I. The Prompt That Broke the Provision

The facts read like a parable. A student pursuing an LLM in Intellectual Property and Technology Laws sits for an online exam. The university’s Unfair Means Committee runs his answers through an AI-detection tool, concludes that 88% of the script is machine-written, and marks him failed. The student, himself a practising IP lawyer, files a writ in the Punjab and Haryana High Court.[1] His defence is not the predictable one that he did not use AI. It is the cleverer one: that even if he did, it does not matter, because under section 2(d)(vi) of the Copyright Act, 1957, the author of a computer-generated work is “the person who causes the work to be created”,[2] and an author, he reasoned, cannot plagiarise himself.[3]

The argument was never adjudicated. The University quietly issued him a revised transcript, and on 18 November 2024 the High Court disposed of the petition as infructuous. But the claim he made that prompting a generative model is indistinguishable from typing has lodged itself in the Indian copyright imagination like a piece of gravel in a shoe. It deserves to be taken seriously. It also deserves to be taken apart.

The trouble with the Shakkarwar argument is not that it misreads section 2(d)(vi). It reads it beautifully. The trouble is that section 2(d)(vi) is only half the question copyright law asks.

II. A Statute Written for a Different Machine

Section 2(d)(vi), inserted by the Copyright (Amendment) Act, 1994,[4] defines the author of a computer-generated literary, dramatic, musical or artistic work as the person who causes the work to be created. The drafting was, for its time, visionary. It was also unmistakably an attempt to accommodate word processors, CAD software, and spreadsheets the kind of machines where the causation between the human operator and the output is direct, deterministic, and total. If you typed, the machine printed. If you drew, it rendered. The computer was a docile amanuensis.

The generative-AI computer is a very different animal. Its outputs are probabilistic rather than deterministic; its training data is vast, unlicensed, and opaque; its contribution to the final work is neither mechanical nor marginal. To say that a user who types “write me a 1,500-word essay on the doctrine of privity in the style of Lord Denning” has caused the work to be created is technically correct in roughly the same sense that ordering a pizza causes a pizza to be created. The provision, drafted when “computer-generated” meant a mail-merge letter, is being asked to carry a model it was never engineered to carry.

There is, in addition, a silent assumption inside section 2(d)(vi) that the drafters never interrogated: that whoever causes a computer-generated work to exist has, by that causation alone, contributed enough original effort to deserve copyright. In 1994 this assumption was cheap. In 2026 it is the whole fight.

III. The Missing Half of the Sentence: Originality

Copyright does not protect authors. It protects original works of authors. Section 13(1)(a) of the Copyright Act places originality at the door of protection,[5] and the Supreme Court in Eastern Book Company v DB Modak clarified exactly what that door looks like.[6] Rejecting both the low bar of the “sweat of the brow” doctrine and the higher “modicum of creativity” standard, the Court adopted the Canadian “skill and judgment” test from CCH Canadian:[7] the author’s work must involve skill that is not trivial, and judgment that goes beyond the mere application of labour.[8]

Apply that test to the Shakkarwar fact pattern and the elegance of his argument begins to unravel. Section 2(d)(vi) tells us who owns copyright in a computer-generated work if copyright subsists. It does not tell us whether it subsists in the first place. Ownership rides on originality; it does not create it. And originality, after EBC v Modak, is no longer an assumption we can smuggle into authorship it is a threshold the author must cross.

A one-line prompt of the “write me an answer on constitutional morality” variety is, by almost any honest reading, not skill and judgment. It is an instruction. The skill and judgment, if any, live inside the model trained on a corpus of human writing, refined through techniques the user does not touch, and deployed through weights the user cannot see. The operator, in such a case, is closer to a commissioning patron than to an author. Copyright has never protected patrons.

But and this is where the honest analysis must part ways with the lazy one not all prompts are one-liners. A user who constructs a multi-stage prompt with specific doctrinal framing, iterates through a dozen drafts rejecting each on substantive grounds, edits the final output, reorganises its structure, and fact-checks its citations is engaged in something the Supreme Court would recognise as skill and judgment. There is a real intellectual scaffolding behind that output, even if no single sentence of the final text was typed by the human. Between the one-line prompter and the twelve-draft curator lies the entire unresolved question of contemporary copyright. Section 2(d)(vi) cannot distinguish between them. EBC v Modak demands that we do.

IV. The Comparative Lens: Three Jurisdictions, Three Confusions

India is not alone in this muddle. In March 2025, the United States Court of Appeals for the District of Columbia Circuit, in Thaler v Perlmutter, affirmed that a work authored entirely by an AI system Dr Stephen Thaler’s Creativity Machine cannot be registered for copyright because the Copyright Act of 1976 presupposes a human author.[9] The Supreme Court declined certiorari on 2 March 2026, leaving the ruling intact.[10] The American position is philosophically firm but practically evasive: it refuses copyright to pure machine works but leaves unanswered how much human involvement elevates an AI-assisted output into authorship.

The United Kingdom’s section 9(3) of the Copyright, Designs and Patents Act 1988 the drafting ancestor from which India’s section 2(d)(vi) descends vests authorship of a computer-generated work in the person by whom the arrangements necessary for its creation are undertaken.[11] It is marginally better worded than the Indian version because “arrangements” at least gestures at effort, but it was drafted in the same pre-generative era, and British courts have hardly used it. The provision is a conceptual orphan.

India’s own flirtation with this question is most visible in the Suryast saga. In November 2020 the Indian Copyright Office granted registration for an AI-generated artwork listing the RAGHAV painting tool as co-author alongside its human user; a year later, the Office issued a withdrawal notice.[12] The United States Copyright Review Board refused the same work in December 2023;[13] the Canadian Intellectual Property Office, through its automated system, registered it. Three offices, three outcomes, one painting. If copyright’s plumbing is leaking this visibly at the international level, the case for new categories is not speculative it is overdue.

