Generative AI and Copyright Law in India: Emerging Challenges and the Need for Reforms

Author: Parakh Saxena, City Law College

---

Abstract 

Generative artificial intelligence (GenAI) questions existing Indian copyright laws, as it raises issues regarding authorship and ownership, AI training, infringement and liability. This doctrinal research assesses the Copyright Act of 1957 in addition to the case law, policy documents and comparative we studies in the United States of America, the United Kingdom and European Union for suggesting required legal reforms.

Moreover the authors suggest the implementation of a human creativity control system as a guiding principle to decide the level of human creative contribution in the works of AI. Under this approach, copyright protection would be provided where a human exercises meaningful creative control while making the  prompts and further in the selection modification finalisation of the outputs. Such a framework would enable courts and policy makers to differentiate between work created by humans and content produced by AI, thus, promoting legal certainty while maintaining the fundamental objective of the copyright law to protect human intellectual creativity.

The paper concludes that the current copyright laws lay a foundation for protecting human creativity however, needs legislative clarifications and policy reforms to efficiently balance innovation creators right and public interest in the age of generative AI.

Keywords– Generative AI, Copyright Law, Ownership and Authorship, Artificial Intelligence, Copyright Reforms in India.

Introduction 

The sudden advancement of generative AI has brought copyright law to a critical turning point. One of the major event that mark the concern was when The New York Times filed suit against open AI (alongside Microsoft) alleging that the defendant unlawfully using the times work to create artificial intelligence product that compete with it and threatens the times ability to provide the nature of service they are known for providing.  The plaintiff pointed out that defendant’s generative AI tools rely on “large language models” (LLM) that were built by copying and using millions of The Times copyrighted news articles, indepth investigations, opinion pieces, reviews, how-to-guides and more.

Now the question arises whether the use of copyrighted content even for training of AI is legal? whether such use of data is infringing the copyrights of the plaintiff? And the consent or authorisation does not have a role? 

As the Generative AI continuous to grow and shape creative industries these questions are no longer confined to a single jurisdiction however, have become a topic of debate all over the globe. For countries like India where copyright laws were formed way before the emergence of autonomous AI system, the debate highlights the pressing and necessity to evaluate if current laws are equipped to address the distinct challenges presented by generative AI.

Generative AI refers to an system that is capable of producing original content including text, images, music, software code, audio and videos by learning pattern from vast data sets. This is different from traditional AI system which primarily analysis data or automate predefined tasks, generated AI studies patterns from the data and create original content which closely resembles human expression.

The use of generative AI across different discipline like education, medicine, reporting, growing business, writings software programs and providing legal services has changed the way creative and intellectual work are created and distributed. But this technological advancement has pointed out a number of important gaps in traditional copyright law. The section 13 of copyright act, 1957 provide the works in which the copyright subsists. They are: original literary, dramatic, musical and artistic works, cinematograph films and sound recordings.            

It clearly mentions “original work” that is - creativity stems out from human and hence must be protected.

Consequently copyright law provides limited insight into matter such as the legal status of copyright protected material when used in training AI systems, copyright ownership and attribution in terms of work generated by AI and questions of liability for violating copyright law when works of created with the involvement of AI.

With the growth of generative AI comes considerable confusion of copyright laws, namely in India, where the Copyright Act of 1957 predates AI technology. The law does not specify any information about copyright status with regard to AI-generated works or any processes for using copyrighted works to teach AI algorithms. This brings about ambiguity regarding authorship, ownership, infringement, and liability. Accordingly, this research paper looks at these questions, reviews global experiences, and suggests ways for updating copyright laws in India.

Literature Review 

The rise of Generative Artificial Intelligence (GenAI) has caused a shift in copyright law. With the ability to create text, images, music, software code, and other creative works with little human involvement, significant legal concerns about the relevance of copyright laws are raised. This is because even though basic copyright principles are still the foundation of the law, they are not enough according to experts to solve the problems faced in the generation of AI-generated materials.

At the forefront of the relationship between Artificial Intelligence (AI) and the field of Intellectual Property (IP) is the World Intellectual Property Organization (WIPO). The organization understands that AI has influenced creativity and brought about legal issues such as the use of copyrighted materials for AI training, who owns AI-created products, and who should be awarded credit for creativity. Instead of presenting a common uniform solution for the whole world, WIPO has encouraged its members to come up with reasonable and balanced solutions in the area of IP law.

