Can AI Training on Copyrighted Content Amount to Copyright Infringement in India?

Can AI Training on Copyrighted Content Amount to Copyright Infringement in India?. Generative artificial intelligence has created one of the most important copyright questions of the digital age: can developers use copyrighted books, news reports, photographs, music and other protected material to train AI systems without permission? The answer under Indian law cannot be reduced to a simple yes or no.
AI training may involve obtaining, storing and processing protected works and can therefore raise questions under the Copyright Act, 1957. At the same time, Indian copyright law contains statutory exceptions, particularly Section 52. Whether a particular use amounts to infringement depends on the statutory rights involved, the facts of the use and the applicability of any exception.
The Statutory Framework
The Copyright Act, 1957 grants copyright owners exclusive rights in protected works according to the category of work involved. Where copyrighted material is copied or stored during AI training, those activities may engage the copyright owner's exclusive rights and the infringement provisions of the Act.
Section 52 identifies certain acts that do not constitute copyright infringement. Section 52(1)(a) deals with fair dealing with a work, other than a computer programme, for purposes including private or personal use, including research. Its Explanation also addresses the storing of a work in an electronic medium for the purposes mentioned in that clause.
ANI Media Pvt. Ltd. v. OpenAI OpCo LLC
The most significant Indian judicial development so far is ANI Media Pvt. Ltd. v. OpenAI OpCo LLC, CS(COMM) 1028/2024, decided by the Delhi High Court on 24 July 2026. Justice Amit Bansal delivered the judgment while deciding ANI's application for an interim injunction.
ANI alleged that its copyrighted news content had been used without authorisation in connection with training the large language models underlying ChatGPT. The Court separately examined the storage of ANI's literary works for training and the allegation that ChatGPT outputs reproduced ANI's protected expression.
On a prima facie basis, the Court held that OpenAI's act of storing ANI's original literary works for training the LLMs underlying ChatGPT fell within Section 52(1)(a) and therefore did not amount to infringement under Section 51. The Court also found, at the interim stage, that the challenged RAG-based outputs were not substantially similar to ANI's original literary works and that ANI had not established memorisation and regurgitation through ChatGPT's responses.
This qualification is crucial. The decision concerned an interim injunction application. The Court expressly stated that its observations were only for deciding that application and would not determine the final outcome of the suit.
Training Data and AI Outputs Are Different Questions
One question concerns training: whether copyrighted works may lawfully be copied or stored for training an AI model. A separate question concerns outputs: whether the system subsequently reproduces protected expression in a manner that infringes copyright. A conclusion on one issue does not automatically resolve the other.
Why the Legal Position Remains Developing
The ANI judgment is a major development, but it should not be treated as a blanket licence for every AI developer to use all copyrighted material without restriction. Its findings were prima facie findings based on the facts before the Court and were made at the interim stage.
Future disputes may involve different kinds of works, different technical systems and different methods of obtaining data. Courts may need to examine the purpose and scale of copying, the applicable statutory exception, the fairness of the dealing, economic effects and whether protected expression is substantially reproduced in outputs.
Arguments on Both Sides
AI developers may argue that training is different from publishing or distributing the original works and that models process large quantities of material to identify patterns and relationships. Copyright owners may respond that large-scale unauthorised copying can interfere with licensing opportunities and that memorisation or substantial reproduction may cause direct commercial harm.
Practical Lessons
AI developers should not assume that simply describing an activity as research automatically eliminates copyright risk. Clear records concerning data sources, permissions, licences and technical safeguards remain important. Copyright owners should preserve evidence of ownership and publication and distinguish carefully between concerns about training and concerns about allegedly infringing outputs.
Conclusion
Can AI training on copyrighted content amount to copyright infringement in India? Yes, it can raise a genuine copyright issue because training may involve copying and storing protected works. However, copying alone does not provide the complete legal answer because statutory exceptions, particularly Section 52, may apply depending on the facts.
As of September 2026, the Delhi High Court's decision in ANI Media Pvt. Ltd. v. OpenAI OpCo LLC is the most important Indian judicial development on this issue. The Court took the prima facie view that OpenAI's storage of ANI's literary works for training the LLMs underlying ChatGPT fell within Section 52(1)(a), and it dismissed ANI's interim-injunction application. The Court also made clear that its observations would not determine the final outcome of the suit.
The safest conclusion is therefore balanced: AI training is not automatically lawful in every situation, but neither is every use of copyrighted material for AI training automatically infringement. Indian law is developing a judicial framework for analysing these disputes, and future cases will depend on the statutory provisions, technology and evidence in each case.

