OPINION | Evaluating India’s generative AI copyright framework and the DPIIT proposal

3) Centralised Collection Entity: A centralised non-profit entity named Copyright Royalties Collective for AI Training (CRCAT) is responsible for the collection and distribution of royalties. This entity would have Copyright Societies and Collective Management Organisations (CMOs) as members (one member for each class of works), providing a single umbrella organisation for administration of copyrighted works and royalty payments.

4) Royalty Distribution: Member organisations would allocate royalties to both members and non-members who register their works for the purpose of licensing for AI training.

5) Rate Setting: Royalty rates would be fixed by a committee appointed by the government, subject to judicial review. The royalty rates may vary based on the revenue generated from the monetisation of the AI tool and the usage of the copyrighted work for the same.

6) Revenue Share Model: The royalty would be payable annually on a recurring basis rather than as an upfront payment. The rationale is to ease access to licensing of the content, thereby making it possible for AI developers to access all requisite content for training AI systems without first having to bear the burden of upfront payments. This approach avoids creating barriers for start-ups in India while ensuring ongoing compensation.

Key Observations:

While the model aims to encourage innovation and, at the same time, protect copyrighted works, the approach raises fundamental issues, briefly discussed below. The model is also based on multiple assumptions with a large gap between evidence and the realities of how enforcement of copyright operates in practice.

I. Uneven Playing Field for Creators

By mandating a blanket licence and offering royalties only to registrants, the model effectively takes away an owner’s ability to withhold their work from use for training AI systems or negotiate royalty rates individually. This approach creates a natural imbalance, where the lack of clarity in determining royalty rates is likely to discourage owners from being part of the proposed ecosystem. Royalties would flow through CMOs and copyright societies, but India’s experience with collective management has been mixed. Smaller authors, independent artists, and those outside organised sectors may struggle to benefit. Since only registered works qualify for royalties, unregistered authors would suffer and would be compelled to register themselves, with uncertainty as to when they would receive royalties. This undermines the principle of equal protection and fails to offer a level playing field across diverse creative communities.

II. Environment for Smaller AI Developers

The model proposes payment of royalties tied to a percentage of global revenues earned by AI companies. However, instead of fostering a competitive ecosystem, the model risks consolidating power among big players, reducing diversity and innovation in the AI sector. Unlike large global firms, which can easily absorb high royalties and compliance overheads, smaller bootstrapped companies may find themselves struggling for margins in an unclear environment. A blanket licence may seem like it reduces legal risks on paper but adds an additional layer of compliance. For smaller Indian AI start-ups, this means paperwork and reporting, which can slow them down and strain budgets.

III. Royalty Rate Determination

The model envisions “flat” royalty rates set by a government-appointed committee. While simple in theory, this approach risks mispricing. High rates could stifle AI innovation, while low rates may undercompensate creators. More nuanced rationale is needed to balance interests, taking into account the diversity of both creators and AI developers. Without careful rate-setting, disputes over fairness and adequacy of compensation are inevitable, potentially burdening courts rather than reducing litigation. There is also heavy reliance on self-disclosure by AI developers, which raises transparency, enforcement, and audit concerns.

Personality Rights Left Unaddressed

The paper focuses narrowly on copyright, leaving personality rights such as likeness, voice, and identity outside its scope at this moment. As generative AI increasingly replicates human attributes, the absence of clear rules on personality rights, membership of persons who can license the personality right, and royalties for the same creates a significant gap. Without guidance, disputes over misappropriation of personal attributes are likely to arise, adding another layer of complexity to the legal landscape.

The DPIIT’s Working Paper is a welcome acknowledgment of the need to balance innovation and creative rights. However, while the proposed framework may provide some structure and predictability, new questions have emerged for policymakers’ consideration. The hybrid licensing model raises concerns around infringement, dilution of rights, royalty determination, and personality rights. Unless these gaps are addressed, the framework risks being challenged both in courts and in practice, potentially undermining the certainty it aims to provide. It remains to be seen whether Part II of the series (scheduled to deal with the copyrightability of AI-generated outputs) brings more clarity on the highlighted concerns to offer a balanced and sustainable framework.

Disclaimer: This article is purely for informational purpose and does not constitute legal advice or opinion by Khaitan Legal Associates.

(Boski Sharma, Partner, Khaitan Legal Associates.)

Views are personal, and do not represent the stand of this publication.