Reviving ‘computer-generated works’: should Hong Kong copyright law adapt the rule to harness AI opportunities?

In July 2024, the Hong Kong government released a public consultation paper on copyright and artificial intelligence (AI), addressing emerging copyright issues in the era of generative AI. A key focus is the government’s interpretation of the existing Copyright Ordinance provisions on computer-generated works (CGWs), concluding that these provisions already support copyright protection for AI-generated content (AIGC), including literary, dramatic, musical and artistic (LDMA) works and non-LDMA works. The government claims that the current framework provides sufficient incentives for AI investment and creativity.

However, this article challenges the government’s position by identifying three major concerns. First, applying the CGWs provisions to AIGC under a plain reading leads to an ‘originality paradox’ that cannot be resolved by reframing the doctrine as entrepreneurial rights. Second, this approach risks a ‘contributor thicket’, where too many individuals involved in the AI creation chain could qualify as authors. Third, protecting AIGC, irrespective of human input, undermines the policy objectives the government seeks to promote; any potential benefits may be outweighed by resulting harms.

The article urges the government to adopt a more cautious and nuanced approach. Rather than relying on ambiguous provisions to advance distant policy goals, the government should use forthcoming guidelines to promote a refined interpretation of the CGWs doctrine. More importantly, it should initiate a new round of consultation to reconsider the viability of the CGWs provisions, with serious consideration given to their revision, repeal or replacement with a new rights model tailored for the AI era.

Artificial intelligence (AI) is reshaping the real world. As the public becomes increasingly familiar with accessible generative AI models like ChatGPT and DeepSeek and views them as useful tools for boosting productivity,1 one may still be surprised by how recently developed AI agents like Manus can autonomously manage tasks for humans. AI is likewise transforming the legal world by rendering certain pre-existing doctrines obsolete, prompting their reconsideration or reform and even facilitating the emergence of new regulations or legal doctrines. There is extensive discussion on whether and how governments should regulate AI development to mitigate potential risks while preserving innovation. Attention has been given to possible solutions for deepfake and related personality interest protection issues. Scholars are rethinking key doctrines in trade secrets law, including ‘readily ascertainable’, ‘reasonable secrecy measures’ and ‘improper means of acquisition’. The enforceability of anti-competition and anti-reverse engineering clauses in widely used AI models is also under scrutiny. Some commentators have raised concerns about the declining trust in trade marks due to AI-generated fake reviews influencing consumer perception. The question of whether AI-assisted or fully AI-generated inventions can qualify for patent protection remains under debate.

Beyond all these issues, copyright law has garnered equal—or, to put it more boldly, even greater—attention from scholars, professionals and policymakers alike since the advent of generative AI. Scholars have explored copyright infringement issues concerning AI model training using copyrighted materials and related exceptions. This article strives to dissect the intricate copyrightability of AI-generated content (AIGC).

The complex question of whether AIGC can receive copyright protection has been preoccupying scholars and policymakers across jurisdictions. For instance, the US Copyright Office initially responded unfavourably to AIGC, rejecting registrations of works created solely by AI and requiring AIGC to be dictated rather than merely influenced by users’ prompts to receive protection. In 2022, the Office denied registration for the artwork A Recent Entrance to Paradise, which the applicant, Thaler, claimed was generated entirely by a computer program without any human involvement. Both the district court and the appellate court upheld the decision, emphasizing the human authorship requirement for copyright protection. The circuit court, however, does not draw a line to address the question of the extent of human contribution required for copyright protection. In contrast, the Office has attempted to delineate this boundary more precisely by recently permitting certain registrations—so long as applicants can demonstrate meaningful human involvement, such as the selection, arrangement and coordination of AI-generated elements in producing the final work. As a comparison, Mainland China has indicated its lenient approach towards the copyrightability of AIGC by granting copyright protection for images generated by AI models through users’ text prompts. Jurisdictions like the USA and China focus their discussions on whether human authors have a sufficient original contribution to the resulting works generated by AI (ie the human authorship and originality requirements).

Other jurisdictions that follow the unique computer-generated works (CGWs) doctrine have engaged in different discussions. In response to the challenges AI technology poses to copyright law, the UK government launched two public consultations: one in 2022 to solicit public views on whether and how the CGWs doctrine can protect AIGC and another in 2024 expressing more explicit concerns about applying this doctrine to protect AIGC. Similarly, the Hong Kong Copyright Ordinance (CO) contains an identical CGWs provision. The Hong Kong government’s recent public consultation shares a similar focus, although it takes a more opportunistic approach. By analysing the provisions of the existing CO related to CGWs, the government concludes that ‘the existing provisions of the CO already provide the backbone for copyright protection of AI-generated works, covering both AI-generated LDMA works and AI-generated non-LDMA works’. The government claims that the current CO offers sufficient incentives to invest in AI technology, advancing creative endeavours and encouraging creativity through AI.

This article examines the CGWs provision in Hong Kong and the government’s stance reflected in its recent public consultation paper. It critiques the assumptions underlying the government’s position and raises concerns about the resilience and efficiency of applying the CGWs provision to protect AIGC in Hong Kong, as envisioned by policymakers.

The article begins with a brief introduction to the CGWs provision, which originates from UK law, and explores how this provision can be understood based on its wording and the case law example. Section II analyses the Hong Kong public consultation paper, elaborating on its perspectives regarding the applicability of the CGWs provision to AIGC. This analysis is followed by a qualitative empirical account of public comments submitted during the consultation period, summarizing key perspectives on the CGWs provision. Building on these findings, the article critically examines how the government and the public may have misinterpreted the CGWs provision, identifying unsound assumptions and potential problems arising from these misunderstandings. The article concludes by proposing possible paths forward for AI and copyright law reform in Hong Kong.

Source: https://academic.oup.com/jiplp/advance-article/doi/10.1093/jiplp/jpaf035/8131776