Generative artificial intelligence has created one of the most important copyright questions of the digital age: can developers use copyrighted books, news reports, photographs, music and other protected material to train AI systems without permission? The answer under Indian law cannot be reduced to a simple yes or no.

AI training may involve obtaining, storing and processing protected works and can therefore raise questions under the Copyright Act, 1957. At the same time, Indian copyright law contains statutory exceptions, particularly Section 52. Whether a particular use amounts to infringement depends on the statutory rights involved, the facts of the use and the applicability of any exception.

The Statutory Framework

The Copyright Act, 1957 grants copyright owners exclusive rights in protected works according to the category of work involved. Where copyrighted material is copied or stored during AI training, those activities may engage the copyright owner's exclusive rights and the infringement provisions of the Act.

Section 52 identifies certain acts that do not constitute copyright infringement. Section 52(1)(a) deals with fair dealing with a work, other than a computer programme, for purposes including private or personal use, including research. Its Explanation also addresses the storing of a work in an electronic medium for the purposes mentioned in that clause.

ANI Media Pvt. Ltd. v. OpenAI OpCo LLC

The most significant Indian judicial development so far is ANI Media Pvt. Ltd. v. OpenAI OpCo LLC, CS(COMM) 1028/2024, decided by the Delhi High Court on 24 July 2026. Justice Amit Bansal delivered the judgment while deciding ANI's application for an interim injunction.

ANI alleged that its copyrighted news content had been used without authorisation in connection with training the large language models underlying ChatGPT. The Court separately examined the storage of ANI's literary works for training and the allegation that ChatGPT outputs reproduced ANI's protected expression.

On a prima facie basis, the Court held that OpenAI's act of storing ANI's original literary works for training the LLMs underlying ChatGPT fell within Section 52(1)(a) and therefore did not amount to infringement under Section 51. The Court also found, at the interim stage, that the challenged RAG-based outputs were not substantially similar to ANI's original literary works and that ANI had not established memorisation and regurgitation through ChatGPT's responses.

This qualification is crucial. The decision concerned an interim injunction application. The Court expressly stated that its observations were only for deciding that application and would not determine the final outcome of the suit.

Training Data and AI Outputs Are Different Questions

One question concerns training: whether copyrighted works may lawfully be copied or stored for training an AI model. A separate question concerns outputs: whether the system subsequently reproduces protected expression in a manner that infringes copyright. A conclusion on one issue does not automatically resolve the other.

Why the Legal Position Remains Developing

The ANI judgment is a major development, but it should not be treated as a blanket licence for every AI developer to use all copyrighted material without restriction. Its findings were prima facie findings based on the facts before the Court and were made at the interim stage.

Future disputes may involve different kinds of works, different technical systems and different methods of obtaining data. Courts may need to examine the purpose and scale of copying, the applicable statutory exception, the fairness of the dealing, economic effects and whether protected expression is substantially reproduced in outputs.

Arguments on Both Sides

AI developers may argue that training is different from publishing or distributing the original works and that models process large quantities of material to identify patterns and relationships. Copyright owners may respond that large-scale unauthorised copying can interfere with licensing opportunities and that memorisation or substantial reproduction may cause direct commercial harm.

Practical Lessons

AI developers should not assume that simply describing an activity as research automatically eliminates copyright risk. Clear records concerning data sources, permissions, licences and technical safeguards remain important. Copyright owners should preserve evidence of ownership and publication and distinguish carefully between concerns about training and concerns about allegedly infringing outputs.

Conclusion

Can AI training on copyrighted content amount to copyright infringement in India? Yes, it can raise a genuine copyright issue because training may involve copying and storing protected works. However, copying alone does not provide the complete legal answer because statutory exceptions, particularly Section 52, may apply depending on the facts.

As of September 2026, the Delhi High Court's decision in ANI Media Pvt. Ltd. v. OpenAI OpCo LLC is the most important Indian judicial development on this issue. The Court took the prima facie view that OpenAI's storage of ANI's literary works for training the LLMs underlying ChatGPT fell within Section 52(1)(a), and it dismissed ANI's interim-injunction application. The Court also made clear that its observations would not determine the final outcome of the suit.

The safest conclusion is therefore balanced: AI training is not automatically lawful in every situation, but neither is every use of copyrighted material for AI training automatically infringement. Indian law is developing a judicial framework for analysing these disputes, and future cases will depend on the statutory provisions, technology and evidence in each case.