V. A Proposal: The AI-Assisted Authorship Category

The doctrinally cleanest way out of this tangle is to stop pretending the binary human/computer-generated classification is adequate. It is not. What Indian copyright law needs is a third category AI-assisted authorship sitting between the purely human author under section 2(d)(i)–(v) and the “person who causes the work to be created” under section 2(d)(vi). The category would recognise, on the face of the statute, that generative-AI outputs are neither fully human-created nor mechanically produced, and that authorship in such cases must be earned, not presumed.

Three practical criteria could define it. First, creative control: the human contributor must demonstrate concrete exercise of skill and judgment prompt design, iteration, curation, structural editing rather than a one-line request. Second, attributable contribution: the human-created elements of the work must be separable in principle, even if not in appearance, from the machine-generated mass. Third, disclosure: the author must declare, at the point of registration or publication, the nature and extent of AI assistance. This is not a burden on authors, it is an insurance policy for them, because undisclosed AI use is precisely what exposes a copyright claim to later invalidation.

Such a category would absorb the Shakkarwar problem without legitimising its shortcut. An exam script generated by a one-line prompt would fail the first criterion and forfeit any claim to originality. The academic-integrity question would then live where it ought to, which is with the university, not the copyright office. A genuinely iterative AI-assisted work the kind a thoughtful researcher, screenwriter, or designer actually produces would clear the threshold and receive protection calibrated to the human contribution.

The move would also align India with the preliminary direction of the Expert Panel constituted by the Ministry of Commerce and Industry in May 2025 to review the Copyright Act for generative-AI realities.[14] That panel has reportedly been weighing a distinct statutory chapter on AI-generated works, including a working distinction between AI-assisted and AI-generated output. The proposal advanced here is, in that sense, not radical but overdue.

VI. Closing: The Operator’s Burden

Authorship has always been, at its best, a verb disguised as a noun. The Romantic-era fiction of the solitary genius producing unaided works has never reflected how books, paintings, or films are actually made editors, assistants, tools, and collaborators have been silent co-authors for centuries. Generative AI is only the latest, and loudest, of those silent collaborators. What it exposes is not that copyright’s human-author foundation is wrong, but that the foundation has been resting on unarticulated assumptions that no longer hold.

Shakkarwar’s argument, properly understood, is not a triumph but a provocation. He showed, perhaps unintentionally, that section 2(d)(vi) read in isolation produces an absurd conclusion that anyone who types a prompt becomes an author. The correction is not to abandon the provision but to complete it, by writing into Indian copyright law what EBC v Modak already says everywhere else: that authorship without originality is not authorship at all.

The operator is not yet the author. She can become one. But only if she does the work the machine has made it tempting to skip.

[1] Kaustubh Anil Shakkarwar v Jindal Global Law School CWP-29690-2024 (O&M), Neutral Citation 2024:PHHC:150307 (Punjab and Haryana High Court, 18 November 2024) <https://indiankanoon.org/doc/149565231/> accessed 19 April 2026.

[2] Copyright Act 1957, s 2(d)(vi) <https://indiankanoon.org/doc/1655540/> accessed 19 April 2026.

[3] See the reporting in Aroon Deep, ‘OP Jindal University Student Files Suit Against AI-Based Plagiarism System’ (Medianama, 5 November 2024) accessed 19 April 2026.

[4] Copyright (Amendment) Act 1994, inserting s 2(d)(vi) into the principal Act.

[5] Copyright Act 1957, s 13(1)(a).

[6] Eastern Book Co v DB Modak (2008) 1 SCC 1 (Supreme Court of India) <https://indiankanoon.org/doc/1062099/> accessed 19 April 2026.

[7] CCH Canadian Ltd v Law Society of Upper Canada [2004] 1 SCR 339 (Supreme Court of Canada), adopted by the Supreme Court of India in Eastern Book Co (n 6) [38]–[40].

[8] ibid [38].

[9] Thaler v Perlmutter No 23-5233, 2025 WL 839178 (DC Cir, 18 March 2025) <https://media.cadc.uscourts.gov/opinions/docs/2025/03/23-5233.pdf> accessed 19 April 2026.

[10] Thaler v Perlmutter, certiorari denied, No 25-449 (US Supreme Court, 2 March 2026).

[11] Copyright, Designs and Patents Act 1988 (UK), s 9(3) <https://www.legislation.gov.uk/ukpga/1988/48/section/9> accessed 19 April 2026.

[12] See the reporting on India’s November 2020 registration and subsequent withdrawal notice in Rani Mehta, ‘Exclusive: India recognises AI as co-author of copyrighted artwork’ (Managing Intellectual Property, 5 August 2021) accessed 19 April 2026.

[13] United States Copyright Office, Review Board, Second Request for Reconsideration for Refusal to Register SURYAST (Correspondence ID 1-3ZPC6C3; SR 1-11016599571, 11 December 2023) <https://www.copyright.gov/rulings-filings/review-board/docs/SURYAST.pdf> accessed 19 April 2026.

[14] Ministry of Commerce and Industry, Government of India, expert panel constituted in May 2025 to review the adequacy of the Copyright Act 1957 for generative-AI works. See <https://pib.gov.in> for official releases; and secondary reporting on preliminary proposals in ‘Generative AI & Copyright Law in India’ (Lexology, 16 July 2025) accessed 19 April 2026.

Eashan Goyal is a third-year law student at O.P. Jindal Global University. His areas of interest include Intellectual Property Rights, Legal Tech, and Constitutional Law.

Purvi Bansal is a second-year law student at O.P. Jindal Global University. Her areas of interest include Intellectual Property Rights, Taxation and Fashion Law.