Published on: 17th July 2026
Authored by: Saumik Sah
Raffles University
I. Introduction
In late 2024, the Delhi High Court initiated proceedings in one of the most significant intellectual property disputes in India’s digital history.[16] Asian News International (ANI), a prominent domestic wire agency, filed suit against OpenAI Inc. and OpenAI OpCo LLC, alleging that the generative artificial intelligence platform, ChatGPT, had been trained on its copyrighted journalistic content without prior authorization, licensing, or commercial compensation.[16]
While the dispute appears on the surface to be a standard claim of unauthorized reproduction, it presents a fundamental challenge to domestic intellectual property frameworks.[16] A statute drafted in 1957, long before the advent of the internet, is now required to determine whether scraping copyrighted news content at scale for commercial machine learning training constitutes a permissible statutory exception under the doctrine of “fair dealing” or amounts to copyright infringement.[16] The final determination of this matter will carry profound implications for all publishers, musicians, filmmakers, and authors whose creative outputs are integrated into generative AI development models.[16] This article analyzes the ANI v. OpenAI litigation up to March 2026, details the core legal issues framed by the Delhi High Court, evaluates the statutory rigidity of the Copyright Act, 1957, contrasts the dispute with comparative Western legal frameworks, and outlines the urgent legislative reforms required at this technological crossroads.[16]
II. Factual Background and Procedural History
ANI’s primary allegation centers on the unauthorized extraction of its digital properties.[16] The plaintiff claims that OpenAI utilized automated web-crawlers to scrape its online articles, bypassing subscriber paywalls to ingest the text into the training datasets for ChatGPT.[16] ANI further demonstrated that the AI platform’s outputs occasionally reproduced its proprietary text nearly verbatim and, significantly, generated artificial “hallucinations” by attributing completely fabricated news stories to ANI as the source.[16] This false attribution creates distinct reputational risks for the news agency.[16] Consequently, ANI sought permanent injunctive relief alongside 2 crore rupees in damages.[16]
Recognizing the novel legal questions at stake, the Delhi High Court appointed two Amici Curiae on the initial day of hearings: Adarsh Ramanujan, a senior intellectual property litigator, and Dr. Arul George Scaria, a professor at NLSIU Bengaluru.[16] The positions adopted by the Amici Curiae reflected the broader global debate, with Ramanujan emphasizing the protection of the economic rights of creators, and Scaria cautioning that an overly rigid application of copyright law could restrict technological innovation.[16]
As the litigation progressed, the scope of the dispute expanded beyond journalistic print media.[16] The Federation of Indian Publishers and the Digital News Publishers Association (DNPA) intervened in early 2025, arguing that mass web-scraping caused systematic economic harm across the publishing industry.[16] By February 2025, major entertainment conglomerates, including T-Series, Saregama, and Sony, sought to join the suit, demonstrating that the data-scraping crisis impacted copyrighted musical and audiovisual works as well.[16] In response, OpenAI blocklisted ANI’s digital domains from future training cycles while maintaining that its models extract only “non-expressive” statistical patterns rather than protected creative expressions.[16] Although the Delhi High Court declined to grant an interim injunction at the preliminary stage, it issued formal summons to the defendants.[16] Following thirty-two hearings, the Court officially reserved its order on interim relief on March 27, 2026.[16]
III. Core Legal Issues and Statutory Analysis
The Delhi High Court has narrowed the litigation down to four core legal questions that test the structural limits of the Copyright Act, 1957:[16]
A. Infringement via Digital Storage and Data Ingestion
The first issue considers whether the temporary or permanent storage of copyrighted works within an AI training corpus constitutes infringement under Section 13 of the Copyright Act, 1957.[16] ANI argues that compiling articles into a training dataset constitutes unauthorized “reproduction,” which violates the exclusive rights of the author.[16] OpenAI contends that the ingestion process transforms the underlying text into mathematical parameters and machine weights, which do not constitute a readable or actionable copy under traditional copyright definitions.[16]
B. Infringement via Generative Outputs and Moral Rights
The second issue examines whether the generation of user responses that draw on copyrighted training data constitutes an independent act of infringement.[16] This question is distinct from the data ingestion stage, as generative outputs may reproduce protected expressions verbatim.[16] Furthermore, the generation of fabricated news stories falsely attributed to the plaintiff engages Section 57 of the Act, which protects the author’s moral right of integrity against acts prejudicial to their honor or reputation.[16]
C. The Rigid Limits of the Fair Dealing Doctrine under Section 52
The third and most critical question is whether commercial machine learning qualifies for protection under Section 52 of the Act, which governs exceptions to copyright infringement.[16] In Super Cassettes Industries v. Chintamani Rao, the Delhi High Court established that Section 52 is statutorily exhaustive, meaning its exceptions are strictly confined to the categories explicitly listed by the legislature.[1] OpenAI’s commercial model cannot easily fit within the statutory exemptions for “private study,” “research,” or “news reporting” under Section 52(1)(b).[16]
