I. Introduction: Art, AI, and Authors^[1]

Art is a tool to communicate messages. In the intellectual property (IP) world, the First Amendment affords some protection to copyrighted works because of the creative works’ communicative nature.^[2] Such a balance of First Amendment protections and rights extends to intellectual property that may not comply with traditional societal ideas of what “art” is.^[3] This is the case with the Rogers test, which is used to afford trademarks First Amendment protections.^[4] In fact, the Copyright Act provides copyright protection for “works of authorship” that includes a list of eight categories of protected art forms.^[5]

Understanding what a work of art is raises a second question: who (or what) is the author of a work of art? Remarkably, the Copyright Act does not provide a definition of “author,”^[6] despite authorship’s status as one of three requirements to register a work’s copyright with the U.S. Copyright Office.^[7] While the generality of the requirement that a work must be authored has remained consistent throughout U.S. copyright law history,^[8] who that author is has been anything but consistent.^[9] The idea of an author has shifted from the publisher,^[10] to the creative genius,^[11] and to communal authorship.^[12] With the advancement of technology, U.S. copyright law faces this question with Artificial Intelligence (AI) art.^[13] Currently, the U.S. Copyright Office does not afford AI art copyright protections.^[14] Once AI art is created, it enters the public domain.^[15]

Despite the lack of legal protections, AI art has taken off.^[16] There are three primary types of AI systems used to create AI art: (1) General Adversarial Network (GAN); (2) Convolution Neural Networks (CNN); and (3) Neural Style Transfer (NST).^[17] Generally, AI art generators use a combination of deep learning and machine learning algorithms to translate text prompts into images.^[18] The provided prompt is compared with massive datasets to determine which images correspond to the objects or aspects in the prompt.^[19] Then, the AI generator uses several algorithms to deconstruct the images it selects, reassembling them into something a human can recognize.^[20] From DALL⋅E 2^[21] to Google’s Deep Dream Generator^[22] and many others,^[23] artists now have a new method to create art.^[24] In fact, the art world has recognized that AI-generated images are art. At the time of writing, several creations have been sold at various auction houses,^[25] with one AI painting titled Portrait of Edmond Belamy, selling for $432,500.^[26] Outside of big-name auction houses, smaller artists are embracing AI art. A wealth of online communities are devoted to sharing AI art and the methods made to reach that AI art result,^[27] and many other creators are selling their AI art.^[28] Further, it seems that humans are poor evaluators of whether something was created by AI.^[29] With this widespread recognition of AI art, the U.S. Copyright Office’s exclusion of AI art from copyright protection is a failure to evolve—the U.S. Copyright Office should recognize the registrability of AI art.^[30]

Perhaps AI is just a tool used by a human? If creating AI art is a human using a tool, then this would likely be registrable based on Section 313.2 of the Copyright Compendium.^[31] Some draw a distinction between AI art generated when the AI acts as a tool for creation (meaning only part of the work is created by AI) and AI art generated almost entirely by AI (meaning the art itself is the product of a human working with an AI).^[32] There is an argument to be made, however, that AI cannot “create” art without the interaction and intervention of a human; consequently, the relationship between the two is always a human using a tool.^[33] No matter the distinction between tool or non-tool use, the authorship determination poses a barrier to registrability for AI art.^[34]

This Note argues that the end user should hold copyright in AI art. Part II of this Note lays the foundation for AI art authorship by examining the history of U.S. copyright law and case law, revealing the law’s transformative and evolving understanding of authorship. Part III follows this legal discussion with a survey of different philosophical theories of authorship, including those of John Locke, Georg Wilhelm Friedrich Hegel, and Immanuel Kant. Asking the question, what is an author, is not a new mental exercise. Philosophers have wrestled with the question and come to varying conclusions, as Part III of this Note shows. As such, this Note suggests that a melting pot of theories might best allow U.S. copyright law to reframe its definition of “author.” Part IV then overviews the different AI art author possibilities based on the views of many legal scholars—the developer, the AI itself, and the end user. This Part concludes that the end user should be the author of AI art based on practical, legal, and philosophical understandings of copyright authorship. The end user, the author, infuses personality, labor, and communication into the AI art via their creative processes.^[35] Lastly, Part V overviews some previously proposed solutions to the authorship issue, introduces a new understanding of authorship, and applies it to the Thaler^[36] case.

