Copyright Office’s 2025 Copyright and Artificial Intelligence report similarly rejects the idea that prompts alone constitute creative authorship.[2]
Because AI developers rely heavily on large datasets, the United States has leaned on the doctrine of fair use to resolve questions surrounding training data. Scholars such as Joshua Landau emphasize that AI relies on non-expressive data processing and suggest that fair use is the doctrinal mechanism most likely to govern training data.[3] Others propose limited statutory reforms that would recognize the human operator as the rights holder when AI acts autonomously.[4]
The result is a flexible but uncertain system—one that protects human creativity but offers little clarity for AI-generated outputs.
II. European Union: Moral Rights, Transparency, and Text-and-Data Mining
The European Union approaches copyright through a rights-based framework grounded in human dignity and moral rights.[5] Under the Digital Single Market Directive, the EU permits text-and-data mining (TDM) for research and limited commercial uses, but rightsholders may opt out.[6]
The EU AI Act adds another layer by regulating AI developers through transparency and dataset-disclosure obligations.[7] Scholars such as Marcus von Welser note that the EU treats AI training as a separate, regulated act that requires its own legal justification.[8]
Compared to the U.S., the EU model prioritizes authorship integrity and data protection—important goals, but ones that can significantly increase compliance burdens for cross-border AI development.
III. China: Administrative Pragmatism and State Supervision
China’s regulatory model blends industrial policy with administrative oversight. The 2023 Interim Measures for the Management of Generative AI Services require developers to use “lawful and truthful” data, register algorithms, and respect intellectual property.[9]
Chinese courts have shown more openness than U.S. courts to assigning copyright to human operators of generative AI systems.[10] At the same time, scholars warn that China’s approach facilitates both rapid technological deployment and potential state-directed information control.[11]
This system offers regulatory certainty but raises concerns about transparency and human-rights protections.
IV. The Emerging Global Trend: Human-Centered, Innovation-Sensitive Law
Despite their differences, the three systems share a common theme: none recognize AI as an independent author. Instead, they protect human creative agency while experimenting with mechanisms to regulate training data, oversee AI developers, or assign output ownership.
Comparative scholars increasingly call for harmonization through the World Intellectual Property Organization or other multilateral frameworks to avoid conflicting national rules that hinder international innovation.[12] As generative AI becomes integral to global creative industries, cross-border alignment on authorship, training data, and output ownership may become not just beneficial—but essential.
[1] Thaler v. Perlmutter, 130 F.4th 1039 (D.C. Cir. 2025).
[2] U.S. Copyright Office, Copyright and Artificial Intelligence, Part II: Copyrightability (2025).
[3] Joshua Landau, Artificial Intelligence and Intellectual Property, Part III: IP Protection for AI-Assisted Inventions and Creative Works, 14 Am. Univ. Bus. L. Rev. 811 (2025).
[4] Kavya Rallabhandi, The Copyright Authorship Conundrum for Works Generated by Artificial Intelligence: A Proposal for Standardized International Guidelines in the WIPO Copyright Treaty, 54 Geo. Wash. Int’l L. Rev. 311 (2023).
[5] See Charter of Fundamental Rights of the European Union, art. 17(2), 2012 O.J. (C 326) 391 (“Intellectual property shall be protected.”).
[6] Directive 2019/790 of the European Parliament and of the Council of 17 April 2019 on Copyright and Related Rights in the Digital Single Market, 2019 O.J. (L 130) 92 (EU).
[7] Regulation 2024/1689 of June 13, 2024, Laying Down Harmonised Rules on Artificial Intelligence and Amending Regulations (EC) No 300/2008, (EU) No 127/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act) O.J. (L 2024/1689)(EU).
[8] Marcus von Welser, A European View on Generative AI and Copyright, 17 Landslide 23 (2025).
[9] Cyberspace Admin. of China, Interim Measures for the Management of Generative Artificial Intelligence Services (July 10, 2023).
[10] Shenzhen Tencent v. Shanghai Yingxun (Shenzhen Nanshan People’s Ct. 2019) (China).
[11] Liana Edgar, Generative AI and Disinformation: Analyzing China’s Strategy Amidst U.S. Investment and Export Controls, 26 S.D. Int’l L.J. 147 (2025).
[12] Frishta Abdul Wali, Copyright Law and Generative AI: A Comparative Analysis on Fair Use Across the U.S., E.U., and China, 51 Rutgers Comput. & Tech. L.J. 293 (2025).