Comparative studies indicate that various jurisdictions have reacted differently to the issue of Artificial Intelligence and copyright. In the U.S., discussions are held about fair use doctrine and disputes on the use of copyrighted work for AI model training. In the European Union, Directive (EU) 2019/790 has been introduced by the EU to regulate the matter of copyright in the Digital Single Market, where provisions are made for text and data mining and innovations in copyright. In the U.K., Copyright, Designs and Patents Act 1988 accepts computer-generated works, although the legal validity of its provisions is questionable due to the developments of Generative AI models in recent times.

The Copyright Act of 1957 in India lacks any direct clauses regarding the concept of AI-produced works or the concept of copyrighted content in AI training. Since the Copyright Act was passed even before the invention of Generative AI, some pertinent issues such as authorship, ownership, infringement, and responsibility cannot be answered by this legislation. This implies that court cases will be based on traditional copyright principles regarding the changing technologies. It is worth mentioning that some landmark cases such as Eastern Book Company v. D.B. Modak, which emphasizes the concept of originality based on the level of skill, judgment, and creativity, and R.G. Anand v. M/S. Delux Films, which defines the difference between thoughts and expressions, remained important, although they were ruled even before the emergence of AI technologies.

It has been universally noted in the literature that India lacks a comprehensive legislative framework for AI and copyright law. Following up on these studies, this research critically examines the Indian copyright law system in comparison to those in place in the United States, the United Kingdom, and the European Union. The need for reforms in the law is discussed, and the concept of creative control by humans is developed in order to create a balance in handling technological advancement.

Research Methodology

This research uses a doctrinal legal research methodology to study the present legal rules, statutes and judicial precedents related to Generative AI (GenAI) and copyright. The doctrinal approach is suitable for the research because the doctrinal research is based on the examination and analysis of the present laws. The provisions of the Copyright Act of 1957 and the related judicial precedents are studied in this research, including Chief of Police v. St. John’s Ambulance Brigade and State v. Khari Bastey. The international laws and documents published by the World Intellectual Property Organisation (WIPO) are also reviewed to understand the opinion of the international community concerning AI and intellectual property.

The research further makes use of comparative legal analysis in order to study how the USA, UK and the EU regulate copyright issues related to generative AI. The study looks at the application of fair use principle in the USA copyright law, the provision related to the computer-generated works in the UK Copyright, Designs and Patents Act of 1988, and the Directive (EU) 2019/790 on Copyright in the Digital Single Market. By comparing these laws with Indian laws, the study highlights the strengths and weaknesses of the Indian position and the various reforms that could be introduced in India.

The research is based on primary as well as secondary sources. The primary sources consist of statutory provisions, judicial decisions, and applicable laws, namely the Copyright Act of 1957, the Copyright, Designs and Patents Act of 1988 (United Kingdom), and Directive (EU) 2019/790 on copyright in the digital single market. On the other hand, the secondary sources are in the form of books, scholarly articles, commentaries, reports, and WIPO material that deal with the topic and relates to the changing nature of generative artificial intelligence with copyright law. The analysis of these sources has been carried out using doctrinal and comparative methods. The study assesses whether the law is sufficient and suggests reforms.

Generative Artificial intelligence And Thier Training 

AI techniques used in GenAI systems are based on algorithms capable of finding patterns and links within huge volumes of texts, books, articles, visual works, audio and video clips and computer programs. In contrast to traditional software, which operates according to programming instructions, GenAI does not implement precomputed programs but instead learns from the input data and uses it to generate new information according to the requests of its users. The training process typically consists of copying files, tokenizing them, making statistical evaluations and adjusting parameters of the model. Even if the output is not presented exactly same as in the original work, there are some challenges regarding the use of copyrighted materials in GenAI.

The main question concerns the possibility of using copyrighted works for training purposes while being protected by some exceptions granted by law. 

The opposite opinions can be illustrated by the well-known series of court cases.   In The New York Times Company v. Microsoft Corporation & OpenAI Inc., The New York Times accused Microsoft of having used millions of its news articles without permission to create a large language model. According to the claim, ChatGPT has the potential to produce outputs similar to its articles that would undermine subscription income and possibilities for licensing. The mentioned lawsuit raised the important issues connected with the possibility of unauthorized training, the possibility for market substitutions, and the operation of fair use rule.

