The Next Big Copyright Question in India

Generative Artificial Intelligence (AI) has created one of the most significant legal crossroads in modern copyright law. Large Language Models (LLMs) are trained using enormous volumes of text, images, music, software code, and other creative works, many of which are protected under copyright law.

Across the world, authors, publishers, artists, record labels, technology companies, and AI developers are engaged in legal disputes over a fundamental question that could shape the future of artificial intelligence:

Does training an AI model on copyrighted content require permission from copyright owners, or can it be justified under existing copyright exceptions?

While courts in the United States, United Kingdom, and European Union are beginning to examine this issue, India has yet to provide a definitive legal answer.

Understanding India's Copyright Framework

The starting point for any analysis is Section 52 of the Copyright Act, 1957, which contains India's statutory fair dealing exceptions.

Unlike some jurisdictions that recognize broader judicial doctrines such as Fair Use, India's copyright law specifies only certain circumstances in which copyrighted works may be used without permission from the copyright owner.

This immediately raises an important legal question:

Can developers argue that training Large Language Models using copyrighted material amounts to "research" or another protected activity under Section 52?

At present, the Copyright Act provides no explicit answer. A close reading of Section 52 suggests that Parliament did not specifically contemplate commercial artificial intelligence systems when drafting the legislation.

Consequently, extending Section 52 to cover AI training would likely require either judicial interpretation or legislative amendment rather than a broad reading of the existing provision.

Does Section 52 Protect AI Training?

India follows a statutory Fair Dealing model rather than the more flexible Fair Use doctrine applied in the United States.

Section 52 lists specific situations where copyrighted works may be used without constituting infringement, including:

• Private or personal use
• Research and private study
• Criticism or review
• Reporting of current events
• Educational purposes

Because these exceptions are expressly listed by Parliament, Indian courts generally cannot create entirely new categories of permissible use simply because technology evolves.

Instead, the courts are expected to interpret the statute according to its language, purpose, and legislative intent.

This makes India's copyright framework significantly narrower than jurisdictions that permit courts to balance competing interests through broader judicial doctrines.

How Indian Courts Interpret Fair Dealing

One of the most important decisions interpreting Section 52 is the Delhi High Court's judgment in University of Oxford v. Rameshwari Photocopy Services.

The dispute involved educational photocopying of academic books by a university photocopy shop.

The Court interpreted the educational exception under Section 52 broadly enough to achieve Parliament's objective of facilitating education, but importantly, it remained within the boundaries of the statutory language.

The judgment demonstrates that Indian courts may interpret statutory exceptions purposively, but they do not possess unlimited authority to create new exceptions that Parliament never enacted.

This principle becomes particularly relevant when considering AI training, since no provision of Section 52 expressly addresses machine learning or artificial intelligence.

Can AI Model Training Qualify as Research?

Supporters of expansive AI exceptions argue that training Large Language Models is fundamentally a research exercise.

According to this view, AI systems do not merely reproduce books or articles for public consumption. Instead, they identify statistical relationships and linguistic patterns across millions—or even billions—of pieces of content.

Developers therefore argue that the purpose of copying is analytical rather than expressive.

However, critics identify an important legal distinction.

Human researchers generally read copyrighted works directly. AI developers, by contrast, typically create extensive digital copies of copyrighted material during the training process before any statistical analysis occurs.

Those reproductions themselves may constitute acts restricted by copyright law.

Furthermore, many modern AI models are no longer experimental academic research projects.

Today's LLMs power subscription-based AI assistants, enterprise software, search engines, customer service platforms, content creation tools, coding assistants, and commercial products generating billions of dollars in revenue.

Given their commercial nature, equating industrial-scale AI training with the "research" or "private study" contemplated by Section 52 becomes considerably more difficult.

What the Civic Chandran Judgment Says

The Kerala High Court's decision in Civic Chandran v. Ammini Amma further illustrates how Indian courts examine fair dealing claims.

The Court emphasized that determining whether a use qualifies as fair dealing requires consideration of the purpose and character of the impugned use.

This suggests that merely labeling an activity as "research" may not automatically bring it within the statutory exception.

Where AI development forms part of large-scale commercial operations, courts may examine whether the commercial objective outweighs claims that the copying occurred solely for research purposes.

Accordingly, Civic Chandran does not necessarily resolve the AI training debate but reinforces the importance of carefully analyzing both the nature and purpose of the use.

Why the AI Copyright Debate Matters

The outcome of this debate extends far beyond technology companies.

If AI developers may freely train models using copyrighted material without authorization, millions of creators could see their works incorporated into AI systems without compensation.

Those affected may include:

• Authors
• Journalists
• Publishers
• Musicians
• Record Labels
• Film Studios
• Artists
• Photographers
• Software Developers
• Academic Researchers

At the same time, requiring licenses for every work used during AI training could significantly increase development costs and potentially slow innovation.

The interpretation ultimately adopted by Indian courts—or Parliament—will therefore shape both India's copyright landscape and its rapidly growing AI ecosystem.

Innovation vs. Copyright Protection

Arguments Supporting a Broader Interpretation

• Encourages innovation in artificial intelligence.
• Supports Indian AI startups with limited financial resources.
• Reduces barriers to AI research and development.
• Promotes technological competitiveness.
• Accelerates domestic AI capabilities.

Arguments Supporting a Narrow Interpretation

• Strengthens copyright protection for creators.
• Ensures authors and rights holders retain control over their works.
• Encourages licensing markets for AI training datasets.
• Provides greater legal certainty.
• Creates opportunities for fair compensation to creators.

Conclusion: The Road Ahead for AI and Copyright in India

India now faces one of the most significant copyright questions of the artificial intelligence era.

Should commercial AI training using copyrighted material be interpreted as falling within Section 52 of the Copyright Act, or should Parliament enact entirely new statutory provisions specifically addressing artificial intelligence?

The answer will influence not only future copyright litigation but also India's innovation strategy, investment climate, digital economy, and the balance between technological progress and the rights of creators.

Around the world, courts are beginning to establish legal principles governing AI training. As similar disputes inevitably reach Indian courts, judges will likely need to reconcile decades-old copyright legislation with technologies that lawmakers could scarcely have anticipated.

Until either Parliament amends the Copyright Act or Indian courts deliver a definitive ruling, the legal status of AI training using copyrighted works will remain uncertain—making it one of the defining copyright issues of the AI age.