THE AUTHORSHIP PARADOX
Balancing Machine Execution Against the Mind That Causes It
By M. Murali Mohan
Introduction: The Question Behind the Machine
Generative artificial intelligence has unsettled one of the foundational assumptions of copyright law: that a work of authorship can be attributed to a legally cognisable human creator whose intellectual contribution finds expression in a protected work. A machine can now produce literary passages, images, musical compositions and other expressive material in response to instructions supplied by a human being. The resulting work may reflect an elaborate human conception, yet its immediate verbal or visual form may be determined, in substantial measure, by computational processes.
The resulting controversy is frequently presented as a choice between two propositions: either the artificial intelligence system is the author, or the work is incapable of copyright protection. That formulation is incomplete. It overlooks a third possibility: the system may be an instrument through which a human being causes a work to be created, while the extent of copyright protection depends upon the applicable statute, the nature of the human contribution and the originality of the resulting expression.
The distinction is particularly important under Indian law. Section 2(d)(vi) of the Copyright Act, 1957 expressly provides that, in relation to a literary, dramatic, musical or artistic work which is computer-generated, the author means “the person who causes the work to be created”. The provision supplies a statutory point of departure that cannot properly be disregarded merely because the technology employed is capable of producing expressive material with considerable autonomy.¹
The difficult question is not whether a machine possesses consciousness, personality or moral responsibility. It is whether, and in what circumstances, the human agency that initiates, directs, controls, selects or materially transforms computer-generated expression falls within the statutory conception of authorship.
This article advances the argument that copyright analysis must distinguish the machine's execution of a process from the human agency that may bring the protected work into existence. It examines the statutory framework under Indian law, the originality doctrine, the allocation of first ownership, and the comparative significance of the United States decisions concerning photography, computer-generated works and artificial intelligence.
The argument is not that every prompt necessarily creates a copyright, nor that payment for an AI subscription automatically confers ownership of every output. It is that the use of an artificial intelligence system should not, by itself, extinguish a claim to human authorship where the statutory requirements are otherwise satisfied.
Part I — The Judicial and Administrative Position: The Case for Caution
1. The distinction between conception and expression
Copyright protects expression, not an abstract idea, intention or general instruction. A person may conceive a novel, formulate its central theme, determine its characters and establish its moral architecture without thereby acquiring copyright over every sentence subsequently written by another person. The same distinction becomes more difficult to apply when the person receiving the instructions is not another human being but a generative system.
A prompt may range from a general request—such as asking for a poem about the sea—to an elaborate set of instructions specifying structure, imagery, character relationships, tonal progression, recurring motifs and successive revisions. These situations should not be treated as legally identical.
In the first situation, the user may supply little more than an idea or subject. In the second, the user's contribution may extend to the detailed design of the work, the repeated correction of its direction and the deliberate selection or transformation of particular expressions. Even then, the degree of control over the final expression must be established rather than presumed.
The relevant inquiry is therefore one of attribution and degree. What did the human actually contribute? Which expressive elements were determined by the human, and which were generated independently by the system? Did the human exercise meaningful control over the final form, or merely request a result and accept what the system supplied?
These questions are central to a sound copyright analysis. They prevent two opposite errors: treating every user as the author of everything generated following a prompt, and treating every work produced with AI assistance as incapable of containing protectable human authorship.
2. The United States position after Thaler v. Perlmutter
In Thaler v. Perlmutter, the United States courts considered an application for copyright registration in relation to an image claimed to have been generated autonomously by an artificial intelligence system. The United States Court of Appeals for the District of Columbia Circuit upheld the requirement of human authorship under the United States Copyright Act. The United States Supreme Court subsequently denied certiorari on 2 March 2026 in Thaler v. Perlmutter, No. 25-449.²
The procedural consequence must be stated accurately. A denial of certiorari leaves the judgment under review undisturbed; it is not itself a Supreme Court judgment on the merits affirming every proposition advanced by the lower court. Nevertheless, the lower-court decision remains an important comparative authority concerning the requirement of human authorship under United States law.
It is equally important not to transpose that conclusion mechanically into Indian law. The Indian Copyright Act contains an express provision addressing computer-generated works and identifying the author by reference to the person who causes the work to be created. Whether a particular AI-assisted work falls within that provision requires an examination of Indian statutory language, the nature of the contribution and the relevant principles of originality.
