Who Owns The Input - Prompt Engineering And The Expanding Edge Of Copyright Law
Author: LawSphereX
Category: Intellectual Property
Published:
Updated:
WHO OWNS THE INPUT? PROMPT ENGINEERING AND
THE EXPANDING EDGE OF COPYRIGHT LAW
Parvati Arun, Institute of Law, Nirma University 1 I.
INTRODUCTION
In March 2023, the US Copyright Office held that while AI-generated works are not generally eligible for copyright protection, a copyright application for a work created with the help of AI can support a copyright claim if a human “selected or arranged” it in a “sufficiently creative way that the resulting work constitutes an original work of authorship. ”2 Despite this ruling , the criteria for what constitutes a “creative work” hinges on an ambiguous territory necessitating the need for further scholarship regarding the complex intertwining of human creativity with AI authorship. The advent of generative artificial intelligence (Gen AI) systems has fundamentally altered the creative landscape as users are able to interact with advanced AI-inputs by crafting effective prompts resulting in structured high-quality outputs.
AI systems such as GPT-4, DALL-E and Midjourney are trained to generate new texts, images or audio outputs employing their data models through well structured prompts. However, these prompts which are often overlooked in legal scholarship and policy debates, are sophisticated, formalized expressions that require human creativity, skill and iteration to produce accurate, stylised or user-targeted outputs. This emerging practice which is popularly known as prompt engineering occupies a liminal space between authorship and instruction, creativity and function, user and system. 3 This paper aims to focus on under explored questions regarding the legal status of the prompt such as can the prompt-as the textual seed of a creative work be recognised as a protectable
1 Author is a 5th year law student at Institute of Law, Nirma University
2 Adrian Bowyer, Human Creativity and AI: The Foundation That Makes the Difference, LINKEDIN (May 15, 2025), https://www.linkedin.com (accessed July 16, 2025).
3 Irene Solaiman et al., Gradient Institute’s Responsible AI Prompting Guide (2023),
https://www.gradientinstitute.org (accessed July 16, 2025).
expression of authorship under copyright law? 4 Should the prompt engineer be regarded as a contributing author or simply a user directing an automated tool? 5 This paper centres upon prompt engineering within a legal framework- particularly in the context of elements such as originality, ownership and fixation. It contends that certain categories of prompts which are designed for with aesthetic and narrative intent may amount to literary or artistic expression under copyright laws. This inquiry is conducted through doctrinal and comparative analysis. Part I of the inquiry consists of a technical overview of how prompt engineering models across major Gen AI platforms. Part II examines the legal tests governing copyrightability under the Indian copyright regime and evaluates their applicability to prompts.
6 Part III reflects upon comparative jurisprudence across United States, European Union and the United Kingdom, including developments through judgements and policy guidelines. Part IV addresses the larger implications of considering prompt engineers as authors (or co-authors), especially in cases involving multi-modal collaborative work, multi-source input or automated recombination. 7 Finally, the paper suggests policy recommendations to extend limited copyright protection or sui-generis recognition for highly structured prompts while cautioning against granting such protection to purely generic or functional outputs. 8 The goal of this paper is to shift copyright law's analytical lens away from output and toward input, away from simply what the machine is outputting and instead to examine what the human is contributing.
Given the fact that AI technologies increasingly mediate creativity, prompt engineering is the latest frontier for human authored works and requires the law to be nuanced and adaptable. 9 II.
WHAT IS PROMPT ENGINEERING?
Prompt engineering refers to the process where generative artificial intelligence solutions such 4 Pamela Samuelson, Allocating Ownership Rights in Computer-Generated Works, 47 U. PITT. L. REV. 1185 (1985). 5 World Intellectual Property Organization (WIPO), Revised Issues Paper on Intellectual Property Policy and Artificial Intelligence (2023), https://www.wipo.int/about-ip/en/artificial_intelligence/ (accessed July 8, 2025).
