Who Owns AI-Generated Content - A Critical Analysis Of Copyright Authorship Under The Indian Copyright Act, 1957
Author: LawSphereX
Category: Intellectual Property
Published:
Updated:
WHO OWNS AI-GENERATED CONTENT? A CRITICAL
ANALYSIS OF COPYRIGHT AUTHORSHIP UNDER THE
INDIAN COPYRIGHT ACT, 1957
Shivam Kumar Mishra, PGDIPR, National Law School of India University
ABSTRACT
The rapid proliferation of Generative Artificial Intelligence (GenAI) has precipitated a profound structural crisis within global intellectual property regimes. AI models, capable of producing sophisticated literary, artistic, and musical works with minimal human prompting, have triggered unprecedented legal and philosophical inquiries regarding authorship, ownership, and the fundamental nature of creativity. The Indian Copyright Act, 1957, predicated exclusively upon an anthropocentric framework, is currently straining under the weight of autonomous machine creation. This article critically examines the existing legislative architecture in India, specifically deconstructing Section 2(d)(vi) of the Act, which attributes authorship of computer-generated works to the person who causes the work to be created.
By juxtaposing this statutory provision against the Indian judiciary’s established "skill and judgment" standard for originality, this paper identifies a critical legal vacuum wherein AI-generated outputs seemingly default to the public domain. The article navigates the regulatory ambiguities highlighted by the ephemeral registration of the "RAGHAV" AI as a co-author, evaluates the landmark ANI Media Pvt. Ltd. v. OpenAI Inc. litigation, and considers the 2026 IT Rule amendments. Ultimately, this article argues that attempting to retrofit autonomous AI outputs into archaic legal structures engenders severe commercial uncertainty and advocates for a sui generis legislative intervention.
INTRODUCTION:
For centuries, the philosophical underpinning of copyright law has been tethered to the human condition. Whether viewed through the lens of John Locke’s labour theory1 or utilitarian concepts designed to incentivise culture, the foundational premise has remained constant: creativity is a uniquely human endeavour. Technology, historically, functioned as a passive conduit—a digital pen or a camera—operating under the deterministic control of a human user. The advent of Generative Artificial Intelligence (GenAI)—spearheaded by language models and diffusion networks like ChatGPT and DALL-E—has ruptured this continuity. GenAI systems are not passive tools; they are autonomous engines that ingest vast repositories of scraped data to produce poetry, scripts, and art with minimal human intervention. This technological reality has violently collided with the Indian Copyright Act, 1957.
Drafted in an era long preceding algorithmic autonomy, the Act’s architecture rests on the assumption that only a natural person can be an "author. " As India navigates this transition, the legal and commercial conundrum is clear: who owns the copyright when an AI system independently generates a work? Does ownership vest in the developer who trained the model, the user who drafted the prompt, or does the output belong to the public domain? As of mid-2026, the absence of clear legislative directives threatens to stifle both investment and artistic enterprise. This article dissects the inadequacies of the current Act, examines the "skill and judgment" test, and evaluates recent domestic developments to propose a path forward. THE ANTHROPOCENTRIC ARCHITECTURE AND 2(d)(vi): Under the Indian framework, authorship and ownership are distinct.
Section 17 establishes that the author is the first owner2. To identify the author, one must look to Section 2(d), which defines authorship based on the work’s nature3. For traditional works, the author is simply "the person who creates the work. " The primary provision concerning technological creation is Section 2(d)(vi), inserted in 1994, which states that for "computer-generated" works, the author is "the person who causes the work to be created. " At a cursory glance, this seems to resolve the dilemma by pointing toward the human user. However, a deeper functional analysis reveals deep flaws. This 1994 amendment4 was intended for deterministic tools like spreadsheets— 1 John Locke, Two Treatises of Government § 27 (Peter Laslett ed. , Cambridge Univ. Press 1988) (1690).
2 The Copyright Act, 1957, § 17, No. 14, Acts of Parliament, 1957 (India).
3 Id. § 2(d).
4 The Copyright (Amendment) Act, 1994, § 2, No. 38, Acts of Parliament, 1994 (India).
where the machine merely acts as a surrogate. Generative AI breaks this chain of causation. When a user inputs a broad prompt into an image generator, the resulting artwork is dictated by the AI’s neural network interpreting statistical probabilities, not by the user’s specific creative vision. The user provides the idea , but Indian jurisprudence, such as in R.G. Anand v. Delux Films5 , confirms that ideas cannot be copyrighted; only the expression is protectable. Since the AI independently formulates the expression, categorising the prompter as the author stretches the statutory language beyond its breaking point.
