GENERATIVE AI AND THE MODICUM OF CREATIVITY: EVALUATING ART CREATION AND COPYRIGHT ELIGIBILITY

ABSTRACT

The evolving paradigm of Generative Artificial Intelligence (GAI) has ignited a profound global discourse regarding copyright eligibility, fundamentally challenging the long-held requirement of human authorship.[1] This paper critically examines the threshold of originality—the “minimum modicum of creativity” as interpreted and applied under United States, European Union, and Indian copyright laws—specifically through case studies of AI-generated and AI-assisted artistic works.[2] The analysis reveals that works emerging from mere suggestion or generic prompts typically fail to meet the requisite standard for creative expression and thus remain unprotected.[3] Conversely, works demonstrably shaped by substantial, iterative human control, or those where GAI serves as a sophisticated enhancement tool for pre-existing original material, unequivocally satisfy the originality threshold and warrant copyright protection.[4] The decisive factor in copyright eligibility is thus firmly established as the demonstrable degree of human contribution and authorship over the work’s expressive elements, alongside the inherent “human touch” principle that underpins copyright jurisprudence across these jurisdictions.[5]

Keywords: Generative Artificial Intelligence, Originality Threshold, AI-Generated Works, Human Contribution, Copyright Eligibility.

I. INTRODUCTION

The rapid ascent of Generative Artificial Intelligence (hereinafter referred to as GAI), capable of autonomously producing intricate artistic outputs, profoundly challenges copyright law’s fundamental tenets globally.[6] Copyright protection, traditionally reserved for “original creative works of an author expressed in a tangible medium,” demands a “minimum modicum of creativity” originating from a human author.[7] Legal systems in the US, EU, and India affirm copyright’s anthropocentric nature, linking protection directly to human intellectual endeavor and expression in order to stimulate and promote human creativity.[8]

This paper addresses the pressing issue of determining authorship and originality when creative output is co-generated by human instruction and algorithmic processing.[9] The central question is whether generic commands suffice to confer protection, or if sustained, iterative human engagement—akin to using a sophisticated tool—is necessary for copyright to subsist.[10] This differentiation is critical given the exponential increase in AI-generated content, which risks saturating the creative market and diluting human creative value without careful management of monopolistic rights.[11] Scholars have underscored GAI’s disruptive challenge to these foundational principles.[12]

Our scope distinguishes genuine human–AI collaboration from minimal intervention by analyzing the creativity threshold across the US and India, where human creative choices remain essential.[13] This paper argues that copyright eligibility for GAI works hinges on human creative control over expressive elements, distinguishing between AI as an autonomous source (unprotected) and AI as a technical tool (where human authorship is preserved).[14]

II. GENERATIVE AI AND MODES OF ART CREATION

Generative AI systems learn from vast datasets of human-created content, synthesizing novel works based on statistical probabilities and explicit instructions.[15] This advanced capability has given rise to two primary modes of art creation that profoundly challenge conventional authorship laws:

A. Prompt-Based Generation (Minimal Intervention)
In this mode, a human provides a concise, often single-step command (e.g., “A vibrant sunset over a futuristic city”) to an AI system. The AI then executes the majority of creative decisions, including style, composition, color, and texture. Here, human input is primarily judged for creativity within the textual prompt itself. The AI’s generative decisions are largely automated and lack autonomous creative intent.[16] Outputs of this nature often lack a strong “causal connection” between the human creator and the expressive aspects of the work due to algorithmic random choices.[17] The United States Copyright Office (USCO) clarifies that if traditional elements of authorship are determined and executed by the technology rather than the human user, the work is generally not copyrightable.[18]

B. Iterative/Assisted Creation (Significant Intervention)
This mode is characterized by continuous and substantial human engagement. It encompasses:
1. Iterative Prompting: The user refines and modifies prompts over numerous steps, actively guiding the AI toward a specific artistic vision.
2. AI-Assisted Enhancement: The AI functions as a sophisticated tool—similar to editing software—to colorize, compile, or enhance pre-existing, human-authored material. This relies on original expressive inputs from the human that are perceptibly maintained and transformed in the final output.