6 The Copyright Act, No. 14 of 1957, §§ 2(o), 2(d), 13, INDIA CODE (1957).
7 Jane C. Ginsburg, People Not Machines: Authorship and AI, 40 CARDOZO ARTS & ENT. L.J. 1 (2022).
8 Andres Guadamuz, AI and Copyright: Ownership and Originality, 44 EUR. INTELL. PROP. REV. 725 (2022).
9 Sonia Katyal, The Paradox of Source Code Secrecy, 104 CORNELL L. REV. 1183 (2019).
as GPT-4, DALL-E and Midjourney are guided through detailed, high-quality instructions to achieve desired outputs.10 It is increasingly recognised not as a mere operational technique, but as a creative process that involves aesthetic, linguistic, and strategic decision-making.11 Prompt engineering is powered by large machine learning models that employ deep neural networks which function on vast amounts of data and produce accurate responses with the aid of systematic and refined prompts. In general, prompts vary in terms of the complexity and degree of human creativity involved and do not exhibit uniform characteristics. This paper distinguishes prompts into three specific categories for analytical clarity such as: -
A. FUNCTIONAL PROMPTS:
Functional prompts are direct instructions which are provided in a simplified format to the AI model which require minimal input and demonstrate a lack of nuance over the subject. Functional prompts are primarily directives and qualify to be functional in nature within reasonable thresholds of copyrightability.
B. EXPRESSIVE OR STYLISED PROMPTS
The second category of prompts consist of prompts that are built based on narrative or creativity. Examples of such prompts include: "Write a poem about solitude in the style of Sylvia Plath, in broken metaphors, irregular meter and or post-war imagery." These prompts are built upon a creative framework and contain tone, pacing, reference points, genre conventions and style within the prompt instruction. In these cases, the prompt becomes more than an embedded tool; rather it emerges as a literary framework–like a screenplay, script or musical score. The process of design for such prompts may involve the exercise of personal judgement, taste and originality which fulfil the criteria for eligibility for authors under many copyright regimes.12
10 Amazon Web Services, What Is Prompt Engineering?, AMAZON WEB SERVICES (2024),
https://aws.amazon.com/what-is/prompt-engineering/ (accessed July 17, 2025).
11 Ethan Mollick & Lilach Mollick, How to Use AI to Do Stuff: An Opinionated Guide, WHARTON
INTERACTIVE (2023), https://writing.wharton.upenn.edu/how-to-use-ai-to-do-stuff/ (accessed July 17, 2025).
12 Eastern Book Company v D B Modak (2008) MANU/SC/0014/200; Infopaq International A/S v Danske
Dagblades Forening (C-5/08) [2009] ECR I-6569.
C. ITERATIVE AND CHAINED PROMPTS
Iterative and chained prompts employ advanced prompt engineering techniques by incorporating sequential control, modular design or refinement loops to improve precision, tone and relevance of the outputs. This is similar to an editorial or directorial process where a final output is co-created via structured and cumulative input. 13 Iterative prompt flows are prevalent in commercial applications, such as: Ø UX writing and developing a brand tone (for example, using AI to create a consistent marketing voice), Ø Game design (prompting for characters, quests, interactions), Ad copy and slogans, where the prompt is essentially the creative concept or big idea. These multi-layered prompts are deserving of legal scrutiny as they reflect a unique interaction of human intent and technical orchestration, which is prevalent over time.
The distinction between these categories is imperative as copyright law is more concerned with the process and expression rather than the outcome of a creative work. 14 If a prompt possessed a sufficient degree of skill, labour, or creativity, it assumes the characteristics of a work of authorship in its own right. 15 Crucially, this leads to a reconstruction of the human-machine relationship. Instead of merely focusing on the AI-generated output (which is denied protection on the basis that it lacks human authorship), this paper asserts that subject to certain conditions, prompts themselves may be literary works, especially when they exhibit:
1. A coherent narrative structure or style,
2. Aesthetic or poetic qualities,
3. Originality in arrangement or tone.
Shifting the focus from output to input creates a potential avenue for recognising human
13 Katyal, supra note 8.
14 Daniel J. Gervais, The Machine as Author, 105 IOWA L. REV. 2053 (2019).
15 Univ. of London Press Ltd. v. Univ. Tutorial Press Ltd., [1916] 2 Ch. 601 (U.K.).
agency in operating GenAI systems. It also challenges the notion that the user is merely “operating” the system as several cases involving advanced involvement of AI and human creativity have demonstrated that a user can “author” with it. III.