While the defendant points to the concept of “transformative use” raised in Syndicate of the Press of the University of Cambridge v. B.D. Bhandari, that precedent was strictly confined to traditional textbook contexts and has never been applied to automated generative models.[2] Unlike the flexible, four-factor fair use standard applied under 17 U.S.C. § 107 in the United States, Section 52 of the Indian Act remains a closed list.[16] The DNPA argued in its submissions that the massive scale of permanent embedding within model weights cannot be characterized as “incidental copying” under Section 52(1)(c), meaning the extraction of billions of tokens from proprietary documents defeats any claim to fair dealing.[16]
D. Extraterritorial Jurisdiction and the Effects Doctrine
The fourth issue concerns whether Indian courts possess jurisdiction over a foreign technology firm that has no registered office or physical servers within Indian territory.[16] OpenAI challenged the court’s jurisdiction on these grounds, noting that the Indian Copyright Act lacks the explicit extraterritorial enforcement provisions found in Section 505 of the US Copyright Act.[16] However, the plaintiff relies on Section 62 of the Indian Copyright Act, which grants jurisdiction to the court within whose territorial limits the plaintiff resides or conducts business, as well as the “effects doctrine,” which permits courts to exercise jurisdiction if foreign conduct causes direct economic or reputational harm within India.[16]
IV. Comparative Global Perspectives
The challenges facing Indian courts reflect a broader global struggle to adapt copyright law to artificial intelligence, though different jurisdictions utilize distinct statutory mechanisms:[16]
* United States: The ongoing litigation in New York Times v. Microsoft and OpenAI centers on the direct reproduction of paywalled articles within training datasets.[16] US courts evaluate these claims using the equitable four-factor fair use test under 17 U.S.C. § 107, which provides significantly more judicial flexibility than India’s exhaustive statutory exceptions.[16]
* United Kingdom: Under Section 29A of the Copyright, Designs and Patents Act, 1988, the UK provides a specific statutory exception for computational data analysis.[16] However, this exception applies exclusively to cases of “lawful access,” meaning that the unauthorized scraping of paywalled or subscription-restricted text remains an actionable infringement.[16]
* European Union: Article 4 of the Digital Single Market Directive 2019/790 establishes a dedicated text and data mining exception for commercial AI models.[3] However, this framework includes a robust opt-out mechanism, requiring AI developers to respect machine-readable reservation notices implemented by rights-holders.[3]
India currently lacks any equivalent, AI-specific statutory provision.[16] Although the Ministry of Commerce and Industry constituted an eight-member expert panel to evaluate potential amendments to the Copyright Act, no formal amending legislation has been introduced to the legislature.[16]
V. Strategic and Systemic Implications
The final outcome of the ANI v. OpenAI suit will have significant implications across several sectors of the digital economy:[16]
A. Impact on AI Developers and Data Acquisition
An adverse judicial ruling would require AI developers to reassess their data ingestion practices.[16] Technology firms might be forced to clear licenses from domestic publishers, remove Indian data from global training corpuses, or build specialized models restricted to consented data, all of which would significantly increase operational costs.[16]
B. Vindication of Content Creators’ Rights
A ruling in favor of the plaintiff would clarify that public accessibility on the internet does not grant a free commercial license for machine learning training, establishing a strong precedent for the economic rights of creators in the digital era.[16]
C. Convergence with IT Law and Data Protection Frameworks
The Court’s jurisdictional rulings will interact directly with India’s emerging digital regulatory framework, including platform liability exemptions under the Information Technology Act, 2000, and data handling mandates under the Digital Personal Data Protection Act, 2023.[16]
VI. Conclusion and Legislative Recommendations
The litigation in ANI v. OpenAI serves as a key test of whether India’s pre-internet copyright laws can effectively address the challenges of generative artificial intelligence.[16] The statutory rigidity of Section 52 prevents courts from easily expanding exceptions to cover unforeseen commercial technologies without risking judicial overreach.[16] However, completely restricting AI systems from accessing public digital content could hinder domestic technological innovation.[16]
The most balanced solution is legislative rather than purely judicial.[16] The Indian Parliament should introduce a specific text and data mining exception to the Copyright Act, balanced by a machine-readable opt-out mechanism for creators, mandatory data transparency requirements for AI developers, and a statutory licensing framework for commercial training datasets.[16] While the Delhi High Court’s upcoming ruling on interim relief will provide an initial indication of how domestic courts view these challenges, comprehensive legislative reform remains essential to resolve the regulatory uncertainty at the center of India’s AI governance landscape.[16]
Bibliography
Primary Statutory Sources
[16] The Copyright Act, 1957, No. 14 of 1957 (India).
[16] The US Copyright Act, 17 U.S.C. § 107 & § 505 (United States).
[16] The Copyright, Designs and Patents Act, 1988, c. 48, § 29A (United Kingdom).
[3] Directive (EU) 2019/790 of the European Parliament and of the Council of 17 May 2019 on copyright and related rights in the Digital Single Market, Art. 4.
Judicial Precedents
[1] Super Cassettes Industries v. Chintamani Rao, (2011) 48 PTC 530 (Del).
[2] Syndicate of the Press of the University of Cambridge v. B.D. Bhandari, (2011) 47 PTC 244 (Del).
[16] Asian News International v. OpenAI Inc. & Anr., CS(COMM) No. 954/2024 (Delhi High Court, Order reserved on March 27, 2026).
[16] The New York Times Company v. Microsoft Corporation, OpenAI Inc., et al., Case No. 1:23-cv-11195 (S.D.N.Y.).
