India's Copyright Crossroads: Delhi High Court Tackles AI Training in Landmark Case

New Delhi – The Delhi High Court is currently presiding over a legal battle that observers have dubbed "a litigation for the coming ages," one poised to define the relationship between artificial intelligence and intellectual property in one of the world's largest digital economies. The case, Asian News International (ANI) v. OpenAI , pits a leading news agency against the creator of ChatGPT, raising fundamental questions about whether training large language models (LLMs) on copyrighted material constitutes infringement or a new form of lawful learning. The outcome could set a globally significant precedent, shaping the future of both generative AI and content creation.

At its core, ANI’s lawsuit alleges that OpenAI unlawfully used its vast archive of copyrighted news content to train ChatGPT without permission, thereby infringing upon its exclusive rights. The court, under Justice Amit Bansal, has framed four pivotal issues: whether the initial storage of ANI’s data for training is an infringement; whether the subsequent use of that data to generate user responses is also an infringement; whether either action can be defended under the "fair dealing" exception in Section 52 of India's Copyright Act, 1957; and whether Indian courts have jurisdiction over OpenAI, whose servers are located in the United States.

"How the Delhi High Court resolves ANI v. OpenAI will determine whether India treats AI model training as lawful learning or copyright infringement," notes one legal analysis, underscoring the case's profound implications. "A choice that could reshape global debates on authors’ rights, innovation, and the future of generative AI."

The Legal Labyrinth: Navigating India's Copyright Act of 1957

Unlike more modern legislative frameworks in the European Union, which explicitly address text-and-data mining, India’s Copyright Act of 1957 was drafted long before the advent of generative AI. This legislative silence forces the court to interpret century-old principles in a novel technological context.

The central conflict revolves around Section 14, which grants authors the exclusive right to reproduce and store their work, and Section 52, which outlines exceptions for "fair dealing." ANI contends that OpenAI's ingestion and storage of its articles—even if converted into numerical vectors—constitutes "reproduction," an act prohibited under Section 14(a)(i).

OpenAI's defense is expected to argue that this process is not reproduction in a legally cognizable sense, as it does not create a human-readable copy of the original expression. Furthermore, they will likely contend that the training process is a form of "research" and thus falls under the fair dealing exception in Section 52(1)(a)(i). This positions AI training as a transformative act analogous to human learning, where a researcher reads countless sources to synthesize new understanding, rather than simple plagiarism.

The debate is further complicated by the fact that Indian law lacks a direct equivalent to the broad "fair use" doctrine found in the United States, which centers on a flexible, four-factor test and the concept of "transformativeness." While U.S. courts have begun to grapple with this, with some judges calling AI training "quintessentially transformative," Indian courts have historically been cautious about importing foreign legal tests that lack a clear statutory basis. The Delhi High Court must therefore navigate this issue using the more rigid, enumerated exceptions provided in Section 52.

Divergent Views from Amici Curiae

The complexity of the case has attracted significant attention, leading to the appointment of two amici curiae ("friends of the court") who have presented sharply contrasting interpretations of the law.

Professor Arul Scaria argued for a more adaptive approach, suggesting that storing copyrighted material for the sole purpose of training AI models should be permissible. In his view, the critical question is whether OpenAI uses ANI's content for anything beyond training. He cautioned that overly restricting the ingestion of data could "hinder knowledge dissemination" and stifle innovation.

In direct opposition, Advocate Adarsh Ramanujan advanced a stricter interpretation, asserting that any unauthorized copying of ANI's data, even once, constitutes a prima facie infringement. He contended that OpenAI's commercial use of ANI's content does not qualify for any of the fair dealing exceptions—it is not for private research, criticism, or news reporting. He further argued that ChatGPT's ability to generate responses that summarize or paraphrase ANI's reporting causes direct economic harm by diminishing the value of ANI's subscriptions.

These opposing views mirror the global policy dilemma at the heart of AI and copyright: how to balance the rights of creators who provide the raw material for innovation against the societal benefits of advancing powerful new technologies.

A Global Context with Local Consequences

The Delhi High Court's deliberations are not occurring in a vacuum. Similar legal battles are raging worldwide. In the United States, cases like The New York Times v. OpenAI are exploring the economic harm to publishers, while a recent settlement in Bartz v. Anthropic drew a line between training on lawfully acquired versus pirated content. Meanwhile, the European Union's 2019 Copyright Directive provides a legislative solution, creating explicit exceptions for text-and-data mining while allowing commercial rights holders to "opt-out."

India, lacking such a bespoke legislative framework, places the judiciary in the powerful but challenging position of acting as a de facto policymaker. A ruling that narrowly interprets fair dealing could vindicate authors' rights but may also create significant barriers for AI development in India, potentially pushing innovation to jurisdictions with more permissive laws. Conversely, a broad interpretation that shields AI training under the "research" exception would be a boon for tech companies but could be seen as devaluing the creative labor of journalists and publishers, who warn that their business models are at risk. The Digital News Publishers Association has submitted to the court that "if ChatGPT freely swallows news, a large chunk of news organisations may die out."

The Path Forward: A Call for Doctrinal Clarity

Beneath the technical legal arguments lies a fundamental question about the purpose of copyright law itself. As the source text aptly puts it, "Indian copyright exists to 'promote the progress of science and useful arts' by rewarding creativity, while not unduly hindering expression and learning." The ANI case forces a direct confrontation with this principle.

The court must decide whether to adhere strictly to the letter of the 1957 Act, potentially finding infringement and leaving it to Parliament to legislate a solution, or to interpret the existing law flexibly to accommodate a technology its drafters could never have envisioned.

Regardless of the final decision, the legal community is watching for clear, principled guidance. A muddled ruling or one that simply borrows foreign concepts without grounding them in Indian jurisprudence could create more uncertainty. The ideal outcome would be a judgment that transparently articulates its reasoning, carefully balances the competing interests of innovation and creation, and perhaps urges Parliament to address the statutory gaps. As the judiciary navigates this uncharted territory, its courage in providing doctrinal clarity will be crucial for creators and innovators alike in the rapidly evolving AI era.