ANI v. OpenAI: CMS INDUSLAW Partner Bharadwaj Jaishankar Explains What the Landmark AI Copyright Case Means for India
As the Delhi High Court hears the landmark ANI v. OpenAI dispute, the case has become India’s first major legal battle over whether AI models can be trained using copyrighted content.
In this interview, Bharadwaj Jaishankar, Partner at CMS INDUSLAW, discusses the key copyright issues before the judiciary, the likely impact on AI developers and publishers, governance measures businesses should adopt, and how India can balance intellectual property protection with AI-driven innovation.
The ANI v. OpenAI case has brought the issue of AI training on copyrighted content before Indian courts for the first time. From a legal standpoint, what do you see as the core question the judiciary will ultimately have to answer?
The legal question that the judiciary will eventually have to answer in the ANI v. OpenAI case is one that has been grappled with by various courts across different jurisdictions. In the Indian context, it would simply be whether use of such copyrighted material for training Large Language Models (LLMs) falls within the ambit of the copyright law in India, i.e., whether it constitutes copyright infringement or the existing exceptions under Section 52 of the Copyright Act could apply. In addressing this, the court is likely to consider several related and key questions. As basic as whether legality should be judged based on the training process itself, or on whether the model generates outputs that reproduce or closely imitate protected expression? That said, while these stand to be fundamental legal issues on the subject, an important consideration before the judiciary is also to strike a careful balance between the interests of legitimate right holders and the broader public interest in fostering innovation and supporting technological advancement.
Whatever the outcome of this case, how do you think it will influence the relationship between AI developers, publishers, media organisations, and content creators in India?
The more lasting effect of this case is the potential to reframe the conversation between AI developers and the content industry from one of confrontation to one of negotiation. The interim ruling suggests that an injunction is a difficult remedy for a publisher to secure. The burden lies on the rightsholder to demonstrate memorisation or substantial reproduction, and the court took note of the fact that ANI had an opt-out available to block scraping and had not used it. For publishers and media organisations, this could make litigation a less attractive solution and a licensing conversation a more realistic one.
The relationship would also become more transactional and increasingly collective. Bilateral licensing deals of the kind now common globally will continue. At the same time, Indian publishers and industry bodies may bargain as blocs to strengthen their position, while developers have their own commercial and reputational incentives to license reliable, high-quality Indian content. We can also expect greater attention to a rightsholder’s use of practical measures such as opt-out protocols, terms of use, and anti-scraping measures, particularly where infringement concerns arise, as these can affect bargaining power. This could lead to a more structured relationship between AI developers and content owners, where licensing, transparency, and technical controls become central to future collaboration.
As businesses rapidly integrate AI into customer service, research, and internal operations, what governance measures should organisations prioritise today to minimise future legal exposure?
On the copyright front, liability can potentially arise at two levels: first, when users feed proprietary data into a model, and second, when AI-generated outputs reproduce protected content. At the training stage, businesses should verify that all datasets used to build or fine-tune AI models have been lawfully obtained and properly licensed. Public availability alone does not authorise unrestricted use, particularly where copyrighted material is involved. Documenting data provenance and examining licensing terms can therefore reduce the risk of infringement claims arising both from model training and from downstream AI-generated outputs.
Additionally, organisations should maintain clear records of the human contributions made throughout the creation of AI-assisted works. Documenting the nature and extent of human involvement may strengthen the case for copyright protection, which depends on demonstrable human authorship. Maintaining this documentation can also assist in establishing ownership, substantiating claims of originality, and preventing unauthorised third-party use of the work.
AI-generated outputs that incorporate third-party logos or taglines may also raise trade mark infringement concerns. Such outputs could amount to unauthorised use and create consumer confusion, especially when used in advertising or other public-facing platforms. Human oversight of AI-generated outputs can also help mitigate these risks.
Establishing internal policies on AI use for employees is also a prudent step. Such policies help raise employee awareness about the appropriate use of AI tools and reduce the risk of employees inadvertently disclosing trade secrets or other proprietary information.
India is simultaneously trying to become an AI innovation hub while strengthening digital regulation. How can policymakers strike the right balance between protecting intellectual property and encouraging technological innovation?
India’s challenge is to protect intellectual property without slowing the development of a competitive AI ecosystem. Tipping the balance too far in either direction will yield unfavourable results. Over-protecting IP could make AI development expensive and weaken India’s innovation ambitions. Under-protecting it could erode the incentives that allow journalists, publishers, authors, filmmakers, and other creators to keep producing original work.
The Department for Promotion of Industry and Internal Trade’s (DPIIT) Working Paper on Generative AI and Copyright suggested a centralised mandatory licensing mechanism through which AI developers could acquire rights to copyright-protected material without negotiating thousands of bilateral agreements. The framework would also ensure that revenue flows back to creators. This is a promising starting point, but any mandatory licensing system must be designed carefully. It should not strip creators of control over their works or reduce licensing to a mere formality. The draft was opened for public and stakeholder consultation, and those responses will be critical in shaping the final approach.
In addition, the Ministry of Electronics and Information Technology’s India AI Governance Guidelines advocate a trust-based, human-centric, sustainable, and accountable approach to AI development. Policymakers should ensure that these guiding principles inform the AI and intellectual property framework.
Ultimately, the goal should not be to make AI training impossible or prohibitively expensive. Instead, it should be to ensure accountability and lawful use. India can lead in AI only if innovation rests on a framework that creators, publishers, developers and users can all trust. Thus, what is required is a system that is transparent, proportionate, and commercially viable.
