The AI Mandate: India’s New Blanket Licensing Framework

The Indian intellectual property landscape is on the cusp of a seismic shift, poised to fundamentally reshape how artificial intelligence (AI), particularly generative AI, interacts with copyrighted content. Moving beyond the protracted and often ambiguous debates around “fair use” or “fair dealing” that have dominated headlines and courtrooms in the West, India, with its ambitious goals for a $1 trillion digital economy and its globally recognized, robust creative sector, is charting a bold new, proactive course. This initiative is best described as The AI Mandate.

At the immediate heart of this mandate is the DPIIT’s Part I Working Paper on Generative AI and Copyright, released in December 2025 by the Department for Promotion of Industry and Internal Trade.[1] This document is far more than just another policy paper for public consultation; it is a foundational blueprint for a Mandatory Blanket Licensing (MBL) framework. Should it be enacted, this framework would fundamentally alter the legal and economic landscape for both AI developers and content creators in India.

The MBL framework proposes to eliminate the need for AI companies to individually negotiate licenses with millions of copyright holders for every piece of data used to train their models. Instead, it envisages a central, collective licensing body—likely an expansion of existing Copyright Societies—that would grant a single, mandatory, non-negotiable license covering all copyrighted works in India for the sole purpose of AI training.[2] In exchange, AI companies would pay a pre-determined, standardized royalty fee into a central pool, which would then be distributed to the original creators based on usage metrics. This ambitious mechanism is designed to strike a crucial balance: ensuring the smooth, unencumbered growth of India’s domestic AI industry by providing legal certainty and access to vast datasets, while simultaneously guaranteeing content creators—from musicians and artists to authors and news publishers—equitable and systematic compensation for the use of their work as training data. This move signals a definitive rejection of the ‘move fast and break things’ approach to intellectual property in the age of generative AI, opting instead for a structured, compensatory ecosystem.

The Paradigm Shift: From Ambiguous “Fair Dealing” to Mandatory “Blanket Licensing”

The DPIIT’s (Department for Promotion of Industry and Internal Trade) recent white paper marks a definitive pivot in India’s copyright jurisprudence concerning Artificial Intelligence.[3] It squarely addresses the long-standing, globally debated question of whether the massive-scale data ingestion required for training AI models—often referred to as ‘scraping’—can legally be accommodated within existing copyright exceptions.

The Inadequacy of “Fair Dealing”

For several years, the Indian legal community explored the possibility of leveraging Section 52(1)(a) of the Indian Copyright Act, 1957, to permit AI training.[4] This section outlines the principle of “fair dealing,” which allows for the limited use of copyrighted material without permission for specific purposes, such as:

• Research: Non-commercial, scholarly investigation.
• Criticism or Review: The use of content to analyze or comment on a work.
• Private or Personal Use: Non-commercial use by an individual.

The government’s analysis, however, delivers an unambiguous “No” to including AI training under this umbrella.[5]

The Commercial and Economic Rationale Against “Fair Dealing”

The rejection is not merely a legalistic quibble but is rooted in a pragmatic understanding of the commercial realities of modern Generative AI. The argument against “fair dealing” rests on two crucial pillars:

1. Commercial Scale of Use: AI training is not a private or limited research activity. It is a mass-scale, industrial process involving the ingestion of petabytes of data, often conducted by well-funded, commercial entities. The scale of this operation fundamentally disqualifies it from being considered “fair.”
2. Substitutive Nature of Output: The primary function of generative AI is to create new content (text, images, code, music) that often directly substitutes the original copyrighted works it was trained on. This creates AI-generated alternatives that compete directly in the marketplace with the economic output of human creators, thereby directly violating the economic rights the Copyright Act is designed to protect. If an AI can generate a commercially viable alternative to an author’s book or an artist’s image, the economic incentive for the creator diminishes significantly.

The Proposed Solution: “One Nation One License One Payment”

In place of the ambiguous “fair dealing” exception, the DPIIT paper champions a bold, centralized licensing framework, succinctly termed the “One Nation One License One Payment” model.[6] This statutory blanket licensing scheme is designed to regularize and monetize the use of copyrighted content for AI training.

Key Features of the Blanket Licensing Framework:

• Centralized System: This model envisions a single, government-backed or designated non-profit collective management organization (CMO) that would act as a clearinghouse for licensing copyrighted works for AI training.
• Statutory Fee Mechanism: AI developers and companies would pay a single, fixed, or formula-based statutory fee to this central body. This payment grants them legal access to a vast, pooled repository of licensed content for their model training.
• Presumption of Use (Opt-In for Creators, Presumed for Users): Crucially, the Indian proposal is distinct from the “Text and Data Mining (TDM)” exception seen in regions like the European Union. The EU model is typically an “opt-out” system, where copyright holders must actively flag or digitally mark their content to prevent AI scraping. In contrast, the Indian framework posits that the use of content for AI training is presumed, provided the content has been “lawfully accessed.”[7] This places the onus on the user (the AI developer) to ensure they are using legally obtained data, but removes the friction of piece-meal, individual negotiation.

