Back
Does Fair Dealing Shield AI Training? Delhi High Court Seeks OpenAI Reply on ANI Plea
The has called upon to respond to a plea by news agency seeking an against the use of its copyrighted content to train and operate ChatGPT. The development marks a significant escalation in the legal battle over the boundaries of copyright protection in the age of artificial intelligence, with the court now poised to examine whether the " " exception under Indian law can shield large-scale of news articles by AI systems.
, the first Indian media house to sue , originally filed its suit in , alleging unauthorised reproduction and storage of its content. In , a of the High Court, presided over by Justice Amit Bansal, refused to grant , holding that an injunction would be against and that 's storage of 's material for training did not appear to constitute . subsequently challenged that interim order before a , which has now sought 's reply.
The Core Legal Question: Fair Dealing vs. Fair Use
At the heart of the dispute lies a fundamental difference between Indian and US copyright law. India's , incorporates a " " framework under Section 52, which provides a closed list of permitted purposes such as , , , and . In contrast, the United States operates under a broader " " doctrine, which allows a more flexible, case-by-case analysis of factors including the purpose and character of the use, the nature of the work, the amount used, and the effect on the potential market.
's counsel has argued that the single judge's reliance on a
‑style balancing test was legally erroneous. As the news agency submitted,
"India has
, not
, a
in Section 52 of the Copyright Act and that reading
at scale into '
, including
' widens that exception beyond what Parliament wrote."
This contention underscores a critical tension: can an AI company that ingests millions of copyrighted articles for training, and then generates revenue from that system, ever claim that its use falls within the narrow confines of "
" or "
"?
Public Interest and the Licensing Paradox
The single judge had reasoned that granting an injunction would harm AI development and the broader
. However,
has pointed out an apparent contradiction in
's own position. The agency noted that if
genuinely believed its use of news content was covered by
, there would be no need to enter into licensing agreements with publishers such as the
.
"If the use is
, it is hard to see what those licences are for and the existence of that market is precisely what fairness analysis is meant to weigh,"
argued.
This is central to the case. , like many AI developers, has struck paid deals with several news organisations to obtain content for training and, in some cases, to display extracts in ChatGPT responses. contends that the very existence of such commercial licences demonstrates that the use of copyrighted news material is not free from copyright liability, and that 's failure to license 's content cannot be excused by a broad reading of .
The Quantifiable Stakes: A Tiny Fraction or a Systemic Threat?
Another twist in the case is the scale of the alleged infringement.
has pleaded that
's content constitutes a "minuscule fraction" of its overall training data.
, however, seizes on this admission to argue that if the use is indeed so small, then an injunction would cause no
to
. The agency submitted that the single judge's finding that an injunction would harm AI development
"sits uneasily with
's own pleading that
's content is a minuscule fraction of its training data."
For legal professionals, this argument raises a nuanced point about the of interim remedies. If the allegedly infringing material is in the context of a vast dataset, the may tilt in favour of the plaintiff, especially when the plaintiff's copyright in each individual article is clear. Conversely, if every copyright owner of a tiny fraction were entitled to an injunction, the cumulative effect could stifle AI development—a concern that the single judge apparently found persuasive.
Implications for AI Training and Copyright Practice
The outcome of this appeal could have far‑reaching consequences for the generative AI industry in India. If the ultimately accepts 's interpretation of Section 52, it may force AI companies to obtain licences for every copyrighted work used in training, significantly increasing costs and legal uncertainty. On the other hand, a broad reading of could allow AI developers to continue using publicly available news content without compensation, potentially undermining the business models of news publishers.
For copyright practitioners, the case serves as a reminder of the importance of jurisdictional differences. While US courts have largely allowed AI training under the doctrine—subject to ongoing litigation—Indian courts must apply a more rigid statutory framework. The 's eventual ruling will likely provide guidance on how Section 52's should be interpreted in the context of .
What Lies Ahead
The has now directed to file its reply, setting the stage for a detailed hearing on the . The court will need to weigh the competing interests of copyright protection, innovation, and public access to information. It will also have to address the fundamental question of whether the exception can ever apply to the commercial training of AI models, or whether such activities require a legislative amendment.
As the legal community watches closely, the ‑ case promises to shape the contours of copyright law in the age of artificial intelligence. The 's decision—whether it grants an or allows the single judge's order to stand—will send a powerful signal to both media companies and AI developers about the legal risks and obligations inherent in training large language models on copyrighted news content.
Conclusion
The 's call for 's reply is the latest chapter in a landmark copyright battle that could define the legal framework for AI training in India. With the doctrine at its core, the case forces a re‑examination of how traditional copyright exceptions apply to novel technologies. For now, 's challenge to the single judge's order remains pending, and the legal community awaits a reasoned decision that will likely influence future disputes across the globe.
