Whose Works Does AI Learn From? Indonesia Begins Rewriting Its Copyright Rules

On 10 September 2026, Deputy Minister of Communication and Digital Affairs (Wamenkomdigi) Nezar Patria stated that the Government is drafting a revision of the Copyright Law and examining regulations governing content created using AI. The Government is also paying attention to the use of AI in the journalism industry. Nezar made the statement during a discussion at Atma Jaya Catholic University of Indonesia, Jakarta, and it was reported by ANTARA on the same day.

Generative artificial intelligence (AI) is changing the way people create works. With several prompts, AI can generate written works, images, music, and even videos within seconds. The problem is that technological developments are advancing much more rapidly than the regulations governing them.

Indonesia already has Law Number 28 of 2014 concerning Copyright. However, the law was enacted before generative AI developed to its present state. Accordingly, the Government now needs to adjust the existing legal framework so that it can address new issues arising from the use of AI in the creation and exploitation of works.

So, what needs to be regulated?

Who Is Considered the Author When a Work Is Created by AI?

Law 28/2014 defines an author as a person or several persons who, individually or jointly, produce a work that is distinctive and personal in nature. This means that Indonesian law currently starts from the premise that there is human contribution by the author.

The issue arises when an image, song, or written work is created almost entirely by AI. Is the copyright owned by the user who writes the prompt, the AI developer, or is there no copyright at all?

The revision of the Law needs to establish clear boundaries. One approach that may be considered is to distinguish works generated entirely by machines from works that continue to involve human creative contribution. In this way, persons who genuinely provide creative contributions can have certainty regarding their rights.

May Copyrighted Works Be Used to Train AI?

This is one of the most significant issues.

AI models require very large amounts of data to learn. Such data may include books, articles, photographs, music, videos, and various other works that may be protected by copyright.

Law 28/2014 does not specifically regulate the use of copyrighted works as AI training data. Although the law contains several provisions limiting copyright for certain purposes, Indonesia does not have a “fair use” doctrine equivalent to that of the United States legal system. Accordingly, the term should not be used as though it constituted a general exception under Indonesian law.

The revision needs to determine when the use of works for AI training is permitted, when permission is required, and when authors are entitled to compensation. The Government should also consider opt-out mechanisms, collective licensing, or particular royalty schemes so that authors do not lose the economic value of their works.

This debate is not occurring only in Indonesia. In the United States, for example, The New York Times’ lawsuit against OpenAI and Microsoft remains one of the significant cases concerning whether the use of articles to train AI can be justified under fair use. Developments in the case demonstrate that this issue does not yet have a simple answer.

How Should Journalistic Works Be Protected?

Media organizations face issues from two directions. On the one hand, media organizations need to be able to use or quote information from other sources for reporting purposes. On the other hand, their journalistic works may also be collected and processed by AI systems on a large scale.

Accordingly, the revision of the Copyright Law needs to clarify the boundaries governing the use of journalistic works, including for quotation, reporting, research, and AI training. The rules must provide sufficient protection for media organizations without restricting press freedom and public access to information.

This issue is also related to Presidential Regulation Number 32 of 2024 concerning the Responsibility of Digital Platform Companies to Support Quality Journalism. Article 7 provides that cooperation between digital platforms and press companies may take the form of paid licensing, revenue sharing, sharing of aggregated news-user data, or other forms as agreed upon. However, these provisions were designed for relationships between digital platform companies and press companies and therefore do not automatically constitute a specific framework for AI model developers.

There Needs to Be a Transparency Obligation

The revision also needs to regulate transparency. The public should be able to know if content has been created or manipulated using AI, particularly in relation to journalistic content, advertising, political content, and public information.

On the other hand, the public also needs certain information concerning the sources or categories of data used to train models, provided that this does not disclose trade secrets or sensitive information. This needs to be considered within the framework of the new intellectual property rights regulatory regime.

Who Is Responsible?

The law needs to determine liability when AI generates content that infringes copyright. For example, where a system generates a work that is highly similar to a particular work or is used to produce commercial content without authorization.

Liability need not automatically be imposed on a single party. The revision may distinguish among the roles of model developers, platform providers, and users based on the degree of control and their respective contributions to the infringement.

Ultimately, revising the Copyright Law is not sufficient merely to answer the question, “Who owns an AI-generated work?” The Government needs to regulate the entire chain of issues: who is considered the author, how works may be used to train AI, when permission is required, how authors receive compensation, how AI-generated content is to be labeled, and who is responsible when an infringement occurs.

If the rules focus only on AI outputs, most of the issues will instead remain beyond the reach of the law. Conversely, if the revision is capable of regulating the process from training data through to the use of AI-generated outputs, Indonesia will have an opportunity to build an AI ecosystem that continues to promote innovation without sacrificing authors’ rights.