The core of the authorship debate identifies the true origin of creative choices.[19] The USCO draws a clear line: AI used as an assistive tool does not impede copyright, but if the AI itself makes the core expressive choices, the work lacks human authorship.[20] Even when AI outputs create a “mirage” of creativity, they often lack the internal understanding and awareness of genuine human expression.[21] Thus, the critical question becomes whether the human exercised sufficient non-trivial skill and judgment to claim authorship.[22] As legal scholarship established early in the computing age, the key lies in determining whether the work is basically one of human authorship with the computer acting as an assisting instrument, or whether the traditional elements of authorship were conceived and executed by a machine.[23]

III. THE MINIMUM MODICUM OF CREATIVITY STANDARD

The “minimum modicum of creativity” standard serves as a universal threshold for copyright protection across major international jurisdictions.[24] This standard mandates that a creation must stem from a human author possessing an “original spark of creativity” distinct from mere mechanical, trivial, or utilitarian functions.[25]

A. United States Law: The “Bedrock Requirement” of Human Authorship
In the United States, this standard is rooted in the landmark Supreme Court decision Feist Publications, Inc. v. Rural Telephone Service Co., which established that originality requires independent creation by the author along with at least a modicum of creativity.[26] While this bar is intentionally low, it is consistently coupled with the requirement of human authorship. The USCO has explicitly affirmed that copyright protection is contingent upon human authorship, denying registration to works produced solely by AI for lacking the fruits of intellectual labor.[27] This position was further cemented in Thaler v. Perlmutter, where the U.S. District Court affirmed that human authorship is a bedrock requirement under the Copyright Act.[28] Consequently, works generated from simple prompts are generally uncopyrightable, as the prompt conveys an unprotectable idea while leaving the AI to dictate expressive elements.[29]

B. Indian Law: “The Person Who Causes the Work to Be Created”
India’s legal framework transitioned from the “Sweat of the Brow” doctrine to the “Modicum of Creativity” approach following the Supreme Court’s ruling in Eastern Book Company v. D.B. Modak.[30] The Court clarified that a work must exhibit a sufficient amount of intellectual creativity and judgment to warrant protection.[31] Crucially, Section 2(d)(vi) of the Indian Copyright Act, 1957 defines the author of a computer-generated work as “the person who causes the work to be created.”[32] This statutory provision provides a legal basis for recognizing human agency by treating AI as a technical instrument assisting a human creator, provided there is a clear demonstration of human skill, judgment, and effort beyond the trivial.[33]

IV. CASE STUDY ANALYSIS OF AI-GENERATED AND AI-ASSISTED WORKS

This section examines three distinct works exemplifying varying degrees of human creative involvement through the analytical lens of established originality standards drawn from judicial rulings and copyright office guidance.[34]

Work 1: Prompt-Based Digital Painting
Legal Analysis: This digital painting represents a pure prompt-based output, presenting an inherently weak case for copyright protection.[35] The human input primarily functioned as an instruction conveying an idea and emotional parameters—concepts unprotectable under copyright law.[36] Key expressive elements such as the color palette, rendering texture, lighting, and compositional choices were entirely conceived and executed by the AI algorithm.

The USCO maintains that a textual prompt, even if highly descriptive, does not confer sufficient creative control when traditional elements of authorship are executed by technology.[37] This aligns with Thaler v. Perlmutter and the foundational principle that copyright protects expression rather than ideas—a doctrine affirmed in Indian jurisprudence in R.G. Anand v. Deluxe Films.[38] International precedents, such as the Chinese decision in Beijing Feilin Law Firm v. Baidu, Inc., similarly hold that selecting an AI output is insufficient for authorship if the AI autonomously executes the artistic expression.[39] Furthermore, as noted in the USCO’s decision regarding Zarya of the Dawn, where output is unpredictable from the prompt, the user lacks sufficient control over the expressive outcome.[40]

Conclusion: Not Copyrightable. The human input fails to constitute an original work of authorship because expressive choices were primarily made by the AI system.[41]

Work 2: Prompt-Based Poetry with Human Intervention
Legal Analysis: This work presents a strong case for protection grounded in human revision and selective modification.[42] While the initial AI draft is unprotectable, the human author’s conscious selection, arrangement, and modification introduce an original contribution that satisfies the modicum of creativity threshold.[43]

The USCO grants copyright where a human modifies AI-generated material to such a degree that the modifications meet standard protection requirements.[44] In the Zarya of the Dawn registration, the USCO accepted human selection and arrangement of elements while excluding individual AI images.[45] Extensive revisions—such as changing key adjectives, introducing new metaphors, and refining cadence—demonstrate human skill and judgment beyond trivial alterations.[46] Editorial contributions to non-human or compilation sources have long been protected in decisions such as Penguin Books USA, Inc. v. New Christian Church of Full Endeavor, Ltd. and Urantia Foundation v. Maaherra.[47]

Conclusion: Copyrightable (limited to human-authored revisions and creative choices).[48]

Work 3: AI-Assisted Collage of Original Sketches
Legal Analysis: This scenario represents AI functioning strictly as an assistive tool for enhancing pre-existing original expression.[49] Originality resides in the human creator’s original hand-drawn sketches.[50] The AI operates as a technical enhancement tool, analogous to digital editing software or camera mechanics.[51]

