ANI vs. OpenAI: Catalyst For Indian Copyright Regime

If your business is building, operating, investing in or deploying AI systems particularly those that rely on large language models (LLMs), the ongoing case of Asian News International (ANI) v. OpenAI¹ before the Delhi High Court could profoundly impact your operational, strategic and legal risk landscape.

This is not just another copyright dispute. It's a litmus test for how India will regulate AI's access to and use of proprietary content with implications that stretch beyond territorial jurisdictions. The AI models trained on third-party content or those entities relying on such tools, understanding this case and its emerging jurisprudence is no longer optional. This case is set to significantly influence the Indian copyright regime and its persisting lacuna regarding AI training data and until future legislation, set precedent for what AI businesses can and cannot do.

What You Need to Know

The Delhi High Court's ongoing proceedings in Asian News International (ANI) v. OpenAI have brought to the forefront an unprecedented question for Indian jurisprudence: Does the training of generative AI models on copyrighted news content amount to copyright infringement? Filed in November 2024, this suit is India's first significant legal challenge involving artificial intelligence, copyright law and the doctrine of fair dealing under the Copyright Act, 1957. The case resonates with parallel lawsuits globally like The New York Times v. OpenAI² in the US and tests how Indian courts will balance the rights of copyright holders with the evolving demands of AI innovation.

ANI argued that OpenAI had scraped and stored large volumes of its news material to train its large language models (LLMs) and produced outputs based on copyrighted news content, thereby infringing its exclusive rights under Section 14 read with Section 2(m) of the Copyright Act.

ANI alleged two primary harms:

1. Unauthorized reproduction and storage of its copyrighted content for AI training.
2. Generation of outputs by ChatGPT that either reproduced or misrepresented ANI's reports, causing reputational and economic harm.

OpenAI responded by challenging the Delhi High Court's jurisdiction, asserting that it neither maintained servers in India nor directly carried out infringing acts within the country. It also argued that AI training involves the transformation of data into machine-readable formats and does not amount to infringement, further relying on the doctrine of fair dealing under Section 52 of the Act. At the interim stage, the Delhi High Court declined to grant an injunction restraining OpenAI from using ANI content but issued summons to the defendants. The matter is now pending for further hearing.³

Why This Matters for AI Businesses Globally

The case raises several legal and procedural questions:

1. Whether the storage and reproduction of ANI's content for AI training constitutes copyright infringement.
2. Whether outputs generated by ChatGPT based on ANI content amount to derivative works or infringing reproductions.
3. Whether OpenAI's use can be justified under India's fair dealing exceptions.
4. Whether the Delhi High Court has territorial jurisdiction over a foreign defendant with no servers or physical presence in India.

Analysis

1. Storage and Reproduction by AI Models

Under Section 14(a)(i) of the Copyright Act, copyright includes the exclusive right to "reproduce the work in any material form" and explicitly states "storing of it in any medium by electronic means" emphasised in Tips Industries Ltd. v. Wynk Music⁴. The central question is whether storing text data to train AI constitutes reproduction.

ANI, relying on MySpace Inc. v. Super Cassettes Industries⁵, argues that persistent storage enabling reuse or transformation goes beyond transient storage and thus amounts to reproduction.

OpenAI counters that data used in AI models being a type of Large Language Model (LLM) becomes part of statistical weights during processing which is suitable for machine learning and is therefore, no longer retaining human-readable form. This falls in line with the argument that storage incidental to technical processes or transformation is not infringement.

The Delhi High Court must decide whether transforming human-readable text into coded or statistical representations retains the character of the original work for copyright purposes which remains a question untested in the Indian copyright regime.

2. Outputs and Derivative Works of AI Models

ANI's claim that ChatGPT outputs reproduce or misrepresent its news reports raises a complex doctrinal question: do AI-generated outputs amount to derivative works under Indian copyright law? Under Section 14(c) of the Copyright Act, copyright owners hold the exclusive right to make adaptations and derivative works. If the AI outputs closely imitate the expression, style or distinctive presentation of ANI's reports, ANI could argue that this amounts to infringement.

However, if these outputs merely summarize facts or provide paraphrased versions of stories, they may fall outside the scope of copyright, given the idea–expression dichotomy which excludes protection for facts and general information.

An added layer is the phenomenon of AI hallucinations, instances where ChatGPT generates false statements wrongly attributed to ANI. While these could damage ANI's reputation and raise potential claims in defamation, they do not automatically amount to copyright infringement because they are not direct reproductions of ANI's copyrighted expression. The court will need to determine whether these AI-generated outputs cross the line from permissible paraphrase or summary into infringing adaptation, a question that sits at the cutting edge of both copyright and AI jurisprudence.

3. Fair Dealing and Lack of TDM Exception in Indian Regime

OpenAI has relied on the exception of fair dealing u/s Section 52 of the Copyright Act. However, OpenAI like several other AI models uses the practice of Text and Data Mining (TDM) and Section 52 does not explicitly cover TDM for AI models. The Indian jurisprudence in this regard has been inclined to incorporate US principles in a manner fitted to the Indian scenario.

The US Supreme Court in Sony Corp of America v. Universal City Studios Inc.⁶ held that "copying a news broadcast may have a stronger claim to fair use than copying a motion picture. This is because copying from informational works encourages the free spread of ideas and encourages the creation of new scientific or educational works, all of which benefit the public." A similar ruling came by the Delhi High Court in Akuate Internet Services Pvt. Ltd. v. Star India Pvt. Ltd.⁷ stating that "even if a work is protected by copyright, the facts and information within it cannot be exclusively controlled." The court stressed that limiting access to such information would restrict the public's ability to share and use knowledge. It also pointed out that information about current events is a report on facts that are already known to the public, essentially the history of the day, rather than something that is created by an author.

