Copyright in AI era: Deeper judicial, legislative scrutiny is needed

Artificial intelligence (AI) has revolutionised the media and information industry, raising hot-button issues such as copyright and sanctity of content creation. The Delhi High Court’s July 24 decision dismissing ANI Media’s application for an interim injunction on OpenAI offers, for the first time, a broad framework for the judiciary to approach the issue in India. In 2024, ANI accused the San Francisco-based AI research organisation of scraping, storing and using ANI copyrighted news articles to train the large language models (LLMs) underlying ChatGPT, reproducing large portions of ANI’s content while generating responses to user prompts. The key issues emerging from this interim order concern the territorial jurisdiction of such disputes and interpretations of Section 52 of the Copyright Act. OpenAI argued that LLM training occurred abroad, and hence applying the Indian Copyright Act to this issue would amount to an impermissible extension of Indian law. Justice Amit Bansal denied this framing on the grounds that ANI’s principal place of business was India and OpenAI also offered and monetised its services in this country. Therefore, Indian courts possess the territorial jurisdiction under both the Copyright Act and the Code of Civil Procedure to entertain this suit. Merely storing the data on servers located in the United States, the court argued, could not insulate the company from Indian law.

 

The substantive issue concerned the applicability of “fair dealing” under Section 52 (1)(a), for which the court has adopted an ample interpretation. ANI had contended that OpenAI was using commercially operated AI systems and should be excluded from the “fair dealing” clause — which states that any work used for (a) private or personal use, including research; (b) criticism or review; or (c) reporting current events will not constitute an infringement of copyright. The court rejected this contention on the grounds that the specific requirement of non-commercial use was absent in the Section concerned. On whether “private use” could extend beyond individual human beings, the court applied the “doctrine of updating construction”, which states that ongoing law should be updated to fit modern facts, technology and social changes. On that basis, the court said, first, the term “private” could apply to private companies and, second, that since LLM training is a form of research and because the training dataset is never made available to the public, its use is “private”. The court outlined three “fairness” factors in determining its decision. One, OpenAI had not appeared to have used ANI content for anything other than training its LLMs. Second, ChatGPT’s functions are fundamentally different from ANI’s business of news syndication. Third, the court pointed to the substantial public benefits of LLMs in health care, education, financial services and so on. 

 

Given the proliferation of parallel cases globally, this decision will be read closely. At the very least, it narrows content producers’ ability to monetise output and raises the spectre of a handful of billionaire founders of AI majors dominating the global information industry. Though this pronouncement is AI-leaning as far as training and research are concerned, it leaves open to interpretation substantive issues of principles and ethics concerning the relationship between owners of AI products and content creators. That demands deeper judicial and parliamentary scrutiny.

 

Disclosure: The Digital News Publishers Association (DNPA), of which Business Standard is a member, is a party in the ANI vs OpenAI case