
Reflecting on President Trump’s first 100 days in office
As artificial intelligence continues to reshape industries, understanding the evolving regulatory landscape is more critical than ever. Our new APAC AI Watch series offers in-depth analysis of key legal developments across Asia-Pacific, helping businesses anticipate compliance challenges and seize emerging opportunities.
Each article will break down legislative updates, policy trends, and enforcement actions across the region, providing practical guidance on how businesses can navigate evolving AI regulations, mitigate risks, and refine their strategies.
Follow our series for the latest insights and expert analysis.
The Hong Kong SAR Government plans to introduce a bill with AI-related amendments to the Copyright Ordinance (“CO”) by July 2025, with a view to securing passage by the current term of the Legislative Council. Once passed, the amendments will have important implications for AI developers and copyright owners. We discuss below the public consultation on copyright and AI and the proposed amendments.
In 2024, the Government conducted a two-month public consultation on copyright and AI, gathering perspectives on copyright protection of AI-generated works, liability for copyright infringement as well as the exceptions under the CO. See our articles here and here.
The Legislative Council released a paper on the outcomes of the consultation in February 2025, setting out the Government’s proposed enhancement of the CO regarding protection for AI technology development, as pledged in the Chief Executive’s 2024 Policy Address.
The responses from the public consultation were considerably mixed where respondents generally supported the Government’s review of copyright laws in light of AI technology. However, there was “no clear consensus” among respondents from different sectors with divergent views. For example, some respondents felt the current CO had sufficient protection for AI-generated works, while others sought clearer guidelines and legislative amendments. Generally, there was an emphasis on the importance of a comprehensive copyright framework to foster innovation in the technology sector, while simultaneously balancing rights of copyright owners and public interest.
A key legislative proposal following the consultation is to introduce a text and data mining (TDM) exception in the CO to facilitate data mining for AI training purposes. Similar exceptions can be seen in other jurisdictions such as the UK, EU and US:
The proposed TDM exception in Hong Kong will allow reasonable use of copyright works for computational data analysis and processing for both commercial and non-commercial uses. To balance the interests of copyright owners, the proposed TDM exception will be subject to restrictive conditions, such as a “three-step test” as set out in the Berne Convention for the Protection of Literary and Artistic Works and the WTO TRIPS Agreement. The three limbs of this test are:
Other restrictive conditions will include:
The opt-out option draws from the EU approach. In this respect, the Government aims to establish a straightforward and accessible mechanism for copyright owners to reserve their rights, while maintaining flexibility in implementation. For copyright works made publicly available online, the opt-out is proposed to be made by machine-readable means.
The proposed TDM exception aims to foster the growth of AI technology by providing legal clarity. It is intended to reduce the administrative burden for AI developers through allowing the access and use of diverse datasets necessary for AI training, while balancing the interests of copyright owners.
At the same time, it remains to be seen how practicable it would be for AI developers to comply with the proposed conditions (for example, the requirement to disclose records of the source of copyright works). Another potential concern is whether the “opt-out” option for copyright owners would essentially undermine the TDM exception. Further, while the “opt-out” option is designed with a view to protecting the interests of copyright owners, further guidance is needed on the implementation of feasible opt-out options.
Authored by Eugene Low and PJ Kaur.