The comparative question is consequently not whether India must adopt the American position. It is whether the Indian statutory text permits, or requires, a different analysis in circumstances where a human being can be identified as the person who caused the work to be created.
3. The significance of Zarya of the Dawn
The United States Copyright Office's determination concerning Zarya of the Dawn provides a further illustration of the distinction between human-authored and AI-generated material. The Office recognised copyright in the human-authored text and in the author's selection, coordination and arrangement of written and visual elements, while excluding the Midjourney-generated images from the scope of the registration on the ground that those images were not the product of human authorship.³
The decision is significant for two reasons.
First, the use of generative AI did not render the entire work ineligible for protection. Human-authored components and qualifying creative arrangements remained capable of protection.
Secondly, the protection recognised for selection and arrangement did not establish that the user automatically owned copyright in every AI-generated image incorporated into the work. The Office distinguished the copyrightable human contribution from the material it considered to lack human authorship.
The decision therefore supports a differentiated approach. It does not conclusively determine how Indian law should interpret Section 2(d)(vi), but it demonstrates why the analysis should identify the particular elements claimed to be protected instead of classifying an entire work as either wholly copyrightable or wholly unprotected merely because AI was used.
Part II — The Structural Tensions in the Conventional Analysis
1. The tool–creator tension
Artificial intelligence systems are not ordinarily recognised as legal persons capable of independently holding copyright, entering into contracts or bearing legal obligations in their own capacity. Yet an objection to human authorship may arise from the proposition that the system, rather than the user, determined the precise expression appearing in the output.
This creates a conceptual tension, though not necessarily a formal contradiction. The law may deny legal personality to a machine while concluding that a particular expression was not authored by the human who initiated its production. The absence of a machine author does not logically compel the conclusion that the user authored the entire output.
The question must instead be approached through the applicable statutory definition of authorship. Where the statute identifies the person who causes a computer-generated work to be created, the absence of machine personhood does not answer the question; it makes the statutory identification of the relevant human agency particularly important.
The distinction is between causal responsibility and authorship. A person who presses a button may cause a machine to operate without becoming the author of every result. Conversely, a person who designs the expressive structure, directs the production process, makes consequential choices and materially shapes the final work may have a substantially stronger claim to authorship.
The law should therefore resist both extremes. It should neither attribute all generated expression automatically to the user nor infer from the system's computational role that human authorship is necessarily absent.
2. The statutory significance of the person who causes the work to be created
Section 2(d)(vi) of the Copyright Act, 1957 states:
“in relation to any literary, dramatic, musical or artistic work which is computer-generated, the person who causes the work to be created.”
This provision is the principal statutory foundation for the argument advanced in this article.⁴
Its language does not expressly require that the person personally type every word, draw every line or manually execute every technical operation involved in producing the work. It directs attention instead to the person who causes the work to be created.
The expression “causes the work to be created” nevertheless requires judicial interpretation when applied to generative AI. Causation in the ordinary sense may be broad: a person may initiate a process, finance it, provide instructions or accept its result. Copyright authorship, however, cannot be determined solely by identifying the earliest or most remote cause of production.
The more persuasive interpretation would examine the nature and effectiveness of the human contribution. Relevant considerations may include:
the specificity and expressive content of the instructions;
the extent to which the human establishes the structure, arrangement, characterisation or other expressive features;
the number and substance of iterative revisions;
the human's selection, rejection and transformation of generated material;
the degree to which the final work reflects the human's demonstrable creative choices; and
the distinction between the material generated by the system and the original expression subsequently contributed by the human.
These considerations are proposed as an analytical framework, not as an established judicial test under Indian law.
A general prompt may establish a causal connection without establishing authorship of the resulting expression. Detailed direction and meaningful editorial intervention may provide stronger evidence, although the legal conclusion must still be reached under the statute and the applicable originality standard.
The critical question is not whether the human contributed something to the process, but whether the evidence establishes the contribution necessary to support the claim made. Section 2(d)(vi) should neither be read out of the statute nor converted into an automatic rule that every person initiating a computer process owns its entire output.