II. Background

A. History of U.S. Copyright

Like much of American law, copyright law originated in England with the Statute of Anne.^[37] The advent of the printing press in the fifteenth century prompted growth of printing presses, and a monopoly of the printers controlled the production along with rights of the works they printed.^[38] Under this system, printers often required authors to sign over the author’s rights of that work prior to publishing.^[39] The 1710 Statute of Anne sought to break up the monopoly that publishers and printers had on the rights to books and give those rights back to the authors who developed them.^[40] In 1787, the first U.S. copyright law was enacted via the U.S. Constitution. Known as the Copyright Clause (and also the Patent Clause), Article 1, Section 8, Clause 8 of the Constitution provides that “Congress shall have [p]ower . . . [t]o promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries.”^[41] This clause was written in the image of the Statute of Anne and defines the purpose of U.S. copyright law—to advance the arts.^[42] This idea embodied in the Statute of Anne and the U.S. Constitution—the idea that copyrights belong to the creator of the work rather than the distributor—remains an undercurrent in modern copyright policy and law.^[43]

Following the acknowledgment of copyright law via the Constitution, Congress enacted the Copyright Act of 1790,^[44] and major revisions to the Copyright Act were implemented in 1831, 1870, 1909, and 1976^[45] respectively. The United States has a history of expanding copyright protections as new technologies were developed, thereby allowing creative expression to develop along with them. The following are a few notable additions: “historical and other prints” (added in 1802)^[46]; “dramatic works” (added in 1856)^[47]; “photographs and photographic negatives” (added in 1865);^[48] “visual art” (added in 1870);^[49] “motion pictures” (added in 1912); ^[50] “sound recordings” (added in 1972);^[51] “computer programs”(added in 1980);^[52] and the Digital Millennium Copyright Act (enacted in 1998);^[53] in addition, the Supreme Court ruled in 2014 that an online streaming service may be a public performance.^[54] While arguably slow to do so, the law has changed to fit the needs of modern technology and address modern copyright issues.^[55]

Copyright registrability and copyright protections are distinct but connected.^[56] While registration is generally a requirement to bring a lawsuit for copyright infringement,^[57] a work need not be registered to have copyright protection.^[58] Copyright protections automatically attach to a work that meets the requirements for protection.^[59]

B. Three Requirements of Copyright

Generally, there are three requirements for copyright protection under U.S. law: the work must be (1) fixed in a tangible form of expression; (2) an “original work;” and (3) created by an author.^[60]

1. Fixation

Fixation is a simple requirement, defined in Section 101 of the Copyright Act.^[61] Works meet the fixation requirement when they are embodied in a medium such that the work “is sufficiently permanent or stable to permit it to be perceived, reproduced, or otherwise communicated for a period of more than transitory duration.”^[62] The method of fixation is of no moment.^[63] For example, a painting is fixed as soon as the artist puts their paint-covered brush on the canvas. However, neither a living wildflower garden^[64] nor a perishable bowl of Vietnamese food met the fixation requirement.^[65] At one point, there was debate regarding the fixed nature of computer games generated by computer code.^[66] This requirement represents the crossover from idea to expression.^[67]

2. Originality = Independent Creation + Minimal Degree of Creativity

The originality requirement “is the very ‘premise of copyright law.’”^[68] As the Supreme Court explained in Feist Publications v. Rural Telephone Service Co., “[o]riginality is a constitutional requirement” that does not consider the effort, time, or cost of creating the work.^[69] Nor is the “author’s skill, experience, or artistic judgment” considered.^[70]

There are two elements to the originality requirement.^[71] The first is independent creation.^[72] Generally, if the work is not a copy of another, it is an independent creation. This too is a low bar; the work need not be inventive or novel and can even closely resemble other works.^[73] The U.S. Copyright Office has stated that the author’s intent or inspiration behind the artwork is irrelevant.^[74] In short, the golden rule of independent creation is that one must not copy, but create one’s own work instead.