The same issues were voiced by other  authors. In Tremblay v. OpenAI, Inc., Paul Tremblay and Mona Awad claimed that OpenAI exploited Tremblay and Awad’s copyrighted text without their permission to develop ChatGPT. In the same way,  the case of Silverman v. OpenAI, Inc. Consists of claims of writers like Sarah Silverman that their books were copied for the purpose of the training of the AI, which constitutes a breach of authors’ rights guaranteed under the United States copyright law.

In comparison to the U.S., it can be seen that  the European Union has introduced various provisions concerning text and data mining through the provisions under the DSM Directive, which contain the exceptions to the regulation that takes account of the economic rights of copyright owners. Similar to the position of the United States and the EU, it is important to note that in the United Kingdom, there are provisions under the  Copyright, Designs, and Patents Act 1988 which allows different forms of text and data mining for purposes of non-commercial research. Consequently, there is value in undertaking a comparison to determine whether or not the copyright system of India has effectively addressed the issue of AI training through the provisions of the Copyright Act.

Authorship, Ownership, Infringement and Liability 

The emergence of Generative AI (GenAI) has tested the concept of copyright by complicating the topic of authorship in AI-generated works. Copyright law typically applies under the assumption that a human being has created a work through creative and intellectual efforts. These uncertainties also extend to issues of copyright infringement and liability, particularly where AI systems are trained on copyrighted materials without the permission of rights holders.

Authorship 

According to the Copyright Act, the notion of authorship is closely linked with human intellect. Section 2(d) defines the author for different types of works but does not mention authorship by AI. Thus, presently Indian law does not consider AI as a legal author. Even though Section 2(d)(vi) states that in case of computer-generated works, the author is the one who creates the work, this provision was introduced before modern generative AI developed and therefore it is unclear how it applies to autonomous AI-generated works.

Ownership 

Copyright ownership for works generated by machines creates further concerns. “When a machine produces a work, various parties can claim ownership, including the developer of the machine, the person using the machine, or the owner of the data or corpus used to train the model. As a result, current laws do not specify who should or should not own the work. Therefore, many jurisdictions have opted not to give copyright protection to works that do not involve human creativity.”

Infringement 

Generally, most robust language models, as well as image generator systems use huge amounts of materials for their training that could contain copyrighted books, newspapers, pictures, and others. Copyright holders claim that reproduction of such materials for training purposes of machines is illegal since it violates their properties. However, AI companies say that the usage of such works is a transformational activity and therefore, it should be allowed according to different theories of legality in use: fair use, text mining, and so on, depending on the current legislation.

Liability 

The various interests involved have led to a considerable amount of litigation across the globe. In one case, the complainants claimed that the utilisation of articles of newspapers was done without proper approval from the copyright owners. The same issue arose in another matter, whereby visual creators contested the infringing usage of images for training purposes, thus emphasising the issue of copyright infringement during the training process.

Again, the question of liability is left unanswered: who shall be liable for some kind of infringement in relation to AI-generated creations? Is it the AI creator? The service provider? Or the end-user? Current copyright laws are targeted towards humans and do not give any guidance on how to deal with independent AI creations. The current legal situation in India does not provide any case law or legal acts in this regard as well.

Analysis of the United States, United Kingdom and European Union

Generative artificial intelligence problems have resulted in several legal issues related to copyright laws in different countries especially regarding the legality of using a copyrighted work for the purposes of developing artificial intelligence and protecting the outputs generated through artificial intelligence. 

In spite of the fact that all three jurisdictions, namely the United States, the United Kingdom and the European Union are trying to create a balance between innovation and the rights of authors when it comes to this kind of issue, they adopt completely different approaches in legal terms.

The copyright system in the U.S. is based on the principle of fair use as set out in section 107. If it is legal for AI creators to use copyrighted items depends on the four-factor fair use test which is currently being dealt with by courts in the U.S. in such cases as The New York Times Co. v. Microsoft Corp. and OpenAI. The U.S. Copyright Office has confirmed that only works with enough human input are protected by copyright law.

In comparison, the United Kingdom operates under a narrower set of laws. Specifically, section 29A of the Copyright, Designs and Patents Act of 1988 states that text and data mining can only take place for research purposes which do not involve profit, and only by individuals who have lawful access to the copyrighted data.