Key creative decisions—selecting initial sketches, establishing visual style parameters, and arranging the final collage layout—were executed by the human.[52] This curated arrangement creates a protectable compilation reflecting human skill and judgment, consistent with the principle in Burrow-Giles Lithographic Co. v. Sarony.[53] Under Section 2(d)(vi) of the Indian Copyright Act, 1957, the user who causes the work to be created retains authorship when directing an instrument to refine original human expressions.[54]

Conclusion: Copyrightable as a human-authored derivative work and compilation.[55]

V. COMPARATIVE EVALUATION AND THEORETICAL IMPLICATIONS

The case studies confirm that copyright eligibility for GAI-involved works depends fundamentally on the extent of discernible human creative control.[56]

Tool vs. Independent Creator
When AI autonomously generates fundamental expressive elements (as in Work 1), it operates as the de facto creator, leaving the output uncopyrightable due to the absence of a human author.[57] This principle aligns with rulings such as Thaler v. Perlmutter, Naruto v. Slater (affirming that non-humans lack statutory standing as authors), and Acohs Pty Ltd v. Ucorp Pty Ltd (denying copyright to automated system outputs).[58]

Conversely, when a human maintains creative control by editing AI outputs (Work 2) or applying AI tools to original inputs (Work 3), the technology acts as a modern instrument.[59] Copyright vests in the human author whose expressive choices remain central to the final work.[60]

VI. CONCLUSION

Copyright eligibility for Generative AI outputs fundamentally depends on human creative control.[61] Works produced solely from generic prompts fail to meet the originality requirement, as they represent unprotectable ideas.[62] In contrast, works involving substantial human editorial revision or AI-assisted enhancement of original material satisfy the modicum of creativity standard established in Feist and adopted in India via Eastern Book Company v. D.B. Modak.[63]

These findings reinforce the anthropocentric foundation of copyright law.[64] While GAI can simulate expressive styles, legal protection remains contingent on human agency.[65] Future legislative developments should focus on providing statutory clarity—potentially through sui generis frameworks—to address automated outputs without eroding protections for human creators.[66]

ANNEXURE A: PROMPTS AND INPUTS USED IN CASE STUDIES

Work 1: Prompt-Based Digital Painting
Prompt 1: “Create a digital painting of a girl sitting behind the table with a scared face showing the expression and feelings of social taboo she is facing from the society and hesitating to go out for work.”
Prompt 2 (Iterated): “Sitting behind at the floor hiding herself behind the table legs folding her legs with her hands, and the image must be a dark image with less lighting and picture must be sort of a sketch appearance, a colourful sketch not with black and white.”

Work 2: Prompt-Based Poetry with Human Intervention
Initial Prompt: “Write a poem on father my strength, my calm anger chaos, my everything. I cannot live happily without him, backbone of our family, light and love of our family short 8 lines and it should sound like a writer.”
AI Initial Output:
“My strength when I feel weak,
My calm when storms may speak.
In chaos, you are my ground,
In silence, your love is found.
The backbone of all we do,
Our guiding light shining true.
Without you, joy fades away,
Our everything, day by”
Human Intervention: Revised key adjectives, adjusted rhythm, and added custom metaphors (e.g., modifying lines to “steadfast anchor in life’s turbulent sea” and adding “A guiding star, forever shining bright, illuminating my darkest night”).

Work 3: AI-Assisted Collage of Original Sketches
Inputs: Original hand-drawn pencil sketches.
Prompt Used: “Enhance, improve lighting, colour, and cinematic effect, then make a collage.”

