Artificial Intelligence and Copyright Law: Reconciling Authorship, Ownership, and Innovation in India

Introduction

Generative artificial intelligence (AI) includes systems that can produce text‚ images‚ video‚ or music in a way that challenges established conventions of copyright law‚ including the notion that a work must be of human origin in order to enjoy copyright protection․ In 2023‚ generative AI systems such as large language models and text-to-image generators produced output indistinguishable from that of a human‚ creating legal uncertainty.[1] The issue is further complicated in India by the Copyright Act‚ 1957‚ which does not expressly provide for non-human authors․ The contradiction rests on the fact that copyright law seeks to protect works of authorship‚ whereas AI systems merely process the input of large datasets and generate content in ways that are distinct from human authorship.[2]

Likewise‚ Indian courts have repeatedly stated that originality must entail the “skill and judgement” of the author‚ and pure mechanical labour cannot be copyrighted.[3] Likewise‚ Indian courts have repeatedly stated that originality must entail the “skill and judgement” of the author‚ and pure mechanical labour cannot be copyrighted․

After showing that the Indian copyright law framework is structurally incapable of dealing with AI creativity‚ and that piecemeal judicial interpretation will not be able to adequately address the tension between authorship and automation‚ the article argues that India is best positioned to enact either a legislative amendment recognizing some form of limited attribution of AI authorship or a sui generis regime for protecting AI-generated works․ The article begins with an overview of the Indian copyright law scheme‚ followed by the judiciary’s interpretation of the originality standard‚ the evaluation of AI-generated works‚ and a consideration of the international standards for reform․

Legal Framework Governing Copyright in India

In India, the main law for copyright protection is the Copyright Act, 1957. It gives exclusive rights for original literary, dramatic, musical, and artistic works.[4] According to Section 13 of the Act, only original works that are fixed in material form shall enjoy copyright, thus incorporating originality as a threshold requirement.[5] Section 2(d) of the Copyright Act exhaustively defines “author” in relation to different classes of works, and, in all of these classes, authorship is, by default, attributable only to a human or a legal entity (company/firm) acting through a human agency.[6] Whereas patent law has already seen some commentary on the topic of inventorship in the context of AI systems, the Indian copyright statute does not contemplate non-human authorship of works. The statutory silence on machine-generated works, creates interpretive uncertainty where AI systems autonomously (without human intervention) generate output.[7]

Additionally, the Act under Section 52 provides a narrow list of exceptions available under the doctrine of fair dealing which allows for usage of copyrighted works for the purpose of private study or research, criticism and reporting.[8] But these exceptions were not created for machine learning or computational training methods, so it is legally unclear if they apply to AI training datasets.

Judicial Interpretation of Originality and Authorship

Indian courts have regularly said that to be original, something needs to have more than just hard work or money put in. It should also have skill, creativity and mental effort. In Eastern Book Company v. D.B. Modak, the Court did not accept the old “sweat of the brow” principle and said original works should have “at least some creativity”.[9] This decision significantly raised the threshold for copyright protection in India.

Previously, the Court, in R.G. Anand v. Deluxe Films had reiterated the general principle that copyright does not subsist in ideas, but merely expression, and consequently, the requirement that the protectable work must be an outcome of skill and labour and not mechanical reproduction.  In the same vein, in University of London Press Ltd. v. University Tutorial Press Ltd. has interpreted originality to mean an exercise of skill and judgement apart from copy.[10]

All these legal rules focus on a human-centered idea of being an author. But they don’t talk about cases where AI systems create something on their own. These systems use probabilities to analyse information from datasets made by people. The law doesn’t make clear if creations with little human help can get protection.

III. AI-Generated Works and the Problem of Authorship

Indian copyright law faces a basic problem with AI-generated works challenging its traditional authorial concept. Section 2(d) implicitly requires human authors and, therefore, AI systems cannot currently be legal authors.[11] This leads to the question: is it the programmer, user, or platform provider who should be regarded as an author of the AI-generated outputs?

Attributing authorship to the human who provides prompts or instructions to the AI system is one possible approach. Yet this is problematic when AI outputs are highly autonomous and unpredictable–weakening the causal connection between human inputs and creative expression.[12]

Furthermore, assigning authorship to developers may also be misplaced, as their contribution is collateral to the generated work and generic, not specific. This, in turn, creates a legal vacuum in which AI-generated works fall outside copyright protection altogether. This could stifle innovation and investment in generative technologies.

Training Data, Infringement, and Fair Dealing

The use of copyrighted materials to train AI systems also raises concerns, albeit separate ones. Given the vast data requirements, datasets can contain copyrighted images, books, and articles and raise infringement issues.[13] Litigation over whether such training constitutes fair use has already emerged in jurisdictions such as the United States. However, there is nothing under Indian law.

Fair dealing exceptions are outlined in Section 52 of the Copyright Act, but its application on machine learning remains ambiguous.[14] AI training copies and processes whole works for computation, not for human reading, which is not the same as typical fair dealing.

Some scholars have argued that expanding the scope of the fair dealing principle to include AI training without a statutory amendment would risk original copyright holders’ rights. At the same time, restricting it may hamper our technological progress and India’s participation in the global AI economy. This dilemma highlights the policy conundrum between innovation and protection.

