Introduction

In an era where algorithms increasingly determine what we read, hear, and believe, the lawsuit filed by Asian News International (ANI) against OpenAI poses a foundational challenge to how Indian copyright law accommodates artificial intelligence. At the heart of the matter is a simple yet seismic question: can copyrighted content be ingested en masse by machine learning systems without explicit consent under the guise of “fair dealing”? The answer will likely determine not just the limits of AI development in India, but the very fabric of content ownership and media credibility in a digital society.

Beyond Mere Infringement: A Question of Institutional Ethics

ANI’s grievance is not just legalistic, it is existential. ANI accuses OpenAI of unauthorized scraping of its journalistic output to train ChatGPT, and more damagingly, of generating fictitious or misleading content wrongly attributed to the agency. This dual injury, of intellectual property and journalistic integrity, elevates the dispute from a mere infringement claim to a crisis of institutional ethics in the age of machine-generated knowledge.

Unlike conventional copyright disputes, the case implicates reputational harm in a media ecosystem already plagued by misinformation. In ANI’s framing, OpenAI is not just an infringer but a purveyor of synthetic distortions that masquerade as credible journalism. This reputational dimension, if accepted by the court, could open new tortious avenues in Indian jurisprudence involving defamation and misattribution in AI-generated speech.

The Fragility of Fair Dealing in India’s Copyright Architecture

Indian copyright law is governed by a list-based exception regime under Section 52 of the Copyright Act, 1957. Unlike the open-ended fair use doctrine in the U.S., India’s “fair dealing” defense enumerates specific purposes, private use, criticism, reporting, research, and education. Whether AI training qualifies under any of these is both legally uncertain and doctrinally fraught. There is no clear judicial precedent that interprets “research” to include commercial, large-scale data mining by foreign entities.

The court will therefore have to break new ground: does computational training, by a for-profit corporation using unlicensed media content, amount to “fair dealing” for the purpose of research? And if so, does the transformative nature of AI training offset its commercial intent? These are uncharted waters for Indian courts, which have traditionally been conservative in their interpretation of exceptions.

The Transformative Use Mirage

OpenAI’s defense relies heavily on the notion of “transformative use”, a U.S.-centric doctrine that allows derivative usage if the end product serves a new, socially valuable function. However, the Indian judiciary has not formally adopted the transformative use test. Even if it did, the utility of AI training as transformative is ambiguous.

Unlike the digitization of books for searchability (as upheld in Authors Guild v. Google), AI training absorbs the semantic content of the original works in ways that defy direct comparison. The content is not just used, it is metabolized. What emerges is not a derivative work in the traditional sense but a latent statistical model of language, capable of imitating styles and fabricating plausible news-like outputs. This functional difference challenges the very premise of transformation as it applies to LLMs.

Moreover, the argument that OpenAI only analyzes linguistic patterns and not expressive content is disingenuous. If ChatGPT can simulate ANI’s tone and style or reference it falsely, then it is precisely the expressive dimension, the protected part, that has been appropriated, even if indirectly.

Harm Beyond Economics: Reputational Injury in the AI Era

The traditional economic test of market substitution is ill-suited to measure the harm caused in this case. ANI is not merely alleging revenue loss, but degradation of trust in its brand due to false attributions. The law must therefore expand its toolkit to recognize non-economic injuries in the AI context, including distortion, dilution, and reputational erosion.

Furthermore, ANI’s claim that ChatGPT attributed fabricated content to it, without any user prompt referencing ANI, raises urgent questions about dataset provenance and the lack of safeguards in LLM deployments. In a country grappling with disinformation and hate speech, the ability of AI systems to generate news-like fabrications with credible attributions is not a trivial technical bug; it is a democratic hazard.

Jurisdictional Anchoring in the Age of Stateless Technologies

OpenAI’s global infrastructure and U.S. incorporation could have undermined the ability of Indian courts to claim jurisdiction. Yet the Delhi High Court’s assertion of jurisdiction, based on the place of harm and the Indian domicile of the plaintiff, signals a broader willingness to anchor digital wrongs within national legal systems. This approach is likely to be mirrored in other emerging economies seeking to assert regulatory sovereignty over foreign tech firms.

It also underlines a critical shift: in the borderless architecture of AI, courts must move beyond territorial formalism to impact-based jurisdiction. The ANI case could thus become a blueprint for how India and other jurisdictions impose accountability on stateless digital actors.

Comparative Tensions and the Global Patchwork

Globally, courts are beginning to test the boundaries of AI and copyright. In Getty Images v. Stability AI, The New York Times v. OpenAI, and various author-led lawsuits across jurisdictions, plaintiffs are making a common argument: that training on copyrighted works without consent or compensation is exploitative. These cases, still in progress, will shape the international doctrine of AI fair use.

India’s own alignment remains uncertain. If the ANI court interprets fair dealing expansively, it may open the floodgates for unregulated AI training using Indian content. If it rules narrowly, it could stifle AI innovation but preserve the moral and economic rights of creators. A middle path, such as a statutory licensing regime, would require legislative action, which remains absent.

Charting a Legislative Future: From Ambiguity to Accountability

The ANI case highlights the urgency of legislative clarity. India lacks specific provisions on text and data mining (TDM), unlike the EU’s tailored exception under the Digital Single Market Directive. Nor does it have transparency requirements mandating AI developers to disclose training datasets.

To navigate the future, Parliament must consider:

1. introducing a limited TDM exception for non-commercial AI research with opt-out rights for rights holders,
2. establishing collective licensing mechanisms to ensure compensation, and
3. requiring algorithmic transparency and attribution audits in generative AI models.

These steps would harmonize the twin imperatives of innovation and protection.

Conclusion: ANI v. OpenAI as a Constitutional Moment

At its core, ANI v. OpenAI is not just a copyright dispute, it is a referendum on how Indian law will confront the epistemic and ethical consequences of artificial intelligence. Will the courts reassert the sovereignty of human authorship, or capitulate to the inevitability of machine intermediaries? Will reputational integrity be recognized as a form of digital dignity?

This case forces Indian jurisprudence to catch up with technological reality. If properly adjudicated, it can offer a foundational precedent for how developing countries frame digital constitutionalism: asserting control, demanding transparency, and defending the cultural and economic value of local content in a global algorithmic economy.

ANI v. OpenAI may just be the first act in a long legal theatre, but its script will resonate far beyond the courtroom.

(This post has been authored by Tejas Hinder, an associate at Cyril Amarchand Mangaldas and an editor at TCLF)

CITE AS: Tejas Hinder, ‘’ (The Contemporary Law Forum, 27 May 2025) <https://tclf.in/2025/05/27/scraping-the-truth-ani-v-openai-and-the-fractures-in-indias-fair-dealing-doctrine/>date of access.