AI and copyright upgrade

Abstract

The rapid development of generative artificial intelligence (GAI) has sparked worldwide debates on how copyright law should respond to the challenges it has raised. In Hong Kong (HK), this conversation has taken centre stage in the recently published Consultation Paper on Copyright and Artificial Intelligence.1 With the aim of providing the HK legislator with a complete picture of the global debate, the School of Law at City University of Hong Kong (CityUHK) held an international conference entitled ‘Comparative Perspectives on AI and Copyright Law: Evaluating HK’s Policy Responses in the AI Era’ on 12–13 December 2024. The conference gathered international legal scholars, practitioners and policymakers to examine how copyright law and policy can properly respond to the AI challenges. This special issue, ‘AI and Copyright upgrate’, arises from that conference and presents six selected papers that together illuminate how copyright regimes can be updated for the AI era. Each contribution addresses a distinct facet of the GAI-copyright interface: the overall impact, copyrightability, infringement, intermediary liability, automated copyright enforcement and remuneration and inequality. Together, they offer insights into doctrinal rethinking, policy innovation and the fundamental values at stake. The issue opens with Daryl Lim’s article, which sets an ambitious tone by examining the extractive dynamics of GAI and their impact on core copyright assumptions. Lim deploys a vivid metaphor—Maurizio Cattelan’s Comedian (the infamous banana duct-taped to a wall)—to illustrate how GAI’s rise exposes structural inequalities in the creative economy. Lim highlights the extractive practices by which AI developers leverage vast amounts of human-created work without due credit or compensation, thereby amplifying existing power disparities between tech companies and individual creators. Lim argues that these inequities call for a recalibration of copyright law: rather than viewing AI as a neutral tool, the law must recognize and address the imbalance it creates. His contribution sets an equity-focused agenda for copyright reform, suggesting that any legislative responses must account for fairness to human artists and authors in an AI-driven marketplace. By rethinking foundational assumptions, Lim’s piece compellingly frames the normative stakes of AI’s impact on copyright and sets the stage for the more targeted analyses that follow. Following this broad structural critique, Chen Yang’s article turns to the issue of copyrightability of AI-generated content (AIGC). The focus is on HK’s ‘computer-generated work’ (CGW) doctrine under the HK Copyright Ordinance (HKCO), casting a critical eye on its ability to properly cover AIGC. Chen analyses HKCO, which the government asserts already, provides a backbone of copyright protection for AIGCs. Chen challenges this optimistic view by unpacking the doctrine’s limitations and the questionable assumptions behind it. In particular, he questions whether traditional requirements like originality or the so-called ‘necessary arranger’ rule can seamlessly extend to AIGCs by comparing the UK experiences. His paper argues that, without careful reconsideration, simply relying on the existing CGW framework is insufficient. While an overhaul may not be imminent, Chen’s piece underscores the need for a more nuanced approach if HK’s copyright regime is to truly harness AI’s creative opportunities. In his paper, Jiawei Zhang focuses on the much-debated issue of the potential copyright infringement risk of training AI using copyrighted works. He advocates a fundamental shift in regulatory perspective from inputs to outputs in the context of AI and copyright. Zhang argues that current debates fixate too much on the input side—the masses of copyrighted works ingested to train AI models—instead of focusing on the output—the contents that AI systems generate. He argues that an output-oriented approach would better calibrate copyright law to the realities of GAI. By judging AIGC on its own merits (for instance, whether an output unlawfully reproduces copyrighted works), policymakers can move away from abstract concerns over training data and towards concrete criteria for copyright infringement determination. This shift, he suggests, would lead to more balanced outcomes: it preserves incentives for human creativity while still allowing AI technology to flourish under clearer rules. The next article by Taorui Guan and Yang Lin tackles the issue related to the safe-harbour regimes for internet intermediaries. Their paper examines whether the safe-harbour regimes can be upgraded to accommodate the challenges raised by GAI through role-specific obligations. They note that the traditional Digital Millennium Copyright Act (DMCA)-style safe harbour—where internet services avoid liability by promptly removing infringing user uploads—does not translate neatly to AI systems, which do not store content in discrete files that can simply be taken down. To resolve this, they envision a reconfigured framework assigning tailored responsibilities to different players in the AI ecosystem. For example, AI model developers, platform providers and end-users would each have defined duties (such as monitoring, transparency or responsiveness to complaints) commensurate with their role in generating or disseminating AI content. This differentiated safe-harbour regime aims to maintain the DMCA’s innovation-friendly spirit while strengthening accountability: it would continue to shield good-faith innovators from crippling liability, but only on the condition that they proactively mitigate copyright risks appropriate to their function. Their contribution thus sketches a blueprint for legal reform that balances the protection of rights with the realities of AI-driven services. Connected to the previous article about intermediaries, Jesse Lu’s article focuses on the issue of platform governance and