3. The originality requirement: Eastern Book Company v. D.B. Modak
In Eastern Book Company v. D.B. Modak, the Supreme Court of India examined the originality required for copyright protection in edited and reported judicial decisions. The Court rejected the proposition that labour and expenditure alone were sufficient and required the exercise of skill and judgment accompanied by a sufficient degree of creativity.⁵
The judgment is important to AI-assisted authorship, but its significance must be carefully confined. It does not decide whether a person who prompts an AI model is the author of its output. It establishes principles relevant to the originality of the claimed contribution.
The case supports the proposition that copyright cannot be claimed merely because a person has expended time, money or effort. The contribution must satisfy the applicable standard of originality. At the same time, the requirement of originality directs attention to intellectual choices embodied in the work rather than to the physical mechanism by which the work was produced.
An author may exercise creative judgment through selection, revision, arrangement and refinement. Those activities may be relevant to copyright protection where they result in original expression. But their mere existence is insufficient: the claimed contribution must be identified, and its originality must be demonstrated.
Applied to AI-assisted writing, the distinction is between the effort of operating a system and the intellectual contribution expressed in the resulting work. Repeatedly generating alternatives and choosing one without more may not establish authorship of the selected output. Substantive rewriting, original composition, creative arrangement or other demonstrable human expression may support a claim to protection in the material contributed by the human.
The principle derived from Eastern Book Company is therefore not that extensive prompting automatically creates copyright. It is that human contribution must be assessed for its intellectual substance and originality rather than dismissed merely because technological assistance was employed.
Part III — Comparative Precedents and Their Proper Reach
1. Burrow-Giles Lithographic Co. v. Sarony: technology as an instrument of expression
In Burrow-Giles Lithographic Co. v. Sarony, 111 U.S. 53 (1884), the United States Supreme Court considered whether a photograph could qualify for copyright protection. The Court upheld protection for the photograph at issue, in which the photographer had exercised creative judgment in posing the subject, selecting and arranging accessories, managing light and shade, and producing the final composition.⁶
The decision is an important historical example of copyright law responding to a technology that could be characterised as mechanical. It demonstrates that the use of an apparatus does not necessarily displace human authorship where the resulting work embodies the photographer's original intellectual conception.
The analogy to generative AI is persuasive but not complete. In the photograph before the Court, the photographer's creative contribution to the composition was established by evidence. In a generative AI process, the system may determine expressive details that the user neither selected in advance nor controlled during production.
The correct lesson from Burrow-Giles is therefore not that every machine-generated result is copyrightable because a human operated the instrument. It is that the technological character of the process does not, by itself, settle the question of authorship. The actual contribution of the human must be examined.
The decision strengthens the argument for a functional inquiry into creative control. It does not eliminate the requirement that the claimed human authorship be established.
2. The analogy to photography and other creative instruments
The history of copyright includes works produced through cameras, recording equipment, editing systems and computer software. In each instance, the instrument may perform technical operations without necessarily becoming the author of the resulting work.
Generative AI presents a more difficult problem because the system may do more than record or execute a predetermined human composition. It can produce expressive alternatives that are not individually specified by the user. That difference matters.
The legal question should not be resolved merely by describing AI as a tool, because the label does not establish how much of the expression the human determined. Nor should the technology be treated as an independent legal author simply because it generates details beyond the user's precise instructions.
A principled analysis asks whether the human's contribution, considered in relation to the particular work claimed, meets the statutory and originality requirements. The instrument may be technologically sophisticated; the legal inquiry remains one of attribution, expression and proof.
3. Selection and arrangement as independent human contributions
A person may assemble a larger work from material obtained through multiple sources, including material generated by AI. The selection and arrangement of that material may contain original human expression. Under the principles illustrated by Eastern Book Company and the United States Copyright Office's decision in Zarya of the Dawn, such contributions must be examined independently from the underlying material.
This distinction is especially important for authors who use AI during the development of a manuscript. The human may establish the architecture of a book, determine the sequence of chapters, create characters and themes, reject unsuitable passages, rewrite generated text and construct a coherent narrative. The resulting work may contain protectable human expression even if some constituent material was generated by a machine.
The extent of protection will depend on the actual contribution and applicable law. Copyright in an original selection or arrangement does not automatically confer exclusive rights over every constituent element, particularly where those elements are themselves unprotected or independently created by another source.
