22 Sep 2025
Intersection of Artificial Intelligence (AI) and Copyright: A Look at the Uncharted Territory of the Law

1. Introduction

The development of Artificial Intelligence (“AI”) in recent years has shown remarkable progress, particularly in the field of generative AI. Generative AI refers to systems capable of creating “new” content, such as text, images, or audio – by transforming one type of data into another (e.g., text to text, text to image, or image to image).[1] This capability is achieved through training on vast datasets, enabling the AI to identify patterns and correlations in existing data and use them, via user prompts, to generate creative outputs. The current state-of-the-art “multimodal” models can even process and synthesize information from across text, images, and audio – further enhancing their versatility.[2] 

With the rise of generative AI, a central legal issue is how it intersects with copyright law, particularly (i) potential infringement when copyrighted works are used for training; and (ii) liability for infringement in AI-generated content. As AI-related litigation grows worldwide (see, for example, Andersen v. Stability AI Ltd. in Northern District of California and Disney & Universal v. Midjourney in Central District of California), the adequacy of existing copyright frameworks to address these challenges has become increasingly pressing.

2. Indonesian Legal Framework on Copyright Protection

Copyright protection in Indonesia is regulated by Law No. 28 of 2014 on Copyright (“Copyright Law”). It was formulated with reference to international instruments ratified by Indonesia, including the Berne Convention for the Protection of Artistic and Literary Works, adopted through the Presidential Decree No. 18 of 1997 (“Berne Convention”).

Under the Copyright Law, protection covers works of science, art and literature resulting from inspiration, ability, thought, imagination, dexterity, skill or expertise expressed in tangible form. The Copyright Law further provides an exhaustive list of nineteen categories of copyrightable works.

The Copyright Law adopts the concept of “derivative works” in line with the Berne Convention. Under Article 2(3) and Article 12 of the Berne Convention, derivative works, such as translations, adaptations, musical arrangements, and other alterations of a literary or artistic work, are considered copyrightable and reserved as an exclusive right of an Author. In the same vein, Article 40 (1)(n) and Article 9(1)(d) of the Copyright Law list translations, interpretations, alterations, anthologies, databases, adaptations, arrangement, modifications and other transformed works as copyrightable, with their creation likewise reserved to the Author. In essence, both regimes recognize derivative works as creations based on preexisting copyrighted material.

While the law recognizes a broad scope of copyrightable works, it also provides exceptions commonly referred to as “Fair Use” or “Fair Dealing”. Under the Copyright Law, three cumulative conditions for the Fair Use exception must be met: (i) the use and purpose of use must fall within the defined categories, such as education, research, scientific, or national interests, or free public performances; (ii) credit must be given to the Author by citing the source; and (iii) the use must not harm the Author’s reasonable interests – meaning that even non-commercial use may constitute infringement if it undermines the Author’s economic benefit. 

3. Use of Copyrighted Works for AI Training

The key legal issue at the intersection of AI and copyright is whether using copyrighted works for AI training constitutes infringement. Generative AI relies on vast datasets to produce “new” content, and if copyrighted works are included without authorization, infringement may arise. Such outputs can be understood as derivative works—creation of which is the exclusive right of the Author—thus requiring the Author’s consent.

In that case, additionally, it is difficult to argue that Fair Use under the Copyright Law applies. Most generative AI systems are commercialized, and even in non-commercial settings the exception is narrow: it requires that the Author’s economic rights not be harmed and that the original source be cited. Given the scale of AI training datasets and the error-prone nature of referencing, these conditions are rarely met in practice. Accordingly, it is unlikely that AI training involving copyrighted works falls within the Fair Use exception.

4.  Issues of Liability in Cases of Copyright Infringement in AI-generated Content

Another legal risk concerns liability when AI-generated content produced from a user’s prompt infringes a copyrighted work. This is complex, as the law does not clearly define how liability should be allocated between AI companies and users.

At present, the rules on how AI shall be operated only exist in Minister of Communication and Information Circular Letter No. 9 of 2023 on Ethics of Artificial Intelligence (“MOCI Circular 9/2023”). The MOCI Circular 9/2023 provides general ethical guidelines—covering values such as safety, transparency, privacy, and intellectual property rights. However, it does not establish concrete legal rules on how AI companies should prevent copyright infringement or what obligations users must follow when operating generative AI.

In the absence of such provisions, liability for copyright infringement in AI-generated outputs will be determined case by case, leaving attribution between companies and users unpredictable.

5. Conclusion

Given the current regulatory landscape, AI companies in Indonesia must tread carefully in both the development and operation of generative AI. Training processes and practical uses must be managed to avoid copyright infringement. Clear preventive measures and good-faith compliance efforts to reduce legal risks must be taken as existing laws make AI companies highly vulnerable to copyright claims, leaving little room for defense. Looking ahead, however, traditional copyright principles, such as what counts as a derivative work and how far Fair Use extends, are likely to evolve, and companies will need to be ready to adapt.

By partner Ayik C. Gunadi (agunadi@abnrlaw.com), associate Evelyn Irmea (esinisuka@abnrlaw.com), and trainee associate Jyestha Herawanto.

This ABNR News and its contents are intended solely to provide a general overview, for informational purposes, of selected recent developments in Indonesian law. They do not constitute legal advice and should not be relied upon as such. Accordingly, ABNR accepts no liability of any kind in respect of any statement, opinion, view, error, or omission that may be contained in this legal update. In all circumstances, you are strongly advised to consult a licensed Indonesian legal practitioner before taking any action that could adversely affect your rights and obligations under Indonesian law.