CAN PROMPTS BE COPYRIGHTED: THE DOCTRINAL TEST WITHIN
INDIAN CONTEXT
In determining the question of copyright eligibility for a prompt, it is necessary to assess it against the baseline doctrinal requirements of originality, authorship and fixation. While Indian copyright law does not specifically address generative AIs and prompts generated through such systems, it may extend protection to certain forms of prompt engineering under the Copyright Act,1957 with respect to comparative jurisprudence.
A. LITERARY WORKS AND THE INDIAN LEGAL FRAMEWORK:
Under Section 2(o) of the Copyright Act,1957; “literary works” consist of computer programmes, tables and compilations including computer databases apart from written works. 16 Prompts of a complex and stylistically unique nature imbibe the nature of written expression. Furthermore, when a prompt demonstrates creativity, linguistic expression and narrative intent, it can be regarded as a literary work, provided that it meets the criteria of originality under the Copyright Act,1957. In Eastern Book Company v. D B Modak, the Supreme Court of India adopted a “modicum of creativity” test for originality which adopts the ‘creativity’ test applied by the United States Supreme Court in Feist Publications Inc v Rural Telephone Service Co.
The Court determined that the characteristics of skill and labour, while relevant, is insufficient by itself as the work must demonstrate independent intellectual effort possessing a certain level of creativity. This signifies that while functional or mechanical prompts which are primarily directives may feel to meet the criteria, highly iterative or chained prompts which constitute narrative structure, stylistic expression, tone control or multi-modal layering may satisfy this.
B. FIXATION AND THE PROMPT AS WORK:
Unlike US law, Indian copyright law does not strictly require fixation in a material medium.17 However, written documentation, digital logs, or platform storage are frequently used to
16 The Copyright Act, No. 14 of 1957, § 2(o), INDIA CODE (1957).
17 17 U.S.C. § 102(a) (2024).
illustrate fixation in practice. Typically, history logs, chat threads, or downloadable project files contain the results of prompt engineering carried out on platforms such as OpenAI or Midjourney.18 In accordance with the fixation standards employed by Indian courts in cases involving ephemeral or computer-generated works, prompts are able to meet the requirements of identifiability and reproducibility because it allows to be digitally traced.19
C. THE AUTHORSHIP CHALLENGE:
Section 2(d) of the Copyright Act,1957 that a literary work author consists of the person who develops the work. The legal issue surrounding prompt engineering centres on whether prompts exist as creative works or they function solely as commands. The legal status of authorship depends on whether the prompt functions as creative expression or operates as a command. A prompt which involves detailed curation and execution of multiple iterations through stylistic prompts or chained prompts results in the user developing the original intellectual framework of the output. 20 The user can claim authorship of the prompt when it stands as an independent literary work even though the AI-generated output lacks protection under existing legal frameworks.
21 The legal framework might find the distinction between protected prompts and unprotected derivative outputs important from a doctrinal standpoint. The prompt functions as an independent expression which stands alone regardless of whether the final generated image or paragraph or design possesses copyright protection.
D. COMPARATIVE DOCTRINAL INSIGHTS:
The United States Copyright Office maintains that works created solely by artificial intelligence without human intervention cannot receive copyright protection.22 The Copyright Office recently denied full protection to Zarya of the Dawn by making a clear distinction between human-written text and computer-generated images.23 The CJEU established in 18 OpenAI, Managing Chat History, OPENAI (2024), https://platform.openai.com/account/history (accessed July 17, 2025).
19 Civic Chandran v. Ammini Amma MANU/KE/0036/1996
20 Solaiman et al., supra note 2
21 Ginsburg, supra note 6.
22 U.S. COPYRIGHT OFFICE, Copyright Registration Guidance: Works Containing Material Generated by
Artificial Intelligence (Mar. 2023). 23 Letter from U.S. Copyright Office to Kris Kashtanova (Feb. 21, 2023) (regarding Zarya of the Dawn), https://www.copyright.gov/ai/zip/kashtanova-letter.pdf.