Comparative Perspectives on AI and Copyright

Different regions in the world have taken different paths when it comes to AI-generated works. The Copyright, Designs and Patents Act 1988 has granted authorship to the person who made the “necessary arrangements” for creation under Section 9(3) in the United Kingdom.[15]

In contrast, the United States has a very strict human authorship requirement, recently reiterated by the U.S. Copyright Office in guidance rejecting copyright protection for AI-generated works only.[16] The European Union has also stressed that human intellectual creation is required for copyright protection.

By considering comparative approaches, it is clear that there is no global consensus. India does not have a legislative equivalent to the UK’s computer-generated works provision, thereby leaving AI outputs in a legal grey zone.

Critical Evaluation and Doctrinal Gaps

When AI-generated works are considered, the Indian copyright framework is found to have significant structural limitations. Specifically, the human authorship requirement, while doctrinally coherent from the perspective of a classical theory of copyright, is unable to respond to the reality of autonomous machine creativity. It is stuck between the competing tensions of either an overextension of copyright to non-human outputs, or of non-application of copyright to a wide category of creative outputs.

Judicial interpretation alone cannot solve this tension, as judges in India are limited by statutory definitions in the Copyright Act, 1957. While the Supreme Court in Eastern Book Company v. D.B. Modak evolved a new standard of originality, statutory interpretation on this front does not account for fundamental questions of non-human authorship.[17]

Moreover, AI developers and creative content owners have no idea about their liability, if any, for infringement use of copyright material in training AI. Until Parliament intervenes, India will lag behind in creating an AI-IP regime.

VII. Policy Recommendations

Legislative reform to expressly recognize AI-generated works under the Copyright Act, 1957, should be contemplated by India. A provision for deeming the person causing the work to be created as the author could be introduced. Alternatively, India could consider affording sui generis protection to generative AI works, in the manner of database protection in the European Union.

Further, express statutory guidance regarding the use of copyrighted materials for training of generative AI systems, recognising fair use principles without impacting upon the rights of creators, should be considered. Regulatory clarity will ensure that innovation in the space of AI does not come at the expense of protection.

Conclusion

Basic principles of copyright law are challenged by artificial intelligence as it disrupts conventional perspectives on authorship and creativity. Current Indian legal framework is tailored to and is insufficient in dealing with the predominant complexities associated with AI-generated works as it is based on a settled notion of human authorship. Judicial precedents have defined the concept of originality in certain cases, however, is yet to account for machine creativity.

The above article has established that although AI has proved to be essential for innovation, it has also proven to be doctrinally incompatible with the provisions laid down in the Copyrights Act, 1957. Select comparative law, can be inferred that legislative intervention is the only solution to fixing such doctrinal inconclusiveness. Therefore, India needs to adopt a futuristic regulatory approach which ensures a balance between technological advancements and safeguarding intellectual property rights.

Reference(S):

Cases

Eastern Book Co. v. D.B. Modak, (2008) 1 S.C.C. 1 (India).

G. Anand v. Deluxe Films, (1978) 4 S.C.C. 118 (India).

University of London Press Ltd. v. University Tutorial Press Ltd., [1916] 2 Ch. 601 (Eng).

Legislation

Copyright Act, No. 14 of 1957 (India).

Copyright, Designs and Patents Act 1988, c. 48 (U.K.).

Secondary Sources

S. Copyright Office, Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence (2023).

World Intellectual Property Organization, Artificial Intelligence and Intellectual Property Policy Report (2023).

World Intellectual Property Organization, Artificial Intelligence and Intellectual Property Discussion Paper (2023).

World Intellectual Property Organization, The WIPO Conversation on Intellectual Property (IP) and Artificial Intelligence: Ninth Session – Training the Machines: Bytes, Rights and the Copyright Conundrum (2024)

[1] World Intellectual Property Organization, Artificial Intelligence and Intellectual Property Policy Report (2023).

[2] Copyright Act No.14 of 1957, § 2(d) (India).

[3] Eastern Book Co. v D.B. Modak, (2008) 1 S.C.C. 1 (India).

[4] Copyright Act, 1957.

[5] Id. § 13.

[6] Id. § 2(d).

[7] World Intellectual Property Organization, Generative Artificial Intelligence and Copyright (2023).

[8] Copyright Act, 1957, § 52.

[9] Eastern Book Co. v. D.B. Modak, (2008) 1 S.C.C. 1.

[10] University of London Press Ltd. V. University Tutorial Press Ltd., (1916) 2 Ch. 601 (Eng.).

[11] Copyright Act, 1957, § 2(d).

[12] World Intellectual Property Organization, Artificial Intelligence and Intellectual Property Policy Report, supra note 1.

[13]  World Intellectual Property Organization, The WIPO Conversation on Intellectual Property (IP) and Artificial Intelligence: Ninth Session – Training the Machines: Bytes, Rights and the Copyright Conundrum (2024).

[14] Copyright Act, 1957, § 52.

[15] Copyright, Designs and Patents Act 1988, c. 48, § 9(3) (UK).

[16] U.S. Copyright Office, Copyright Registration Guidance: Works Containing AI-Generated Material (2023).

[17] Eastern Book Co. v. D.B. Modak, (2008) 1 S.C.C. 1.