This blanket licensing approach aims to strike a balance: ensuring AI innovation is not stymied by complex licensing hurdles, while guaranteeing that creators receive collective, statutory remuneration for the value extracted from their work by commercial AI systems.

The CRCAT: India’s New IP Powerhouse? A Deep Dive into the Proposed Mechanism

To successfully administer the unprecedented AI Blanket Licensing Framework, the Department for Promotion of Industry and Internal Trade (DPIIT) has put forth a crucial proposal: the establishment of the Copyright Royalty Collection and Allocation Tribunal (CRCAT).[8] This new statutory body is envisioned as the indispensable linchpin of the entire system, a mechanism designed not just to collect funds but to fundamentally redefine the relationship between AI innovation and creator compensation.

The CRCAT’s mandate is multi-faceted, encompassing two primary, yet profoundly complex, functions:

1. Global Royalty Collection: Securing Value from AI

The first and arguably most revolutionary task of the CRCAT is the Collection of Royalties. The DPIIT’s suggestion outlines a model where the Tribunal will be responsible for determining and collecting a mandatory percentage of an AI company’s global revenue as a royalty for utilizing copyrighted material in its training data or output generation.

This is a crucial distinction that elevates India’s framework beyond a typical domestic licensing scheme. By pegging the royalty to global revenue—and not merely revenue generated within India—the government signals a robust intent to capture the economic value created by AI activities worldwide, irrespective of where the AI company is headquartered. This provision aims to prevent large multinational AI developers from circumventing the framework by simply ring-fencing their Indian operations. The underlying principle is that if an AI model, trained on data that includes Indian copyrighted works, generates global profits, a portion of those profits should rightfully flow back to the original creators.[9] The CRCAT, therefore, is not just a national collection agency, but is positioned to become a significant international player in the intellectual property rights landscape.

2. Allocation and Distribution: The Immeasurable Challenge of Provenance

The second, and perhaps most intricate, function is the Allocation of Funds—the distribution of the collected royalties back to the rightful creators. While the concept is simple in theory, the practical implementation presents a formidable, unprecedented technological and logistical hurdle.

This is where the complexities truly begin. The CRCAT’s process must somehow determine:

• Attribution: Which specific creator’s copyrighted work contributed to the training of a large language model (LLM), a generative image model, or any other AI system?
• Contribution Weight: How much did that specific work contribute to the final, commercial output generated by the AI?
• Fair Compensation: Based on that contribution, what is the appropriate share of the collected royalty pool?

The challenge of “data provenance” will be immense. Modern frontier AI models are trained on datasets often comprising billions—and sometimes trillions—of data points, scraped from the open web, digitized libraries, and various proprietary sources.[10] Tracing a specific phrase in an AI-generated article, or a particular style element in an AI-generated image, back to its original source in the massive training corpus is currently a near-impossible feat. The CRCAT would need to pioneer or adopt novel, scalable auditing technologies—perhaps cryptographic proofs of inclusion or advanced forensic AI tools—to create a system of transparent and auditable attribution. Without such a technological solution, the allocation process risks becoming arbitrary, leading to disputes, and potentially undermining the entire framework’s goal of fair compensation. The success of the CRCAT hinges not just on its legal and financial mechanisms, but fundamentally on its ability to solve this massive data provenance puzzle.

The “Lawfully Accessed” Caveat: A Critical Safeguard Against Piracy in AI Training

While the overarching goal of the Model Blanket Licensing (MBL) framework is to drastically streamline and legitimize the process of training Artificial Intelligence models, a crucial and potentially restrictive detail lies within the “lawfully accessed” clause. This seemingly simple phrase acts as a fundamental safeguard, placing a significant burden of due diligence on AI developers and setting a clear boundary against the wholesale legitimization of pirated content.

The primary implication is that the MBL is not a universal ‘get-out-of-jail-free’ card for AI training data sourced from illicit repositories. AI developers cannot operate under the assumption that the blanket license retrospectively legalizes the use of content procured from illegal or unauthorized platforms. Specifically, this clause directly prohibits the incorporation of material hoovered up from well-known pirated sources—such as academic content from LibGen or large-scale book corpora like Books3—even if the content itself might be theoretically covered by the MBL if it were accessed through legitimate means.

