How AI is changing the legal framework in China, Hong Kong, the EU and the UK?

This article has been co-authored with Partner Ian Liu and Registered Foreign Lawyer Helen Xie from Deacons, Hong Kong's largest independent law firm.

Artificial intelligence (AI) is attracting significant R&D and investment globally, including in China, EU, the UK and Hong Kong. The rapid development of AI has raised significant questions regarding its impact on the legal framework in different jurisdictions, especially the laws and practice concerning copyright in AI-generated content. In this article, we will examine the recent case law and regulatory developments in China, EU, UK and Hong Kong related to AI-generated content (“AIGC”).

What’s happening in China and Hong Kong?

China

The existing PRC law does not specify whether AIGC constitutes a work protected by the PRC Copyright Law. Notwithstanding this, in the past year and a half, the PRC courts have decided three notable cases concerning the copyrightability of AIGC.

The first case, Li v Liu, was decided by the Beijing Internet Court in November 2023. The plaintiff, Li, used a generative AI software called Stable Diffusion to generate an image by inputting text prompts. After generating multiple images, Li selected and published a satisfactory one online. The defendant, Liu, used this image as an illustration for his online article without permission. Li sued Liu for copyright infringement. The court considered that the key issue in determining whether the generated image published by Li enjoyed copyright protection was whether it possessed originality and was a human intellectual creation. The court found that Li had provided significant intellectual input throughout the generation process, including inputting around 30 prompts to determine the output of the AI-generated image. This process reflected Li’s aesthetic choices and personal judgment, making the generated image not merely a “mechanical intellectual creation”, but one with originality. Therefore, the court ruled in favour of Li.

The second case, Lin Chen v Hangzhou Gaosi Membrane Technology, was decided by the Changshu Court in October 2024. The plaintiff, Lin Chen (“Chen”), used other generative AI software  Midjourney and Photoshop to create a 2D image featuring a half-heart on a lake. The defendant, Hangzhou Gaosi Membrane Technology (“Gaosi”), installed a 3D device resembling the half-heart in a lake and published pictures of it online for promotional purposes.

The court found that although Chen’s picture was created via AI software, Chen had amended the prompts to refine the image. Therefore, Chen’s picture was protected by copyright as it reflected his selection and arrangement of scene, environment, colour, light, shade, and angle of various elements in the picture. The court ruled in favour of Chen, finding that Gaosi’s publication of pictures constituted infringement.

Hong Kong

On 8 July 2024, the Hong Kong Government launched a two-month public consultation on amending the Copyright Ordinance in light of the importance of AI to Hong Kong’s technological and industrial transformation and the development of a digital economy.

The consultation paper proposes to introduce a new exception to copyright infringement for certain “text and data mining” (“TDM”) activities for both commercial and non-commercial use. Such TDM activities would include not only conventional text and data mining, but also computational data analysis and processing for enhancing the performance of a computer program, such as the development, training and enhancement of AI models.

Regarding copyright subsistence and ownership, the consultation paper considers AIGC to be covered by the provisions governing “computer-generated works” under the existing Copyright Ordinance (Cap. 528), such that the “person by whom the arrangements necessary for the creation” of an AI-generated literary, dramatic, musical, or artistic work will be taken to be the author and first owner of copyright in it. This position is in line with the UK, but contrasts with jurisdictions such as the US, where the strict requirement for “human authorship” of works currently means that AI-generated works are not eligible for copyright protection.

During the two-month consultation period, the Government received numerous responses from various industries with differing opinions. Generally, the creative industries, such as the Music Publisher Association of Hong Kong and the International Publishers Association, disagreed with the TDM exception. On the other hand, the software industry, including the Software Alliance, generally supported the TDM exception.

One of the concerns about a TDM exception is that it will prejudice the ability of copyright owners to exploit their works. The consultation paper suggests certain conditions to help mitigate this concern including requiring lawful access to a copyright work, so that the TDM exception will not apply if licensing schemes are available, or the copyright owners have expressly reserved their rights (giving owners an option to “opt out”).  David Wong, the Director of Intellectual Property Department, said on a recent radio program that copyright holders can simply declare their refusal to allow others to use their works, which constitutes an “opt-out”. However, Johnny Ng Kit-chong, the Deputy Chairman of the Panel on Commerce, Industry, Innovation and Technology, expressed concern that if a significant number of copyright holders choose to opt-out, it could hinder the development of AI in Hong Kong.

As regards copyright subsistence and ownership, the Hong Kong Bar Association – Committee on Intellectual Property considers that human creativity must be behind the work in order to attract copyright protection.

