Abstract
Generative artificial intelligence (AI) can produce text, images, music and other creative material with varying levels of human involvement. This development raises difficult questions for copyright law, particularly authorship, originality, ownership and infringement. India’s Copyright Act, 1957 addresses computer-generated works in Section 2(d)(vi), which identifies the author as the person who causes the work to be created. Yet the provision does not expressly explain how it applies when a user supplies prompts, a developer builds the system and an AI model generates the final expression.
This paper uses doctrinal legal research to examine the Indian statutory framework, the originality principles recognised in Eastern Book Company v. D.B. Modak, and the Copyright Office’s 2026 decision concerning the AI system DABUS. It distinguishes copyright in an output from questions concerning the use of protected works to train AI systems. The paper argues that existing law provides a starting point but needs clearer guidance on attributing authorship to legally recognised persons and identifying protectable human contribution. It recommends a contribution-sensitive approach, proportionate transparency, careful contractual drafting and balanced treatment of creators’ rights and technological innovation.
1. Introduction
Generative artificial intelligence has become a practical tool in writing, illustration, music, design and other creative fields. A user may produce a work by entering a short prompt, repeatedly refining instructions, selecting among many outputs or substantially editing the material generated by a system. These different levels of involvement make it difficult to apply copyright rules developed around human creators.
Copyright law protects qualifying expression rather than ideas in the abstract. In India, the Copyright Act, 1957, is the principal legislation. Section 2(d)(vi) expressly addresses computer-generated literary, dramatic, musical and artistic works by identifying the author as the person who causes the work to be created. [1] This wording is relevant to generative AI, but its application raises questions about the role of the user, developer and other participants in the creative process.
The main research question is whether the current Indian framework provides sufficient clarity to determine authorship and copyright protection for AI-generated works. The paper has three objectives: to interpret the statutory framework, to examine originality and authorship in AI-assisted creation, and to recommend practical legal clarification. It adopts a doctrinal method based on legislation, judicial decisions, administrative materials and official publications. It does not claim to rely on interviews, surveys or empirical data.
2. Indian Legal Framework
2.1 Computer-Generated Works and Copyright Subsistence
Section 13 of the Copyright Act identifies categories of works in which copyright may subsist, including original literary, dramatic, musical and artistic works, cinematograph films and sound recordings. Section 2(d)(vi) addresses authorship for computer-generated works in the specified categories. The provision therefore offers a statutory starting point, but it does not specifically define generative AI or set out a detailed test for determining the human contribution required in every situation. [1]
Section 14 sets out the exclusive rights associated with copyright. Section 17 generally deals with the first owner of copyright, subject to statutory exceptions. Sections 18 and 30 address assignment and licensing. These provisions matter when AI-assisted material is created in employment, under commission or for commercial use. However, a contract allocating rights cannot by itself establish that copyright subsists in material that does not satisfy the law’s requirements.
2.2 Infringement and Exceptions
Sections 51 and 52 address infringement and specified acts that do not constitute infringement. Whether a particular use of copyrighted material by an AI developer is lawful depends on the facts and the relevant statutory provisions. Public availability of a work does not automatically mean that it is free to copy or use for every purpose. Equally, it would be premature to say that every use of copyrighted material in AI training necessarily infringes copyright. The issue requires analysis of the act performed, the rights involved and any applicable exception.
It is essential to distinguish two questions: whether an AI-generated output qualifies for copyright protection and whether the training or generation process has infringed rights in pre-existing works. The answers may differ. A protectable human contribution to an output would not automatically make all training practices lawful; a dispute over training data would not automatically decide authorship of every output.
3. Authorship and Originality
3.1 Who Causes the Work to Be Created?
The phrase ‘the person who causes the work to be created’ in Section 2(d)(vi) is central to the analysis. Consider three situations.
First, a user provides detailed creative instructions, selects an output from many alternatives and substantially edits it.
Second, a user enters a short prompt and accepts an output without further intervention.