Although the UK Government once thought about expanding this exception to cover AI development, their proposal was met with rejection from the writers, publishers, and creative industries.

“The European Union has introduced an organized regulatory model in the form of the Directive (EU) 2019/790 which deals with copyright in the Digital Single Market. The articles 3 and 4 in this directive state that text and data mining is allowed for research institutions and others unless the copyright holder has opted out.” This framework seeks to promote technological innovation whilst ensuring adequate protection of copyright.

In contrast to the jurisdictions, India has no law regarding AI training datasets or AI-generated materials. There are no legal norms about intellectual rights and use of copyrighted materials in the development of AI. Therefore, a framework is suggested that will bring balance between AI innovation and rights of copyright owners.

Comparative Analysis of Copyright Protection for Generative AI: United States, United Kingdom, European Union With India

In the United States, laws regulating AI are based on what is known as the doctrine of fair use. Instead of trying to create regulations specifically for AI, courts in the U.S. determine the legality of training AI systems using copyrighted material by applying the standards of fair use.

The UK has an even tighter set of legislative restrictions. Under section 29A of the Copyright, Designs and Patents Act 1988, text and data mining is only permitted for circumstantial research carried out by users with legal access to copyrighted materials. The UK government has chosen to uphold the existing legislation and therefore has not extended the exception to apply to commercial AI developments, thus taking a stronger stand in favour of copyright holders.

“The EU has made some changes in the legislation by adopting Directive (EU) 2019/790 concerning copyright in the Digital Single Market whereby it made provision for exceptions with respect to text and data mining. This legislative framework provides for mandatory exceptions in Articles 3 and 4 on text and data mining that allows a copyright owner to have an opt-out option.” 

This model seeks to strike a balance between technological innovation and the economic interests of authors and other rights holders.

On the contrary, the Copyright Act, 1957 still is interpreted using the classical doctrine of original and authorship. In Eastern Book Company v. D.B. Modak, the Supreme Court emphasized that copyright protection requires a minimum degree of creativity and intellectual effort. The ruling did not deal with artificial intelligence, however, it is likely that the principles set by this case will influence future rulings regarding cases related to AI generated content. 

Indian Copyright Law and the Need for Reform

There is no legal framework that comprehensively governs the connection between generative artificial intelligence and copyright law in India. The Copyright Act, 1957 of India is the primary statute governing copyright law but was enacted prior to the advent of contemporary AI technologies, thus the issues of AI works, usage of copyright materials for building AI, and responsibilities of AI infringers remain unregulated in Indian law. The biggest problem in the Indian regulatory framework is whether there is a copyright infringement on the use of copyrighted material for training AI. Copyright gives the right to the owners of copyrighted works to reproduce, communicate, adapt, and distribute the works.  

A lot of copyrighted material will be reproduced in a normal AI training process. Thus, there can be possible copyright infringement on the use of the copyrighted works unless there is a statutory exemption. Indian law does not provide for any AI-specific exemption allowing text and data mining for any commercial AI purposes.

Another issue that is yet to be resolved is the issue regarding the rights with respect to authorship of  “computer generated works”. Section 2(d)(vi) of the Copyright Act defines the author of a computer-generated work as the individual who causes the work to come into being. However, this provision was enacted for traditional computer generated work and does not solve the issue of authorship with respect to works that are produced by an autonomous AI system with minimal human intervention. There is stil ambiguity whether copyright belongs to the user or the developer while confusion exists as to whether AI-generated works are copyrightable.

Indian courts have held the consistent view that copyright protects originality in expressions instead of ideas. In 1977, the Hon’ble Supreme Court of India speaking at the standard of copyright of originality stated the requirement of “modicum of creativity”, which means exercising skill and judgment of the creator instead of mere effort.

In a similar fashion, in , the Supreme Court re-confirmed that copyright protects the expression of the ideas and not the ideas as such. This difference could function to help courts in deciding whether AI-generated results simply copy the existing expression that is protected by the copyright law or give rise to a new expression that could be considered original.

Discussion 

The conducted research proves that generative artificial intelligence has contradicted classical principles of copyright law, as it has made the border between concepts of authorship, originality, ownership, and infringement indistinct. Existing laws in the field of copyright were established under the assumption that novels are written by human intellect. However, nowadays, AI systems can produce creative works without much human intervention, thus causing ambiguity concerning copyright protection and application of copyright material for AI training. 