Liability and Disincentivization:

The “lawfully accessed” requirement dictates that the source and method of data acquisition must be legitimate. If an AI development team trains its model on a corpus of data where the origin is demonstrably illicit (i.e., a pirated dataset), the blanket license offers no protection. In such a scenario, the AI developer remains fully liable for copyright infringement, even if they have paid the MBL fee.[11] This dual-pronged approach is a deliberate policy choice, designed to strike a delicate and essential balance:

1. Legitimization of AI Training: By offering a streamlined license, the MBL encourages innovation and provides a clear legal path for ethical AI development.
2. Disincentivizing Piracy: By tying the license’s validity to the legality of the acquisition, the framework actively discourages the creation, distribution, and use of pirated content as a cost-effective shortcut for AI data pipelines. It transforms the legal risk associated with using pirated data from a gray area into a clear-cut liability.

Operational Challenges for Developers:

For AI developers, this clause introduces a critical operational and compliance challenge. They are no longer just responsible for the content of their training data but must also maintain rigorous provenance tracking and audit trails for how that data was acquired. This necessitates robust internal processes to vet data sources, confirm licensing compliance upstream, and ensure that data procurement methods align with all applicable laws. Any failure to do so converts the MBL from a shield of protection into a potential paper-thin defense against copyright litigation.

 The AI Blanket Licensing Framework of India: Significance, Stakeholder Implications, and Global Context

The proposed AI Blanket Licensing Framework, an initiative spearheaded by the Department for Promotion of Industry and Internal Trade (DPIIT), signifies a fundamental paradigm shift in India’s intellectual property rights regime concerning generative artificial intelligence. By instituting a statutory licensing mechanism and establishing a centralized body—the Copyright Royalty and Compensation Administration Tribunal (CRCAT)—to manage the collection and distribution of royalties, India is transitioning definitively from the legal ambiguities inherent in ‘fair dealing’ towards a highly structured and comprehensive regulatory model. This pivotal policy alteration carries profound and divergent consequences for key stakeholders, both within India and internationally.

Implications for the Creative and Copyright Holder Ecosystem

For creators, artists, authors, and copyright holders, this framework represents a potential financial boon and a long-overdue formal recognition of their foundational contribution to the advancement of AI. At present, the vast majority of data utilised to train large language models (LLMs) and other generative AI systems is acquired through web scraping, operating within a legally ambiguous space. The new framework is poised to deliver:

• Assured Revenue Generation: By mandating compensation from AI companies for the use of copyrighted material, the framework establishes a statutory foundation for a royalty stream that was previously non-existent. This has the potential to significantly stimulate the creative economy.
• Centralised Royalty Administration: The CRCAT is envisioned as the principal administrative authority, thereby simplifying a process that would otherwise impose insurmountable legal and logistical burdens on individual creators seeking to negotiate licenses with global technology corporations.
• The Criticality of Equitable Distribution: Notwithstanding the framework’s promise, the most pressing concerns pertain to the operational modalities of the CRCAT. Specifically, how will royalties be accurately assessed for works integrated into expansive and diverse training datasets? What precise metrics will guarantee a transparent and equitable allocation of funds to millions of creators? The ultimate success of this endeavour is entirely contingent upon the CRCAT’s capacity to administer this complex system fairly.

Implications for AI Developers and the Technology Sector

For both domestic and international AI developers, the framework presents a strategic trade-off between securing legal certainty and accepting a potential increase in financial liability.

• Legal Clarity Versus Operational Expenditure: The framework furnishes a clear and legally sanctioned avenue for the training of AI models, thereby mitigating the systemic risk posed by perpetual, fragmented litigation and complex individual licensing negotiations. This constitutes a substantial operational advantage. Nevertheless, this certainty is associated with an intrinsic cost.
• The Challenge of Global Revenue Jurisdiction: The most controversial element involves the potential requirement for AI developers to remit a percentage of their “global revenue” to an Indian-based CRCAT. This provision raises intricate questions regarding jurisdictional reach and international enforceability. For multinational entities whose models are trained on a heterogeneous mix of global data, the precise method for accurately assessing the “Indian” component of the input data to determine the proportional royalty remains a critical issue. If the royalty obligation is deemed excessive, this could result in significant compliance and financial strain, potentially deterring certain global entities from developing or deploying their models within the Indian market. The determination of the final royalty percentage and the formal definition of ‘global revenue’ will constitute pivotal points of negotiation.