THE ORIGINALITY STANDARD:
Copyright protection is contingent upon a work being "original. " The Supreme Court’s landmark ruling in Eastern Book Company v. D. B. Modak6 established the "skill and judgment" test. For a work to be original in India, it must originate from the author and reflect a non-trivial degree of independent intellectual effort, skill, and judgment. This test is problematic for GenAI. An AI operates by recognising patterns within its training data and predicting outputs. It does not possess consciousness or the capacity for "judgment" in any legally cognisable sense. Conversely, the human user who types a text prompt applies negligible skill to the final expression. Because the machine cannot legally exercise judgment and the human contributes insufficient intellectual effort to the execution, AI-generated outputs arguably fail the Modak test.
Without legislative reform, most purely AI-generated content in India likely occupies a "legal limbo," effectively defaulting to the public domain.
REGULATORY TURBULENCE: THE RAGHAV CONUNDRUM
The tension between technology and law became reality with the "RAGHAV" incident. In 2020, Indian lawyer Ankit Sahni registered an artwork titled Suryast with the Copyright Office, naming himself and the AI app RAGHAV as co-authors. This registration was a watershed moment, potentially making India the first jurisdiction to formally recognise AI as an author. However, the registration was short-lived. Following global debate, the Copyright Office issued a withdrawal notice, signalling that copyright vests exclusively in human artists. While the validity of this withdrawal remains a subject of scrutiny, the incident underscored the systemic confusion within our regulatory bodies. By oscillating between granting and
5 R.G. Anand v. Delux Films, (1978) 4 S.C.C. 118 (India).
6 Eastern Book Company & Others v. D.B. Modak & Another, (2008) 1 S.C.C. 1 (India).
withdrawing recognition, the registry highlighted the futility of forcing AI into human-centric frameworks. As of April 2026, the Delhi High Court has even had to intervene to force the Copyright Office to resolve pending applications involving AI authorship, such as those filed by Stephen Thaler. INPUT VS. OUTPUT: THE ANI v. OPENAI LITIGATION The ANI v. OpenAI Litigation While authorship debates dominate academic circles, the battle over inputs —the training data—is the real-world crisis. The ongoing ANI Media Pvt. Ltd. v. OpenAI Inc7. litigation before the Delhi High Court is arguably the most influential AI legal case in India today. ANI alleges that OpenAI unlawfully scraped its proprietary news archives to train its models. OpenAI argues that its use is transformative and falls under "fair dealing. " Unlike the U. S.
"fair use" doctrine, India’s Copyright Act relies on a specific, exhaustive list of exceptions under Section 528. The systematic scraping of protected works for commercial AI training does not comfortably fit within these statutory gaps. With judgment currently reserved as of mid-2026, the verdict will determine if AI developers have a "right to train" on Indian data. If the court rules strictly against OpenAI, India may force the industry to rethink its data practices; if it rules in their favour, it may leave human creators without legal recourse.
THE 2026 IT RULE AMENDMENTS AND THE ROAD AHEAD:
The government has recently begun responding. On February 10, 2026, the Ministry of Electronics and Information Technology (MeitY) notified the IT (Intermediary Guidelines)9 Amendment Rules, 2026. These rules incorporate "Synthetically Generated Information" (SGI) into the due diligence obligations of platforms. While these rules focus on deepfakes and misinformation, they represent a significant shift toward mandatory labelling and disclosure. This regulatory impatience with AI harms suggests that legislative reform for copyright is likely the next domino to fall. The path forward requires a three-pronged legislative approach: Defining Human Intervention: Parliament must amend the Act to define the degree of "substantial human intervention" required for copyright. Iterative prompting and post-
7 ANI Media Pvt. Ltd. v. OpenAI Inc. & Anr., CS(COMM) 1028/2024, High Court of Delhi (India).
8 The Copyright Act, 1957, § 52, No. 14, Acts of Parliament, 1957 (India).
9 Ministry of Electronics and Information Technology (MeitY), Notification G.S.R. 84(E), Information
Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026 (notified Feb. 10, 2026).
generation editing should be recognised as human effort, while autonomous generation should not. Sui Generis Protections: For works with minimal human input, we should consider a sui generis right with a shortened duration (e.g., 5-10 years) for the person who made the "necessary arrangements," explicitly denying moral rights which are fundamentally human. Statutory Licensing for Data: To resolve the ANI v. OpenAI conflict, a statutory licensing scheme could allow AI developers to use data while ensuring a portion of corporate revenues is redirected to creators via a collective management organisation.
CONCLUSION:
The Copyright Act, 1957, is a relic in the age of generative algorithms. Section 2(d)(vi) is inadequate for navigating autonomous creation, and the Modak standard creates an insurmountable barrier for AI. The Suryast registration and the ANI litigation demonstrate that ad-hoc interpretations are unsustainable. India stands at a crossroads. To lead in the digital economy while preserving the sanctity of human artistry, we must urgently overhaul our intellectual property regime. Parliament must move beyond human-centric definitions to create a balanced framework that incentivises innovation without disenfranchising the creators whose work serves as the foundation for the AI revolution. Without clear legislative clarity, the intersection of AI and law will remain a labyrinth of uncertainty, stifling the very progress we seek to foster.