Looking at the comparative approach, it can be said that various jurisdictions have adopted different stances regarding copyright issues related to AI. While the US uses the court’s interpretations and fair use doctrine to address the issue, the European Union has certain rules of text and data mining laid out in Directive (EU) 2019/790. While the UK recognizes computer-generated works under the Copyright Designs and Patents Act of 1988, India still does not have the relevant law. Even though the Indian courts have laid down certain copyright principles, like in case of Eastern Book Company v. D.B. Modak and R.G. Anand v. M/S. Delux Films, those precedents do not address the challenges that AI-related copyright raises. Hence, this study recommends adopting a human creative control approach, where the copyright will be dependent on the degree of human creativity. This study also proposes to adopt rules regarding AI training, licensing, and responsibility so that the balanced system of copyright is created in India.

Conclusion 

In this light, it is imperative to have legislative reforms that would establish a definitive legal position on the issue of AI training, the ownership status of AI-generated creations, the factors determining the liability position of AI technologies, and the implementation of transparent licensing procedures for all stakeholders. Most importantly, India should think about enacting a “Human creativity control System” as a guiding principle that would facilitate human creative control as a way of establishing whether copyright applies to AI-generated works or not. It should be noted that the establishment of copyright would be possible in situations where a human being has made meaningful creative contribution in shaping the works. There should be an understanding that mere involvement of AI, as such, in the creation of something cannot be taken to imply the existence of authorship, because the person who is involved in the process must have had a significant say in the manner that that particular work was put together and this should not be determined merely by the involvement of an AI tool.

Moreover, the regulatory framework must promote more transparency in relation to the processes of AI technology advancement through adequate disclosure of the use of copyrighted property for AI training while protecting from misuse of legally protected trade secrets and valuable business information. It is essential to find a balanced approach in the licensing of AI training based on copyrights, which will make sure that copyrights owners are equally acknowledged and financially rewarded while making technology development possible. The implementation of such reforms will increase legal certainty and decrease chances for copyright conflicts between authors, audience, AI engineers, and courts.

References

Cases

• Andersen v. Stability AI Ltd., No. 3:23-cv-00201 (N.D. Cal. Jan. 13, 2023).

• Eastern Book Company v. D.B. Modak, (2008) 1 SCC 1 (India).

• R.G. Anand v. M/S. Delux Films, (1978) 4 SCC 118 (India).

• Silverman v. OpenAI, Inc., No. 3:23-cv-03416 (N.D. Cal. filed July 7, 2023).

• The New York Times Co. v. Microsoft Corp. & OpenAI, Inc., No. 1:23-cv-11195 (S.D.N.Y. Dec. 27, 2023).

• Tremblay v. OpenAI, Inc., No. 3:23-cv-03223 (N.D. Cal. filed June 28, 2023).

Statutes

• Copyright Act, 1957, §§ 2(d), 13 (India).

• Copyright, Designs and Patents Act 1988, c. 48, § 29A (UK).

• 17 U.S.C. § 107 (2024).

International and Regional Instruments

• Directive (EU) 2019/790 of the European Parliament and of the Council of 17 Apr. 2019 on Copyright and Related Rights in the Digital Single Market and Amending Directives 96/9/EC and 2001/29/EC, arts. 3–4, 2019 O.J. (L 130) 92.

Government Reports and Official Materials

• UK Intellectual Property Office, Government Response to the AI and Intellectual Property Consultation (Feb. 2023).

• United States Copyright Office, Copyright and Artificial Intelligence, Part 2: Copyrightability 1–8 (Jan. 2025).

• U.S. Copyright Office, Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence, 88 Fed. Reg. 16,190, 16,191–93 (Mar. 16, 2023).

• World Intellectual Property Organization, Revised Issues Paper on Intellectual Property Policy and Artificial Intelligence, WIPO Doc. WIPO/IP/AI/2/GE/20/1 Rev., ¶¶ 31–48 (May 29, 2020).

• World Intellectual Property Organization, Revised Issues Paper on Intellectual Property Policy and Artificial Intelligence, WIPO Doc. WIPO/IP/AI/2/GE/20/1 Rev., ¶¶ 68–80 (May 29, 2020).