News & Insights
Singapore’s Creative Industry grapples with GenAI – Can Copyright Law help strike a balance?
The intersection of generative artificial intelligence (“GenAI”) and copyright law presents complex legal challenges in Singapore’s evolving digital landscape. This article examines how Singapore’s Copyright Act 2021 applies to GenAI technologies, potential infringement scenarios, and available exceptions that developers and users may potentially rely upon. As recent developments through April 2025 have demonstrated, there is a growing need for clarity in this rapidly changing field.
The Evolution and Impact of Generative AI Technologies
Artificial intelligence refers to technologies that simulate human cognitive functions including knowledge processing, reasoning, problem solving, learning, and planning. These systems analyse input data to produce decisions or outputs with varying degrees of autonomy.[1] GenAI represents an advanced subset of these technologies, capable of creating new content such as text, images, audio, and other media by learning patterns from vast training datasets. [2]The exponential growth of GenAI platforms has transformed digital content creation. Midjourney enables users to generate sophisticated images from simple text descriptions, while ChatGPT, an AI chatbot developed by OpenAI, acquired an unprecedented one million users within just five days of its launch in November 2022.[3] This rapid adoption demonstrates both the extraordinary public interest in these technologies and their potential to disrupt creative industries.
The proliferation of GenAI has sparked significant concerns among Singapore’s creative community. Local artists have expressed apprehension about AI-generated art potentially undermining their livelihoods and creative endeavours.[4] An illustrative example of this occurred in April 2025, when social media platforms were flooded with AI-generated images mimicking the distinctive artistic style of Studio Ghibli animations, igniting intense debate around issues of plagiarism and copyright infringement.[5] As of end-May 2025, Singapore courts have yet to issue any rulings on cases specifically addressing copyright issues arising from the use of GenAI, making this an area of jurisprudence that is in need of development.
Singapore’s Copyright Framework
Copyright in Singapore is governed by the Copyright Act 2021 (“CA”). The CA provides comprehensive protection for original creative works, granting copyright owners exclusive rights to reproduce, publish, perform, communicate, and adapt their creations.
Copyright protection arises automatically upon creation of original works expressed in tangible form – no formal registration is required. The CA protects literary, musical, dramatic, and artistic works, along with sound recordings, films, broadcasts, and performances.[6] Protection extends to the expression of ideas rather than the ideas themselves, a fundamental principle of copyright law globally.
Under the CA, copyright owners have the right to take legal action against infringers and may seek various remedies including injunctions to halt infringing activities and damages to compensate for monetary losses suffered.[7] Singapore’s copyright regime aims to strike a balance between protecting creators’ rights and enabling reasonable use of such works for innovation, education, and the public interest.
Areas of Concern in the Generative AI Ecosystem
Training Data Acquisition and Processing
The development and operation of GenAI systems potentially interacts with copyright law at multiple points. The first critical juncture occurs during the training phase, where developers collect vast datasets to teach AI models pattern recognition and content generation techniques.
Under the CA, copyright infringement occurs when a person copies another’s work wholly or substantially without consent or authorization.[8] The CA defines copying broadly to include:
- Storage of work in a computer or electronic medium[9];
- Conversion of work into digital or other electronic machine-readable form[10]; and
- Making temporary or incidental copies in the course of other uses[11].
When developers engage in web scraping to acquire training data, they may potentially infringe copyright in the materials gathered if they reproduce protected works without permission. This is particularly relevant when training datasets include books, articles, artwork, photographs, or other copyright-protected content obtained from the internet or other sources without proper authorization.
AI-Generated Outputs and Derivative Works
Another area of concern involves the output created by GenAI systems. Copyright infringement concerns may arise when AI-generated content bears substantial similarity to protected works used in training data. For example, an AI system that generates summaries of political events might reproduce text segments verbatim or nearly verbatim from news articles or commentaries used in its training.
Similarly, image generation models might create visual works that incorporate distinctive elements of copyrighted artwork, potentially constituting unauthorized derivative works. The Studio Ghibli-style images that circulated widely in April 2025 exemplify this concern, as they arguably reproduced the distinctive visual style and characteristics of works by the renowned animation studio.
Exceptions and Defences under Singapore Law
Computational Data Analysis Exception
The CA contains an exception for computational data analysis.[12] This exception could be considered to be supportive of AI innovation in Singapore.[13]
Under this exception, both commercial and non-commercial organizations may extract data from copyright-protected works for computational data analysis without obtaining consent from copyright owners, provided certain conditions are met.[14] The CA defines “computational data analysis” to include “using a computer program to identify, extract and analyse information or data from the work.”[15]
To benefit from this exception, developers must ensure:[16]
- The copy is made exclusively for computational data analysis purposes and not used for any other purpose;
- The copy is not supplied to others except for verifying the analysis results;
- The developer has lawful access to the original material from which the copy is made; and
- The original material is not an infringing copy (or if it is, additional requirements must be satisfied).
