I. Introduction

The world is on the verge of a judicial reckoning on the matter of Generative Artificial Intelligence (AI). Over the last two years, there have been several high-profile cases in the West, including the case of The New York Times v OpenAI, that have been brought to court to decide whether or not scraping the internet for training data is illegal or ‘fair use’. These conflicts, however, are based on an institutional order that is rooted in the West’s legal and judicial systems, and so have been speculated upon in the Global South.

The speculation was brought to an end by the Delhi High Court in its awaited ruling in ANI Media Pvt. Ltd. v. OpenAI OpCo LLC. This is India’s first significant legal battle concerning the copyright implications of large language models (LLMs), and it’s not just about Intellectual Property (IP) rights. It places the Delhi High Court at the intersection of legal, political, and social concerns. India has no clear-cut statutory safety valves, so that the forthcoming verdict alone will spell the future of the Indian digital press and the limits of technological innovation within the country’s local terrain of the political economy.

II. The Statutory Void: The Absence of a TDM Exception

It is important to comprehend the enormity of the stakes at hand in the battle of ANI v OpenAI by observing the structural difference between Indian copyright regime and the international copyright regime first. The key to a thriving AI development ecosystem in most jurisdictions is to set a legislative framework that allows for the ingestion of computational data to be done safely. For example, the European Union has an explicit Text and Data Mining (TDM) exception defined in its Digital Single Market Directive that permits commercial AI scraping, as long as rights-holders do not opt out. Likewise, the United States has adopted a far-ranging, four factor test for defining “transformative fair use” to protect technological indexing.

The Copyright Act, 1957 of India does not have any such mechanisms. The Act was drafted before the internet, and defines a very limited list of ‘fair dealing’ exceptions in Section 52.[1] The exceptions are strictly limited to human activities: private research, critique, review and reporting of current events.

OpenAI has been putting commercial models on the free, scraped, journalistic data, which has put them in a statutory round hole. The Delhi High Court is not expected to exercise its imagination in interpreting the existing laws, but it has a huge socio-legal dilemma. It has to either do some radical judicial interpretation to expand the definition of ‘fair dealing’ to also cover automated machine learning or it has to stick to the word of the law and declare the whole concept of generative AI training pipelines as violative.

III. The Socio-Political Dimensions of Data Scraping

In the long and acrimonious arguments before Justice Amit Bansal, the Digital News Publishers Association (DNPA) jumped in with a clarion call: the unchecked and unauthorized ingestion of news data by LLMs poses a threat to the core economics of journalism. This argument shifts the discussion from the technical matter to a socio-political one.  The primary purpose of copyright law is to serve the public interest by offering the economic motivation for creating original works.[2] If newsrooms can be easily mined for content, then it can be combined with others in whatever manner, and AI can then provide instant answers directly to users, breaking the commercial feedback loop. No longer clicking through to the origin, users are killing the money sources behind real-world journalism: ad revenue and subscription models. Moreover, OpenAI’s claim that it only uses “publicly available” facts and that its models are trained elsewhere indicates that there’s a clear geopolitical imbalance. On a political economy level, this is similar to the extraction models of history. Western technology giants increasingly draw on digital resources from the Global South, such as Indian-language databases, to develop proprietary platforms that generate substantial commercial value. When the court considers local, created-for-human consumption content to be a free and unregulated commons, it stands to undermine the economic health of cultural and informational production in the United States.

IV. Predicting the Judicial Horizon

The Delhi High Court’s verdict, which was reserved, will probably create a middle ground between complete ban and blanket use, especially in the absence of any legislative movement on the topic of governance in the area of Artificial Intelligence.

The court is highly unlikely to grant a sweeping interim injunction that completely halts OpenAI’s operations in India, especially given that OpenAI has already proactively blocklisted ANI’s domain from future data scraping. But the central question is whether the court has the authority to decide the legality of the old data ingestion, which was previously ingested by trained data that were stored in the model weights.

The benchmark analysis will probably be divided into two parts: Ingestion and Output. If the output is a regurgitation of copyrighted news text, usually due to having memorized the text from the model, the court will almost certainly find it to be an outright case of copyright infringement.[3]

The real jurisprudential showdown is at the eating phase. The court may conclude that the mere temporary storage of data on a computer for processing and analysis of the data does not constitute an impermissible ‘expression’ of a work, but that a commercially beneficial use of that trained model without a statutory safe harbour would do so.[4] The court has indicated that it feels deeply inclined to build a balanced framework, which is uniquely Indian, by appointing eminent academics as Amicus Curiae. The ruling is likely to establish a precedent for a Collective Licensing Architecture, where tech companies are required to pay licenses to digital publishers into an institutional fund.

V. Conclusion

The upcoming verdict in ANI vs OpenAI holds significant importance for law and order in India. It makes our legal frameworks face a crucial dilemma: does law adapt to accommodate technology, or technology to the laws? As the Delhi High Court is set to render its verdict, it has a heavy burden of responsibility to establish a precedent for the developing world at large. Technology has to be developed within a framework of the law, or it will not be advanced. The only thing the court has to do is make sure that the rise of the artificial machine doesn’t lead to the economic downfall of the human press.

[1] Copyright Act 1957, s 52.

[2] William M Landes and Richard A Posner, The Economic Structure of Intellectual Property Law (Harvard University Press 2003).

[3] Copyright Act 1957, ss 14 and 51.

[4] Copyright Act 1957, ss 2(ff) and 14.

Aparna Tamrakar is a law graduate with an LL.M. in Intellectual Property Rights & Technology Law from O.P. Jindal Global University and is enrolled with the Bar Council of Madhya Pradesh. Her areas of interest include intellectual property law, technology and data protection law, trademark and patent compliance, and legal research.

Aryan Sharma is an advocate enrolled with the Bar Council of Delhi and currently works as an Associate at Udgam Law, with a focus on dispute resolution and arbitration. His areas of interest include arbitration, insolvency and bankruptcy law, competition law, and technology law.