A bill to amend the Copyright Ordinance to support artificial intelligence development will be submitted to the Legislative Council in the first half of this year. While the consultation paper provides some insights into the AI-related copyright issues in Hong Kong, several practical concerns remain unresolved. It is advisable to stay informed about future developments regarding the amendment of the Copyright Ordinance in Hong Kong.

What is happening in EU and UK?

EU

The 'EU Artificial Intelligence (AI) Act" (EU AI Act) came into force on 1 August 2024 and is the first legislation of its kind in the world to regulate the use of AI.

The EU AI Act creates a legal framework for AI safety, protection of fundamental human rights and wider ethical principles. The EU AI Act's provisions will be phased in over time. The first set of requirements which ban "unacceptable risk AI systems" (eg social scoring) came into force in February 2025. This will be followed by additional obligations over the next two to three years, with full compliance for high-risk AI systems expected by 2027.

UK

By contrast, the UK does not currently have any overarching AI legislation. The UK currently adopts a "principles-based framework" for existing sector-specific regulators to interpret and apply to the development and use of AI within their domains. In July 2024, the new UK Labour government suggested that it plans to pass laws to address the risk of the "most powerful forms of AI". However, at the time of writing there are currently no AI-specific laws scheduled to pass through the UK parliamentary system.

In January 2025, the UK government unveiled the far-reaching 'AI Opportunities Action Plan' to unlock long-term productivity benefits across the UK economy. The policy and investment package is based on a 50-point action plan to develop compute infrastructure, boost skills, and open up public sector data sets to train AI models (amongst other measures) in a bid to become a global 'AI superpower'. This is seen as significant shift in tone by the UK government, which has historically focused on increased AI safety and regulation of the most powerful technologies. We anticipate that any new UK regulation, particularly in respect of AI and copyright (see section below), will align with this pro-innovation approach and potentially avoid imposing burdensome regulatory requirements on AI developers in the UK.

In terms of copyright infringement, in December 2024, the UK government opened a policy consultation on AI and copyright infringement, seeking input from businesses and industry stakeholders. The consultation will consider the same questions as a recent HK consultation paper – namely, whether (i) AI models should be prevented from training on any copyright-protected works without a licence, or (ii) if a current "text data mining" exception to copyright infringement (for non-commercial use) should in fact be expanded and codified for training AI models for commercial use.

The UK government's preferred position is a TDM which allows certain rights holder to reserve their rights, meaning they could opt out of available content being accesses for AI training or have the power to impose certain licencing conditions. Resulting legislation expected in 2025 or 2026 will ultimately look to balance the need to incentivise creative industries such as music, publishing and the arts whilst encouraging innovative AI training and development in the UK.

Why is UK and EU’s development important to businesses in Hong Kong/ China?

The EU AI Act takes a risk-based approach to establishing obligations on AI systems and their developers. There are transparency and explainability requirements that will apply broadly, even to users of "low – risk" AI systems such as chat bots, which will mean businesses will need to update how they communicate the fact that AI is being used with their customers. Providers of high-risk systems are subject to stricter obligations, such as mandatory "fundamental rights" impact assessments, data governance requirements and registration in an EU database. High-risk systems are generally those deployed in sensitive sectors such as biometrics, critical infrastructure, education, employment, law enforcement (amongst others).

Importantly for Hong Kong & China businesses is, the EU AI Act has "extraterritorial" effect. The obligations will apply to providers or deployers of AI Systems who are based outside of the EU where the ultimate output of those systems are within the EU. For example, a Hong Kong-based business using an AI system (whether created in-house or by a third party) to service customers in the EU will also have to comply with the various obligations in the Act.

Penalties for breaching the regulations are severe – ranging from EUR7.5m (or 1.5% of global annual turnover) to EUR35m (or 7% of annual turnover), depending on the type of infringement. Clearly robust AI compliance measures are crucial for all firms developing or using AI Systems in the EU market.

By contrast, there is no similar law in the UK as yet, however many UK businesses (who also operate in the EU) are using the EU AI Act as a reference point and "best practice" to establish their compliance programmes for responsible AI.

Conclusion

The emergence of generative AI has far-reaching implications for intellectual property rights, with copyright law being a critical area of concern. In both China and Hong Kong, courts are actively addressing these issues through case law, while regulatory developments across the world are also underway to address the complexities arising from AI-generated content.

Given AI’s designation as a national technological priority in China, it is likely that there will be ongoing advancements in law and policy related to AI-generated content. As such, businesses operating within China, or with interests in the region, should remain vigilant and attentive to these developments, ensuring they are adequately prepared to navigate the evolving technology and intellectual property landscape. In the meantime, while there is little coherent collaboration across jurisdictions to address the risks and issues of AI, global businesses should stay on top of developments in major jurisdictions.