Importantly, the CA clarifies that “lawful access” does not include circumventing paywalls or violating database terms of use[17] which appears to be an attempt to strike a balance between allowing developers of GenAI models use of copyrighted works while ensuring that such use does not go beyond to take advantage of the work itself.
Fair Use Defence
Users of GenAI tools who obtain AI-generated outputs may potentially rely on the fair use defence[18] under the CA. Singapore courts consider several factors when assessing whether the use of copyright-protected elements constitutes fair use:[19]
- The purpose and character of the use, including whether it is commercial or non-profit educational;
- The nature of the copyrighted work;
- The amount and substantiality of the portion used relative to the whole work; and
- The effect of the use on the potential market for or value of the work.
The fair use analysis is highly context specific. For example, text outputs that critique or comment on literary works may qualify as fair use because a critique typically serves a different purpose than the original work and is not considered to be a substitute in the marketplace. Conversely, AI-generated images that preserve substantial similarities to original artwork or merely collage elements from protected works may not qualify as fair use, particularly if they could serve as market substitutes for the licensed works of the original.
Practical Implications for AI Developers, Users and Rights Owners
The intersection of Singapore’s copyright law with GenAI creates both opportunities and challenges. Developers should carefully document their data acquisition and training methods to ensure compliance with the computational data analysis exception requirements. This includes verifying lawful access to training materials and implementing safeguards against reproducing substantial portions of protected works in AI outputs.
For users of GenAI tools, understanding the potential copyright implications of AI-generated content is essential, particularly for commercial applications. Users should:
- Consider implementing human review processes for AI outputs before publication;
- Be cautious when generating content in styles closely resembling known artists or creators;
- Document transformative elements or novel contributions when building upon AI-generated content; and
- Consider licensing content for commercial use if it is likely that any GenAI output may resemble protected works.
Finally, rights owners should also consider taking practical steps to proactively protect their works. Such steps may include:
- Implementing licencing models for developers to easily adopt to facilitate the use of copyrighted works in AI training datasets;
- Using digital watermarks in digital assets to deter unauthorized reproductions in AI outputs and to facilitate tracing of the same; and
- Establishing proactive enforcement policies and mechanisms that leverage on monitoring and detection tools for checking for non-compliance with existing licensing models or unlawful use.
Conclusion
As GenAI continues its rapid evolution, the CA attempts to balance innovation with protecting creators’ rights. The computational data analysis exception represents a possible position that may facilitate AI developers’ reasonable use of copyrighted works for the development of their GenAI models.
However, it is still unclear how courts will interpret terms such as “substantial similarity” between AI outputs and protected works, the scope of “lawful access” for training data, and the application of fair use to various AI use cases. As these technologies become increasingly sophisticated and widespread, apart from further legislative refinements, it will be necessary for judicial pronouncements to complement such statutory enactments by explaining how they are to apply to the various GenAI use cases.
Organizations developing or utilizing GenAI should stay informed about legal developments in this rapidly evolving field and consider consulting with copyright specialists to navigate the intersection of technology and intellectual property law.
Should you have any queries on this update or generally, please feel free to contact any one of the undersigned.
This Client Update was authored by Jevon Louis (Partner), Ted Teo (Partner) and Kuek Kai Liang (Senior Associate).
Click here to view the article.
[1] Model AI Governance Framework (2020): https://www.pdpc.gov.sg/-/media/Files/PDPC/PDF-Files/Resource-for-Organisation/AI/SGModelAIGovFramework2.pdf
[2] Model AI Governance Framework for Generative AI (2024): https://aiverifyfoundation.sg/wp-content/uploads/2024/05/Model-AI-Governance-Framework-for-Generative-AI-May-2024-1-1.pdf
[3] https://explodingtopics.com/blog/chatgpt-users
[6] Parts 3 and 4 of the CA.
[7] Section 305(1) of the CA.
[8] Section 146 read with sections 41, 49 and 112(a) or 113(a) of the CA.
[9] Section 41(2)(a) of the CA.
[10] Section 41(2)(f) of the CA.
[11] Section 50(1) of the CA.
[12] Sections 243 and 244 of the CA.
[15] Section 243(a) of the CA.
[16] Section 244(2) of the CA.
[17] Illustrations under section 244(d) of the CA.
[18] Sections 190 and 191 of the CA.
[19] Section 191 of the CA.