Despite the rulings, the absence of an established statutory TDM exception puts AI developers on uncertain ground in India.

4. Jurisdiction of Indian Courts

To address the issue of jurisdiction, ANI relied on Section 62 of the Indian Copyright Act, 1957 which states that legal proceedings can be instituted in the respective forum where the person instituting the suit resides or carries on business. However, OpenAI contended that as per Section 20 of the Code of Civil Procedure suits are required to be filed either where the defendant resides or where the cause of action arises. Since, OpenAI neither operates its servers nor conducts the training and generation as alleged in India, Delhi High Court has no jurisdiction to which ANI countered that ChatGPT's services are accessible in India, and the harm occurs within India. Delhi High Court precedent in Telegram LLC v. IPRS⁸ suggests courts may assume jurisdiction if harm is localized, even if servers are abroad.

The court's future ruling could clarify whether global AI platforms can be sued in India based on accessibility and impact alone.

What Global AI Businesses Should Track

The stance towards TDM across the globe is varied and at very different stages individually but this should be seen in positive light because the stagnation towards such unexplored areas of technology and law is no longer persisting.

In the United States, the fate of TDM's legal nuances hangs on the verdict of the ongoing case of New York Times v. OpenAI. The EU has codified the exception of TDM with the Directive on Copyright in the Digital Single Market, 2019⁹ and the AI Act. India runs parallel to the US in this regard with little or no clarity in statutes yet and a lot is left on jurisprudential interpretation. The UK government took a step towards bringing a broad TDM exception but post the severe backlash has withdrawn it. China's Cyberspace Administration has issued draft regulations for generative AI services in 2023 placing compliance burden on AI businesses. Japan has one of the most AI friendly regime wherein the copyright law has been interpreted in order to incorporate TDM exception.¹⁰

Strategic Takeaways for AI Businesses

i. Move from "assumed fair use" in the direction of explicit licensing strategies

Courts globally have not yet taken a definitive stance on scraping large amounts of data for training falls under the ambit of fair use or not. The verdict of ANI v. OpeniAI and similar disputes in the US and elsewhere will clarify to some extent whether this assumption is safe or not. The businesses should prepare their strategies for both scenarios and recognise that while negotiating explicit content licenses requires upfront investment, it will drastically reduce litigation risk.

ii. Conduct periodic AI model risk audits

As AI models evolve, so do the legal and reputational risks they carry when built on large and often opaque datasets. AI businesses should implement structured, periodic risk audits that combine technical review and legal analysis to examine what data the model ingested, whether it includes copyrighted or sensitive material and how outputs might recreate or misrepresent protected content. These audits should also assess potential for hallucinations, defamation or biased outputs that could lead to claims beyond copyright. Regular risk audits not only help businesses detect and remediate issues before they escalate into litigation but also demonstrate due diligence to regulators, investors and clients positioning the company as a responsible AI business.

iii .Track evolving jurisprudence

The legal landscape around AI is shifting rapidly, not just in India but worldwide. Indian courts, through cases like ANI v. OpenAI, are grappling for the first time with foundational questions around AI training, fair dealing and output liability. At the same time, international precedents from the New York Times v. OpenAI lawsuit in the US to TDM exceptions around the world are likely to influence local thinking. AI businesses operating in or targeting the Indian market should systematically monitor these developments, through internal legal teams or specialized advisors, to anticipate new obligations or risk exposures. Staying informed isn't just about compliance; it's a strategic advantage that enables companies to adapt business models, refine licensing strategies and engage constructively with policymakers before rules harden into law.

iv. Anticipate jurisdictional reach

A critical and often overlooked risk is that AI companies can face lawsuits in jurisdictions where their products or services are accessible even if they have no physical offices, servers or employees there. Indian courts, like many globally, are increasingly willing to assert jurisdiction based on the effects doctrine. For AI businesses, this means proactively assessing where their models accessible. Understanding and planning for this expanded jurisdictional reach isn't merely defensive, it's essential for responsibly scaling AI offerings into large emerging markets like India while avoiding unexpected legal entanglements that could disrupt growth or investor confidence.

v. Enhance transparency and user controls

The dataset used to train AI models is opaque in the sense that the data used to train is kept hidden from the users. In an environment where AI outputs are increasingly scrutinized for accuracy, provenance and potential infringement, transparency and user empowerment have become essential rather than optional. AI businesses should move beyond generic disclaimers to provide clearer information about how models are trained, what types of data were used, and where licensing has or hasn't been secured.

The Bottom Line

A ruling against OpenAI could mean AI firms must change how they source and use data, create costly licensing processes or even limit certain AI functionalities in India. It may also encourage more content owners to file similar claims.

ANI v. OpenAI presents the Delhi High Court with an opportunity to articulate the scope and limits of copyright law in the context of AI-driven technologies. Beyond the immediate dispute, the case raises nuanced questions which were long due to be deliberated upon.

For AI businesses, this case is more than an ordinary legal dispute as its verdict could set modified compliance obligations, reshape technological processes and affect market strategies. Staying updated on ANI v. OpenAI therefore, is imperative for relevant risk management, strategic planning of future operations and staying legally compliant in the extremely dynamic AI sector.

Footnotes

1. CS(COMM) 1028/2024)

2. The New York Times v. OpenAI (S.D.N.Y July 16, 2025).

3. https://www.barandbench.com/news/chatgpt-data-not-stored-india-copyright-act-does-not-apply-openai-to-delhi-high-court-ani-suit