Implications for India’s Role in Global AI Governance

This ambitious policy initiative extends beyond the regulation of domestic copyright; it strategically positions India as a bold, proactive architect of regulatory norms within the nascent domain of AI governance.

• A Distinct Regulatory Paradigm: While Western jurisdictions, notably the European Union with its AI Act, have primarily concentrated on issues of safety, ethics, and consumer protection, India’s framework directly addresses the foundational economic challenge of copyright and the extraction of content value. It offers a unique, creator-centric counterpoint to the prevailing, often industry-driven, models of AI development.
• Establishment of a Global Precedent: Should this framework prove successful, it possesses the potential to establish a global precedent, particularly for developing nations possessing extensive, culturally significant, and frequently scraped content libraries. It represents a powerful affirmation of a nation’s sovereign right to derive economic value from its intellectual capital when utilised by global technological corporations.

The extension of the public consultation deadline to February 6, 2026, underscores the government’s acknowledgement of the framework’s profound ramifications and its dedication to fostering a robust, inclusive dialogue with all stakeholders. As legal experts scrutinise jurisdictional limits, industry associations advocate for preferential royalty terms, and creators lobby for maximum transparency, the eventual configuration of India’s AI Mandate will critically define the future economic dynamic between human creativity and machine intelligence within the nation. The regulatory shift is unequivocal: the passive legal acceptance of “fair dealing” for AI training is being superseded by a sophisticated, statutory framework engineered both to cultivate the AI economy and to ensure equitable compensation for its creative foundation.

Reference(S):

Primary Authorities: Indian Statutes & Regulations

• The Copyright Act, No. 14 of 1957, India Code (1957).
• The Copyright Rules, 2013, Gazette of India, pt. II sec. 3(i) (Mar. 14, 2013).
• Draft Copyright (Amendment) Rules, 2026 (incorporating the Copyright Royalty Collection and Allocation Tribunal provisions).

Government Publications & Policy Papers

• Dep’t for Promotion of Indus. & Internal Trade, Ministry of Com. & Indus., Gov’t of India, Part I Working Paper on Generative AI and Copyright (Dec. 2025).
• Dep’t for Promotion of Indus. & Internal Trade, Ministry of Com. & Indus., Gov’t of India, National Intellectual Property Rights Policy (2016).
• Ministry of Elecs. & Info. Tech., Gov’t of India, IndiaAI Strategy Report: Toward a $1 Trillion Digital Economy (2024).

International Instruments & Foreign Legislation

• Directive 2019/790, of the European Parliament and of the Council of 17 April 2019 on Copyright and Related Rights in the Digital Single Market, 2019 O.J. (L 130) 92.
• Regulation 2024/1689, of the European Parliament and of the Council of 13 June 2024 Laying Down Harmonised Rules on Artificial Intelligence (AI Act), 2024 O.J. (L 1689).
• U.S. Copyright Office, Artificial Intelligence and Copyright, 88 Fed. Reg. 16190 (Mar. 16, 2023).

Secondary Sources: Books & Journals

• Bina Agarwal, Intellectual Property Rights in the Digital Age: An Indian Perspective (3d ed. 2024).
• Shardul Amarchand Mangaldas, The AI Mandate: Navigating the Mandatory Blanket Licensing Framework in India, SAM Legal Alert (Jan. 2026).
• P. Narayanan, Law of Copyright and Industrial Designs (4th ed. 2017).

[1] Dep’t for Promotion of Indus. & Internal Trade [DPIIT], Part I Working Paper on Generative AI and Copyright (Dec. 2025) [hereinafter DPIIT Working Paper].

[2] Id. at ¶ 4.2 (discussing the expansion of Copyright Societies under § 33 of the Copyright Act).

[3] DPIIT Working Paper, supra note 1, at 12.

[4] The Copyright Act, No. 14 of 1957, § 52(1)(a), India Code.

[5] DPIIT Working Paper, supra note 1, at ¶ 5.1 (concluding that industrial-scale scraping exceeds the scope of “private or personal use”). [^6]: Id. at ¶ 6.3.

[6] Id. at ¶ 6.3.

[7] Compare Directive 2019/790, of the European Parliament and of the Council of 17 April 2019 on Copyright and Related Rights in the Digital Single Market, art. 4, 2019 O.J. (L 130) 92, with DPIIT Working Paper, supra note 1, at ¶ 7.2.

[8] DPIIT Working Paper, supra note 1, at ¶ 8.

[9] Id. at ¶ 8.4 (outlining the global revenue royalty model).

[10] d. at ¶ 9.1 (addressing attribution and algorithmic distribution).

[11] Id. at ¶ 10.2 (defining “lawful access” and exclusion of infringing copies).