Navigating the Algorithmic Muse: India’s Quest for Balance in AI and Copyright Law 

Navigating the Algorithmic Muse: India’s Quest for Balance in AI and Copyright Law 

The release of a working paper by the Department for Promotion of Industry and Internal Trade (DPIIT) on Generative Artificial Intelligence (GenAI) and copyright marks a critical juncture for India’s digital future. This isn’t merely a technical or legal formality; it’s a profound societal negotiation. We are being forced to re-examine the very bedrock of creativity—what it means to author, to own, and to innovate. As GenAI tools blur the line between human artist and sophisticated instrument, India’s proposed regulatory framework seeks to forge a path that neither stifles a transformative technology nor disregards the rights of creators. The central challenge is as delicate as it is urgent: how do we feed the engines of progress without consuming the cultural fuel that powers them? 

Deconstructing the Dilemma: Inputs, Outputs, and Regulatory Gaps 

The copyright controversy surrounding GenAI unfolds on two distinct but interconnected fronts: the input side (training) and the output side (creation). Each presents unique philosophical and legal puzzles. 

1. The Input Side: The “Great Library” DilemmaAt the heart of every powerful GenAI model lies an immense dataset—a digital library often compiled by scraping the internet. This library contains countless copyrighted books, articles, images, and code, typically ingested without explicit permission or licensing. From the AI developer’s perspective, this is “reading” or “analysis” on an unprecedented scale, a necessary process to identify patterns and learn the rules of language, art, or music. From the copyright holder’s perspective, this constitutes massive, unlicensed reproduction of their work for a commercial purpose.

The Indian Copyright Act, 1957, currently offers no safe harbor for this activity. The ambiguity around whether such use falls under “fair dealing” for research (Section 52(1)(a)) creates a legal grey area. This uncertainty is a significant brake on innovation, as developers operate under a cloud of potential litigation. The working paper rightly identifies this absence of a clear “text and data mining” (TDM) exception as a major regulatory gap that needs addressing. 

2. The Output Side: The “Ghost in the Machine” QuestionWhen an AI generates a poignant poem, a striking image, or a functional piece of software, who, if anyone, owns it? Traditional copyright law is anthropocentric, vesting rights in a human “author.” Can an algorithm be an author? Most global jurisdictions, including India’s current stance, lean toward “no,” deeming fully AI-generated works as lacking the necessary human authorship for protection.

This leads to complex scenarios: 

• The Prompt Engineer’s Claim: Is the person crafting the detailed prompt the “author,” or merely a user instructing a tool? 

• Derivative or Infringing Output: When an AI produces content stylistically identical to a living artist or replicates substantial elements of a protected work, does it constitute infringement? If so, who is liable—the user, the developer, or the AI itself? 

• Moral Rights: Copyright includes non-economic “moral rights,” like the right to attribution and to object to distortion. How can these be applied to an AI’s output, which has no personal integrity to defend? 

These aren’t abstract questions. They determine investment, commercial strategy, and the legal risks for a burgeoning creative-tech industry. 

Global Playbook: Contrasting Philosophies in Regulation 

As the DPIIT paper notes, different countries are experimenting with vastly different models, each reflecting a distinct balance between competing interests. 

• The Permission-First Model (Voluntary Licensing): Favored in parts of the US and by rightsholder groups, this model treats AI training as a clear-cut use requiring a license. It gives maximum control to creators but risks creating an insurmountable “transaction cost” nightmare. Negotiating licenses with millions of individual rightsholders is impractical, potentially cementing the market power of a few large tech firms with the resources to do so. 

• The Exception-First Model (TDM Exceptions): Adopted by the EU, Japan, and Singapore, this approach creates a statutory copyright exception specifically for text and data mining for research and, in some cases, commercial purposes. Singapore’s broad, flexible exception is often cited as a pro-innovation benchmark. This model provides legal certainty for AI developers but is often criticized by creators as a “free ride” that devalues their work. 

• The Collective Solution (Extended Collective Licensing – ECL): Practiced in the EU and Scandinavia for specific uses, ECL allows a collective management organization (CMO) to grant licenses for a repertoire that includes both its members and non-members within a category. It simplifies bulk licensing but requires robust, transparent CMOs and mechanisms to ensure royalties reach all rightsholders. 

India’s Proposed Path: The Hybrid Regulatory Model 

The DPIIT working paper’s recommendation of a Hybrid Regulatory Model is arguably its most significant and insightful contribution. It attempts to synthesize the strengths of the global approaches while mitigating their weaknesses. The model has two core, interdependent pillars: 

1. A Right to Train with Fair CompensationThis pillar is revolutionary. It conceptualizes a statutory right for AI developers to use any lawfully accessed copyrighted work for training. This eliminates the legal uncertainty of the current “fair dealing” ambiguity and prevents rightsholders from issuing blanket refusals that could stall technological development.

However, this right is not a free pass. It is intrinsically linked to the second pillar: a guaranteed mechanism for fair compensation. This moves the debate from “if you get paid” to “how you get paid,” framing the use not as an infringement to be prevented but as a new form of cultural value extraction that requires remuneration. 

2. Statutory Remuneration RightsThis is the counterbalance. Copyright owners would be granted a non-waivable statutory right to receive royalties for the use of their works in AI training. This likely would not be an opt-in system but a mandatory collective right, managed through CMOs or a new licensing body. Key implementation questions include:

• Valuation: How is the value of a single work within a training set of billions determined? Should it be a flat fee, or based on the prominence or frequency of use in training? 

• Distribution: How are royalties collected from developers and accurately distributed to millions of rightsholders, many of whom may be unaware their work was included? 

• Transparency: Will developers be required to maintain and disclose detailed, attributable training datasets—a concept known as “dataset provenance”? This is technically challenging but crucial for accountability. 

Beyond Legislation: The Road Ahead and Unanswered Questions 

Adopting a hybrid model would be a bold first step, but its success would depend on nuanced execution and addressing broader challenges. 

• Fostering Human-AI Collaboration: The law must evolve to recognize and protect works that are the product of meaningful human-AI collaboration. A gradient of authorship, rather than a binary human/computer distinction, may be necessary. 

• Liability and Transparency: Clear guidelines are needed on infringement liability for AI outputs. Promoting standards for “AI watermarking” and metadata to identify AI-generated content will be crucial for maintaining trust. 

• The Global Ripple Effect: India’s decision will resonate globally. As a major IT powerhouse and a colossal producer of cultural content, its framework could become a template for other developing nations seeking to nurture both their tech and creative sectors. 

Conclusion: Crafting a New Social Contract for Creativity 

The DPIIT working paper is more than a policy document; it’s an invitation to craft a new social contract for the algorithmic age. The proposed hybrid model acknowledges a fundamental truth: in the 21st century, data and creative expression are not just artifacts of culture but critical infrastructure for innovation. Treating them purely as exclusive property to be locked away could hinder progress. Conversely, treating them as free raw material disrespects the labor and genius of creators. 

India’s potential path—a right to train paired with a right to remuneration—seeks to establish a continuous, systemic flow of value. It aims to ensure that as AI systems learn from and build upon the vast archive of human creativity, they also feed back into its ecosystem, supporting the very creators who make the next cycle of innovation possible. In doing so, India has the opportunity to demonstrate that technological advancement and respect for creative rights are not a zero-sum game, but can be the twin engines of a thriving digital society.