Third, a developer designs a general-purpose AI system that produces outputs for numerous users. It would be difficult to conclude that the developer automatically owns copyright in every output simply because the developer created the underlying technology.
These examples show that causal involvement and creative authorship are not necessarily identical. A person may contribute to the process without contributing original expression to the final work.
A sound analysis should therefore consider the particular work and the roles of the people involved, rather than assume that every user, developer or system operator is automatically the author. This contribution-sensitive approach is a proposed analytical framework, not a settled statutory test expressly laid down for all generative AI cases.
3.2 Originality and Human Creative Contribution
In Eastern Book Company v. D.B. Modak, the Supreme Court of India considered the originality standard and rejected a test based only on the expenditure of labour. The decision is commonly understood to require a sufficient degree of skill and judgement involving creativity, rather than mere mechanical effort. [2]
Although the case did not concern generative AI, it provides a useful framework for evaluating the originality of human contributions to AI-assisted works.
A person may exercise creative judgement by making detailed choices, arranging elements, selecting among outputs or substantially modifying the final expression. By contrast, merely initiating an automated process may provide weaker evidence of original human expression.
The precise legal consequence will depend on the work and applicable law. Originality and authorship should also remain distinct questions: an output may be new or non-copying, yet the law must still identify the author to whom rights can be attributed.
3.3 The Copyright Office’s DABUS Decision
On 31 August 2026, the Indian Copyright Office rejected an application concerning the artwork A Recent Entrance to Paradise that named DABUS, an AI system, as the author.
Reports of the order state that the Office considered Section 2(d)(vi) and concluded that the AI system could not itself be entered as the statutory author. The order also considered the applicant’s role in conceiving and configuring the system and supplying inputs for the particular work. [3]
The decision is significant because it separates the system that immediately generates an output from the person who may legally be identified as having caused the work to be created.
It should not, however, be overstated as a comprehensive judicial rule for every AI-assisted work. It is an administrative decision concerning a particular application, and broader questions about different forms of human contribution remain open to further clarification.
Before submission, the complete order should be obtained and checked against the summary of the decision.
4. Emerging Legal Challenges
4.1 Training Data and Existing Copyright Works
Generative AI models may be trained using large collections of text, images, music and other material. If protected works are included, questions may arise about reproduction, authorisation and statutory exceptions.
The legal analysis should be specific: the relevant rights, acts and exceptions must be identified rather than assuming that all training is lawful or unlawful.
In December 2025, the Department for Promotion of Industry and Internal Trade published Part I of a working paper on generative AI and copyright.
The paper proposed a framework for use of lawfully accessed copyrighted material in AI training through a blanket licensing arrangement and a centralised royalty mechanism.
This was a policy proposal, not an enacted amendment to the Copyright Act. [4] Its policy status must be stated accurately. The proposal also illustrates that training-data questions are related to, but distinct from, authorship of AI-generated outputs.
4.2 Ownership and Contracts
Ownership disputes can arise where AI-assisted material is created by employees, commissioned creators, independent contractors or users of online platforms.
Agreements should distinguish rights in the underlying software, any human-created contribution and any copyright that validly subsists in the final output.
Terms of service may allocate contractual rights between parties, but they cannot automatically create statutory copyright where the legal requirements are not met.
4.3 Similarity, Enforcement and Evidence
An AI-generated output may resemble a pre-existing protected work. Similarity alone does not automatically prove infringement; the analysis may require consideration of protected expression, the extent of reproduction and applicable exceptions.
Enforcement may also be complicated when records of prompts, edits, output selection and source material are unavailable.
Maintaining appropriate records can help clarify how a work was produced and what each participant contributed.
5. Critical Analysis and Recommendations
The existing law has an important advantage: Section 2(d)(vi) already recognises computer-generated works. The challenge is applying its language to modern systems that can generate complex outputs with limited human intervention. Granting rights automatically to anyone who initiates a process could attribute copyright without sufficient creative contribution. Conversely, excluding every AI-assisted work could disregard genuine human creativity in directing, selecting and modifying an output.