The argument for human authorship is strongest when the author can identify the original expression contributed, the editorial decisions made and the relationship between those decisions and the final form of the work.
Part IV — Authorship and Ownership Must Not Be Confused
1. Section 17(c): the employment provision
Section 17 of the Copyright Act, 1957 begins with the general rule that, subject to the Act, the author of a work is its first owner. Section 17(c) provides an exception for a work made in the course of the author's employment under a contract of service or apprenticeship, where the specified preceding exceptions do not apply. In such a case, the employer is the first owner in the absence of an agreement to the contrary.⁷
This provision illustrates an important distinction between authorship and first ownership. The person who creates a work and the person who first owns the copyright need not invariably be the same person. The statute expressly allocates ownership in certain relationships and circumstances.
It would, however, be incorrect to equate a paid AI subscription with employment under a contract of service. A subscriber ordinarily does not employ the AI platform as a servant or apprentice, and the platform's provision of computational services does not by itself establish the legal relationship contemplated by Section 17(c).
The economic analogy has limited but legitimate force. A human author who uses an AI platform may supply the intellectual direction, pay for the service, evaluate the outputs and decide what enters the final work. Those circumstances may be relevant to understanding the human's role in production. They do not, without more, establish ownership under the employment provision.
The sounder argument is that ownership must be determined by the applicable statutory rules, the nature of the work, any relevant agreement and the extent of the human contribution. Section 17(c) cannot be used to create an employment relationship where none exists.
2. The subscription payment and the question of ownership
Payment for access to a tool is not equivalent to payment for copyright in every product generated through it. A subscription may confer contractual rights to use a service and may be accompanied by terms governing outputs. Those contractual terms cannot, by themselves, resolve every question of statutory authorship or copyright subsistence.
The subscriber's payment may demonstrate that the service was acquired for the subscriber's use. It does not establish that every generated output is original, that the subscriber authored every expressive element or that copyright necessarily subsists in the output.
Conversely, the fact that the system was developed and operated by a third party does not automatically establish that the third party is the author of every output. The legal position must be determined by examining the relevant statutory provisions, the human contribution, the nature of the output and any applicable contractual arrangements.
The distinction is fundamental: access to a system, authorship of a work, subsistence of copyright and ownership of the resulting rights are separate questions.
3. The author's claim must be supported by evidence
Where the claim to authorship depends on the human's direction and editorial intervention, the author should be able to demonstrate the nature of that contribution. Drafts, successive versions, records of substantive revisions, structural outlines and evidence of original selection or arrangement may assist in establishing the human role.
Such records do not automatically prove copyright, nor does the absence of a particular type of record necessarily defeat a claim. Their significance lies in their capacity to show how the work came into existence and which expressive decisions are attributable to the human.
A claim based only on the assertion that the user supplied prompts or paid for a subscription may be weak. A claim supported by identifiable original expression, meaningful revision and evidence of creative control may present a stronger case.
The law should assess the substance of the contribution rather than the mere presence or absence of AI in the creative process.
Part V — A Proposed Framework for Indian Copyright Analysis
The existing statutory scheme permits a disciplined analysis without either assuming that all AI outputs are protected or declaring all such outputs unprotected.
The following framework is proposed for judicial consideration and further scholarly debate.
First: identify the work claimed. The claimant should specify whether copyright is asserted in the complete work, particular passages, a selection or arrangement, or human modifications of generated material. A claim framed at the appropriate level avoids treating a composite work as indivisible.
Secondly: identify the human contribution. The inquiry should establish the instructions, design choices, original composition, selection, arrangement, revision and other expressive decisions attributable to the claimant. General ideas and instructions should not be confused with the expression claimed to be protected.
Thirdly: consider Section 2(d)(vi). Where the work is computer-generated within the meaning of the provision, the court should examine who caused it to be created. That inquiry should be informed by the actual production process, without assuming that mere causal initiation necessarily establishes authorship of every expressive element.
Fourthly: apply the originality requirement. Under the principles of Eastern Book Company, labour, expense and technical effort alone are insufficient. The identified human contribution must meet the applicable standard of originality.
Fifthly: distinguish subsistence from ownership. Once the existence and scope of copyright are assessed, the court should determine first ownership under Section 17 and any relevant agreement or other statutory provision. Ownership should not be assumed merely from payment for the technology or from the fact that the claimant initiated the process.