Infopaq International v Danske Dagblades that originality depends on whether a work is the author’s “own intellectual creation”.24 According to the Infopaq test prompts with selection, arrangement and tone and creative judgement could qualify as original works even when the output is AI-generated. Different legal doctrines indicate that specific prompts can exist independently as copyrightable creations separate from their role in AI generation.
E. THE PROPOSED THREE-PART TEST:
The courts should use this proposed test to analyse prompt copyright eligibility which draws from established Indian legal principles.
1. Originality: The prompt needs to demonstrate original thinking through independent
evaluation instead of using basic factual or functional data.
2. Fixation: The prompt needs to be stored in either physical or digital format which
allows identification.
3. Purpose: The prompt must have an expressive goal to direct creative work instead of
performing automated operations. IV.
OWNERSHIP, LICENSING, AND PLATFORM CONSTRAINTS
The issue of ownership and control becomes complicated when prompt engineering takes place on closed AI platforms where terms of use take precedence over user rights, even if prompts satisfy the requirements for copyrightability. In most of these systems, such as Adobe Firefly, Midjourney, and OpenAI's ChatGPT, the platform provider either maintains derivative control or, in accordance with platform regulations, gives users limited-use licenses.25 For instance, OpenAI's terms stipulate that users maintain ownership of the content they submit, including prompts, but also mention that, unless they opt out, the company may use that input to train models.26 On the other hand, Midjourney has different policies based on the level of subscription and has the ability to assert more extensive usage rights over the outputs
24 Infopaq, supra note 11.
25 Rebecca Giblin & Kimberlee Weatherall, If AI Is the Author, Who Owns the Copyright? 37 I.P.J. 1 (2023). 26 OpenAI, Terms of Use, OPENAI (2024), https://openai.com/policies/terms-of-use (accessed July 18, 2025).
produced by user input.27 Observing from the legal prism, this reflects a difference between functional control through a contract and copyright ownership under the law.28 User agreements may limit the enforceability of rights even in cases where a prompt is copyrightable. Furthermore, concerns about joint authorship may surface in collaborative prompt chains or AI-assisted co-authorship, especially when several people provide consecutive prompts to generate a single output.29 Doctrinal clarification is needed for these new use cases on:
1. Whether, like other literary works, prompt authorship can be transferred or licensed.
2. If equal ownership of the output is implied by joint prompt authorship.
3. And if platforms need to record prompt metadata in order to prove provenance.
V.
CONCLUSION
Prompt engineering has emerged to be one of the most vital interfaces between human creativity and technology. The copyrightability of AI-generated outputs has received a lot of attention in legal discourse, but this paper generates attention on the input layer, where human users use narrative structure, aesthetic judgement, and iterative control to influence outcomes. Under the Indian Copyright Act of 1957, such prompts may be considered literary works, especially if they demonstrate originality, fixation, and expressive intent. However, recognising prompts as copyrightable also requires engagement with platform governance, contractual asymmetry, and doctrinal ambiguity in cases of joint authorship.
Indian copyright law must change as the lines separating human and machine creativity become increasingly hazy in order to guarantee that new modes of creative labour, such as prompt engineering, are properly acknowledged and safeguarded. This paper recommends that:
1. a legal test or legislative clarification for identifying prompts as literary works;
27 Midjourney, Terms of Service, MIDJOURNEY (2024), https://docs.midjourney.com/docs/terms-of-service (accessed July 18, 2025).
28 Ginsburg, supra note 6, at 20.
29 TANYA APLIN & LIONEL BENTLY, INTELLECTUAL PROPERTY LAW 126–29 (3d ed. 2021).
2. platform-level requirements, particularly for commercial use, to log and attribute
complex prompts.
3. standards for licensing that allow authors to keep control of prompts used in outputs
aided by artificial intelligence. The goal of acknowledging prompts as creative works is to honour the multi-layered human agency that endures in the era of algorithmic aesthetics, not to provide protection for machines.