enforcement, criticizing the emerging trend of automated copyright moderation. He observes that, as platforms increasingly deploy algorithmic tools (like content filters and copyright bots) to police infringement, these systems often operate with minimal transparency or oversight. Lu argues that such ‘black box’ enforcement can erode due process: users may find their content removed or accounts penalized without a clear explanation or meaningful opportunity to appeal. Moreover, vesting quasi-regulatory power in private algorithms, he suggests, creates an accountability gap—one where corporate interests and error-prone AI can trump lawful user activities (eg, parody) with little recourse. To counter this, Lu calls for stronger regulatory checks on automated enforcement, including requirements for transparency in how infringement decisions are made and avenues for users to challenge wrongful removals. His piece underscores that any upgrade of copyright law in the AI era must not unfairly sacrifice individual rights and freedoms; on the contrary, it should impose ‘algorithmic accountability’ so that efficiency in enforcement does not come at the expense of fundamental rights and public interests. His contribution thus injects a note of caution: even as we adapt laws to govern AI, we must also govern the use of AI in law enforcement itself, keeping fundamental rights and values in sight. Rounding out the special issue, Rostam Neuwirth offers a provocative reframing of the entire AI-and-IP debate by shifting our focus to the overarching issue of global inequality. He argues that current discussions about AI and copyright—from questions of AI authorship to liability for AI-induced infringement—are missing the forest for the trees. The more pressing concern, in Neuwirth’s view, is that GAI is contributing to a widening global gap between those who control technology and the creative labour force that fuels it.2 Interestingly, but not surprisingly, his view echoes Lim’s from a different angle. He calls for rediscovering IP law’s original purpose of rewarding creators: rather than merely tweaking doctrines at the margins, the law should be reoriented to ensure that human creativity is justly compensated when AI systems become increasingly dominant. This could entail new legal mechanisms or reforms that guarantee authors a share in the value derived from AI’s use of their works, thereby preventing what he describes as the ‘plenty’ of AI’s output from making human creators ‘poor’. Neuwirth’s contribution, broad in scope and principle, ties together the theme of this special issue by reminding us that the ultimate goal of any AI-related copyright upgrade should focus on building a more equitable creative ecosystem. Together, these six articles demonstrate the multi-dimensional effort required to ‘upgrade’ copyright for the AI era. They range from rethinking fundamental doctrines (authorship and originality), to proposing new legislative and regulatory frameworks (for copyright infringement and for intermediary liability), to cautioning against unintended consequences of enforcement technologies and finally to re-centring the discussion on fairness and societal impact. Several common threads emerge. One is the importance of balance—balancing incentives for innovators with protection for creators, balancing the benefits of AI’s openness with the rights of those whose works are used, and balancing enforcement of rights with preservation of user liberties and the public domain. Another recurring theme is adaptability: copyright law, often rooted in pre-digital assumptions, must evolve in light of AI’s unprecedented capabilities, whether by updating old rules or by devising novel policy tools. Crucially, the contributions also remind us that copyright does not operate in a vacuum. GAI’s challenges intersect with questions of technology governance, competition and social justice. An ‘AI and Copyright Upgrade,’ therefore, it is not simply about doctrinal analysis—it is about ensuring that the copyright system continues to encourage human creativity and innovation while promoting equity and the public good in this new technological landscape. We hope that the ideas presented in this special issue will inform and inspire policymakers, academics and industry leaders as they work towards a future-proof and fair copyright regime for the AI age. Acting as the guest editor of this special issue, I would like to extend my gratitude to all the authors for their insightful contributions and careful research that made this special issue possible. I also thank the Hong Kong Commercial and Maritime Law Centre under the CityUHK School of Law for supporting the conference, which provided the fertile ground for these wonderful discussions. My gratitude also goes to all the conference participants, including Peter Yu, Guobin Cui, Jyh-An Lee, Yahong Li and Orabhund Panuspatthna, who kindly presented their views and shared their valuable comments. Special thanks to my colleague Yang Chen, our centre secretary Claire Dibo Huang and my PhD students Lingjun Gao and Yiyan Zhang, who co-organized the conference with me, for their hard work in setting up all the details. We are additionally grateful to the editorial team of the Journal of Intellectual Property Law & Practice, especially editor-in-chief Prof. Eleonora Rosati and managing editor Ms. Sarah Harris, for providing the invaluable platform for us, and reviewers who provided valuable feedback and helped shape these papers into their final form. Finally, we acknowledge the support of our institutions and colleagues in fostering an environment where cutting-edge topics like AI and copyright can be rigorously explored. This collective effort has made the ‘AI and Copyright Upgrade’ special issue a reality, and we trust that it will provide useful suggestions for the HK legislators to consider and contribute meaningfully to the ongoing dialogue at the intersection of technology and copyright law.