Sixthly: define the scope of protection. Where a work combines human-authored material and AI-generated content, the analysis should identify the expression in which copyright is claimed. Protection in original human contributions should not automatically be extended to material for which the statutory requirements have not been established.
This framework does not create a new statutory test. It organises the existing questions of authorship, causation, originality and ownership so that each is addressed on its own terms.
Part VI — The Economic and Jurisprudential Stakes
Copyright law must respond to technological change without abandoning the principles that justify protection. If every output generated with AI assistance were categorically excluded from copyright, original human expression embedded in AI-assisted works could be denied protection merely because the author employed a modern instrument.
At the other extreme, if every person who submits a prompt were automatically treated as the author of the resulting expression, the originality requirement could be diluted and claims to protection could extend far beyond the human contribution actually made.
Neither outcome is satisfactory. The first risks disregarding genuine human creativity; the second risks converting the act of operating a service into a claim of authorship over expression determined by the system.
The appropriate response is not to replace the statutory framework with a presumption in favour of either the machine or the user. It is to apply the law to the particular work, identify the contribution claimed and determine whether the statutory conditions for protection have been met.
The question also has economic consequences. Authors, publishers, artists, researchers and businesses require clarity about the rights they may claim in AI-assisted material. Uncertainty can discourage legitimate creative investment, complicate licensing and produce disputes over the use of composite works. Clearer guidance on the relationship between human contribution and computer-generated expression would assist both creators and users of the technology.
Yet economic investment alone cannot determine authorship. Copyright is a statutory right governed by the conditions prescribed by law. The policy objective must be to recognise genuine human creativity while preserving the distinction between original expression and the mere production of material by an automated process.
Conclusion: The Mind Behind the Instrument
The authorship paradox arises when the law is asked to reconcile two facts: a machine may generate the immediate expression of a work, yet the process may have been initiated and shaped by a human being whose intellectual choices are reflected in the final result.
The answer cannot be supplied merely by describing artificial intelligence as a tool. Nor can it be supplied by attributing independent authorship to a system that the law does not recognise as a legal person. The decisive inquiry must be directed to the applicable statute, the work claimed, the human contribution and the originality of the expression.
Indian law possesses a provision of particular significance in this debate. Section 2(d)(vi) identifies, for a computer-generated work, the author by reference to “the person who causes the work to be created”. That language deserves substantive interpretation in the context of generative AI. It should neither be deprived of practical effect nor read as granting automatic copyright to every person who initiates a computational process.
The jurisprudential task is to distinguish mere prompting from meaningful creative direction; technical execution from original expression; authorship from first ownership; and the existence of a work from the scope of rights that may subsist in it.
The law should not presume that the machine has authored everything simply because it has executed the process. Equally, it should not presume that the human has authored everything merely because the human initiated it.
The central principle is narrower, but more enduring: where a human being can establish an original contribution to a work, the use of artificial intelligence should not, by itself, defeat the claim to protection. Where the claim extends to expression determined by the system, the statutory requirements must still be satisfied.
The future of copyright in the age of artificial intelligence will depend not upon choosing between human intelligence and machine execution, but upon identifying, with legal precision, the contribution for which protection is claimed and the person to whom the law attributes it.
Authorities and statutory references
Copyright Act, 1957, § 2(d)(vi). See the statutory text published by India Code and the Government of India's Copyright Office.
Thaler v. Perlmutter, No. 25-449, petition for a writ of certiorari denied 2 March 2026, Supreme Court of the United States; see also the judgment of the United States Court of Appeals for the District of Columbia Circuit, Thaler v. Perlmutter, No. 23-5233 (18 March 2025).
United States Copyright Office, Zarya of the Dawn, Registration No. VAu001480196, letter dated 21 February 2023.
Copyright Act, 1957, § 2(d)(vi).
Eastern Book Company v. D.B. Modak, (2008) 1 SCC 1; AIR 2008 SC 809. The Supreme Court's judgment was delivered on 12 December 2007.
Burrow-Giles Lithographic Co. v. Sarony, 111 U.S. 53 (1884).
Copyright Act, 1957, §§ 17 and 17(c).