REFERENCES

[1] European Parliamentary Research Service, Intellectual Property Rights and Generative AI, EPRS Study PE 754.184 (2023).
[2] U.S. Copyright Office, Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence, 88 Fed. Reg. 16190 (Mar. 16, 2023).
[3] Jane C. Ginsburg & Luke Ali Budiardjo, Authors and Machines, 34 Berkeley Tech. L.J. 343, 355 (2019).
[4] Pam Samuelson, Allocating Ownership Rights in Computer-Generated Works, 47 U. Pitt. L. Rev. 1185 (1986).
[5] Daniel J. Gervais, The Machine As Author, 105 Iowa L. Rev. 2053, 2060 (2020).
[6] Ryan Abbott & Elizabeth Rothman, Disrupting Creativity: Copyright Law in the Age of Generative AI, 2023 Fla. St. U. L. Rev. Online 1 (2023).
[7] 17 U.S.C. § 102(a) (2018).
[8] Christophe Geiger et al., The Human-Centric Approach to Copyright in the AI Era, 52 IIC 1101 (2021).
[9] Ginsburg & Budiardjo, supra note 3, at 380.
[10] Arthur R. Miller, Copyright Protection for Computer Programs, Databases, and Artificial Intelligence Systems, 106 Harv. L. Rev. 977 (1993).
[11] Vicenç Feliú, AI, Copyright, and Market Saturation in the Digital Era, 18 J. Intell. Prop. L. 45 (2022).
[12] Abbott & Rothman, supra note 6, at 12.
[13] Indian Copyright Act, No. 14 of 1957, § 2(d)(vi), INDIA CODE (1957).
[14] U.S. Copyright Office, supra note 2, at 16192.
[15] Murray Shanahan, Talking About Large Language Models, 66 Comm. ACM 82 (2023).
[16] Hako Zurth, Artificial Authorship and the Illusion of Intent, 14 Eur. J. L. & Tech. 112 (2023).
[17] Feliú, supra note 11, at 52.
[18] U.S. Copyright Office, supra note 2, at 16191.
[19] Ginsburg & Budiardjo, supra note 3, at 390.
[20] U.S. Copyright Office, supra note 2, at 16192.
[21] Zurth, supra note 16, at 118.
[22] Gervais, supra note 5, at 2075.
[23] Abraham L. Kaminstein, Register of Copyrights Annual Report, U.S. Copyright Office (1965).
[24] Berne Convention for the Protection of Literary and Artistic Works, art. 2, Sept. 9, 1886, 828 U.N.T.S. 221.
[25] Feist Publ’ns, Inc. v. Rural Tel. Serv. Co., 499 U.S. 340, 345 (1991).
[26] Id. at 346.
[27] U.S. Copyright Office, Compendium of U.S. Copyright Office Practices § 313.2 (3d ed. 2021).
[28] Thaler v. Perlmutter, 687 F. Supp. 3d 140, 146 (D.D.C. 2023).
[29] Id. at 148.
[30] Eastern Book Co. v. D.B. Modak, (2008) 1 SCC 1 (India).
[31] Id. at 102.
[32] Indian Copyright Act, No. 14 of 1957, § 2(d)(vi), INDIA CODE (1957).
[33] Modak, (2008) 1 SCC at 105.
[34] U.S. Copyright Office, supra note 2, at 16191.
[35] Thaler, 687 F. Supp. 3d at 147.
[36] 17 U.S.C. § 102(b) (2018).
[37] U.S. Copyright Office, supra note 2, at 16192.
[38] R.G. Anand v. Deluxe Films, (1978) 4 SCC 118 (India).
[39] Beijing Feilin Law Firm v. Baidu, Inc., Beijing Internet Court (2018) Jing 0491 Min Chu No. 239 (China).
[40] U.S. Copyright Office, Cancellation Decision re: Zarya of the Dawn (Feb. 21, 2023).
[41] Thaler, 687 F. Supp. 3d at 149.
[42] U.S. Copyright Office, supra note 2, at 16193.
[43] Feist, 499 U.S. at 348.
[44] U.S. Copyright Office, supra note 2, at 16193.
[45] U.S. Copyright Office, Cancellation Decision re: Zarya of the Dawn, supra note 40.
[46] Modak, (2008) 1 SCC at 108.
[47] Penguin Books USA, Inc. v. New Christian Church of Full Endeavor, Ltd., 2000 WL 1028634 (S.D.N.Y. 2000); Urantia Found. v. Maaherra, 114 F.3d 955 (9th Cir. 1997).
[48] U.S. Copyright Office, supra note 2, at 16193.
[49] Miller, supra note 10, at 985.
[50] 17 U.S.C. § 102(a) (2018).
[51] Burrow-Giles Lithographic Co. v. Sarony, 111 U.S. 53, 60 (1884).
[52] Id. at 61.
[53] Id.
[54] Indian Copyright Act, No. 14 of 1957, § 2(d)(vi), INDIA CODE (1957).
[55] 17 U.S.C. § 103 (2018).
[56] Gervais, supra note 5, at 2080.
[57] Thaler, 687 F. Supp. 3d at 146.
[58] Naruto v. Slater, 888 F.3d 418, 426 (9th Cir. 2018); Acohs Pty Ltd v. Ucorp Pty Ltd, [2012] FCAFC 16 (Austl.).
[59] Burrow-Giles, 111 U.S. at 59.
[60] Modak, (2008) 1 SCC at 109.
[61] U.S. Copyright Office, supra note 2, at 16191.
[62] Feist, 499 U.S. at 350.
[63] Modak, (2008) 1 SCC at 102.
[64] Geiger et al., supra note 8, at 1105.
[65] Zurth, supra note 16, at 122.
[66] Gervais, supra note 5, at 2095.