First, legislative or authoritative guidance should clarify how the phrase ‘person who causes the work to be created’ applies to generative AI. The guidance should consider the specific work and the respective roles of users, developers and other participants. Second, copyright claims should identify the human contribution relied upon, such as original selection, arrangement or substantial modification. This need not mean that every use of AI requires burdensome documentation; the degree of evidence should be proportionate to the nature and value of the claim.
Third, creators and businesses should maintain suitable records of inputs, edits, output selection and relevant licences where practical. Fourth, commercial contracts should clearly address permitted AI tools, rights in human-created elements, licensing responsibilities and procedures for dealing with infringement allegations. Fifth, any reform concerning training data should balance the interests of copyright holders with innovation and public interest. Policy proposals should be distinguished from existing law until enacted or authoritatively decided.
A separate, automatically protected category for every AI output is not necessarily the best solution. It could create exclusive rights over material lacking meaningful human creativity and complicate access to the public domain. A more measured approach would clarify existing authorship rules, recognise demonstrable human creative contribution and address training practices through a separate, evidence-based policy process.
6. Conclusion
Generative AI challenges copyright law by separating the production of an output from traditional ideas of human authorship. The Copyright Act, 1957 provides a starting point through Section 2(d)(vi), but questions remain about how to identify the person who causes a work to be created and how to assess human originality. The reasoning in Eastern Book Company v. D.B. Modak offers a useful originality framework, while the Copyright Office’s 2026 DABUS decision highlights the need to distinguish a technological system from a legally recognised author.
Indian law would benefit from clearer guidance on authorship, proportionate evidence of human contribution and transparent contractual arrangements. Training-data disputes should be analysed separately from the copyrightability of outputs. The objective should be to encourage responsible technological development without weakening the rights of human creators or granting unsupported monopolies over material that does not qualify for protection.
References
Legislation
- The Copyright Act, 1957, Act No. 14 of 1957, especially Sections 2(d)(vi), 13, 14, 17, 18, 30, 51 and 52. Official text: https://copyright.gov.in/Copyright_Act_1957/chapter_i.html
Judicial Decision
- Eastern Book Company v. D.B. Modak, (2008) 1 SCC 1, Supreme Court of India. A copy is available at https://indiankanoon.org/doc/1062099
Administrative Decision
- Copyright Office, Government of India, order dated 31 August 2026 concerning A Recent Entrance to Paradise, Diary No. 9356/2022-CO/A. Reporting and analysis: Bar & Bench, “AI-generated work can get copyright but AI system cannot be author: Indian copyright office” (31 August 2026), https://www.barandbench.com/news/litigation/ai-generated-work-can-get-copyright-but-ai-system-cannot-be-author-indian-copyright-office; SCC Times, “Original, yes, Author, no: Copyright Office rejects AI authorship, but not AI-generated art” (1 September 2026), https://www.scconline.com/blog/post/2026/09/01/authorship-of-ai-legislative-prerogative-dabus-copyright-office/. Obtain and cite the complete official order if available.
Government Publication
- Department for Promotion of Industry and Internal Trade, Working Paper on Generative AI and Copyright, Part I: One Nation One License One Payment: Balancing AI Innovation and Copyright (December 2025), https://www.dpiit.gov.in/static/uploads/2025/12/ff266bbeed10c48e3479c941484f3525.pdf. Treat it as a policy proposal, not enacted law.
Official Guidance
- Intellectual Property India, “Basics of Copyright,” https://ipindia.gov.in/pages/copyright/learn/basics-of-copyright (accessed 24 October 2026).
Written By: Bhumika Tiwari, B.Com. LL.B. (Hons.), Mangalayatan University, Jabalpur 4th year (7th sem)
Submitted to: Dr. Pooja Agrawal – Academic Year: 2026–2027