The second requirement is creativity. To satisfy this, the work must “possess some creative spark,” though, like its sister requirement, independent creation, creativity is a low barrier to entry—the Supreme Court said as much in Feist.^[75] Creativity is a minimal level of expression that is not a mere copy of another work. Creativity is found in an instruction manual,^[76] a standardized test answer sheet,^[77] and poster art advertising a traveling circus.^[78] Creativity is not found in the arrangement of a phonebook,^[79] random number generation,^[80] nor a functional chart of horse racing statistics.^[81] Nonetheless, this element is the theoretical underpinning of U.S. copyright law.^[82] In Bleistein v. Donaldson Lithographing Co., Justice Oliver Wendell Holmes described this originality requirement in terms of human creativity and personality,^[83] which suggests that Holmes saw creativity as immeasurable by any objective standard.^[84] Holmes suggested that the creative process is within everyone because each individual’s “[p]ersonality always contains something unique. It expresses its singularity even in handwriting, and a very modest grade of art has in it something irreducible, which is one man’s alone.”^[85] Judge Frank followed Justice Holmes’s argument against using the artist’s intentions to judge creativity.^[86] This sentiment rings true in copyright law today, seen especially with accidental works of art.^[87]

3. (Human) Authorship

The current dilemma for artists creating AI art is the (human) authorship requirement. As many scholars have noted, there is no explicit human authorship requirement in the U.S. Constitution,^[88] nor has Congress imposed such a requirement.^[89] While the Copyright Act mentions authorship in section 102(a), requiring that the copyright be attached to “original works of authorship,”^[90] the standard that such author be human is nowhere to be found.^[91] Instead, this is the position of the U.S. Copyright Office.^[92] The U.S. Copyright Office Compendium section 306 states that “[t]he U.S. Copyright Office will register an original work of authorship, provided that the work was created by a human being.”^[93] Why has the U.S. Copyright Office infused humanity into authorship?

The concept of authorship has been scarcely defined by law, requiring direct address by the courts.^[94] However, throughout U.S. copyright law, the notion of authorship is intertwined with that of originality. Authorship is connected to the creative mind behind the art.^[95] The Trademark Cases illuminated the definition of authorship by explaining that originality derives from the author.^[96] Originality—and so too authorship—is found where a work incorporates “the fruits of intellectual labor” that “are founded in the creative powers of the mind.”^[97] While the comingling of originality and authorship established a foundational conception of authorship alone, the first case to clearly define what an author is in the context of a “new” artistic medium (in terms of copyright law) was Burrow-Giles Lithographic Co. v. Sarony.^[98] In this case, Burrow-Giles argued that a photograph was not a writing, nor a creative product, made by an author due to the nature of the camera’s production of the photograph.^[99] Burrow-Giles argued that a photograph, which merely captures the image it is pointed at, lacks the creativity for copyright protection.^[100] The Court did not find this persuasive.^[101] Instead, the Court looked to the authorship requirement of copyright, maintaining that an author must make or originate a work that is of similar kind to those listed in the Copyright Act.^[102] A photograph was of such kind and thus was afforded copyright protection.^[103] The Court explained that photography was protectable by copyright, as the works could be traced back to the deliberate choices of the photographer, such as lighting adjustments or subject positioning.^[104] With this, copyright jurisprudence received a definition—an author constitutes “he to whom anything owes its origin; originator; maker; one who completes a work of science or literature.”^[105] So, authorship exists in the individual who caused the existence of the work.

After Burrow-Giles, the definition of “author” was further explained with the “divine author” cases, Urantia Foundation v. Maaherra^[106] and Penguin Books v. New Christian Church Full Endeavor.^[107] Both cases upheld the idea that authorship is distinctly human, or at least not divine. Urantia Foundation and Penguin Books held that the author of the divine revelation is not the heavenly being who communicated the work; rather, the human that the spirit communicated to was the author.^[108] Similarly, in Naruto v. Slater, the court held that a monkey who took a selfie was unable to hold a copyright in the photograph as the monkey is not human.^[109] It seems then that humanity need not create the work but simply take part in creating the work by integrating intellectual labor.

Nonetheless, the challenge against non-human authorship persists. One of the most recent challenges to copyright law authorship occurred in Thaler v. Perlmutter.^[110] While the case is in the D.C. federal district court, this issue is highly relevant and likely to reach the Supreme Court at some point, even if not with this case specifically. In this case, Thaler challenged the U.S. Copyright Office’s denial of his application for copyright registration of his work A Recent Entrance to Paradise.^[111] As Thaler’s original complaint states, Thaler created this AI art using his AI system, which he refers to as a “Creativity Machine.”^[112] Thaler states that through the work-for-hire doctrine, he is entitled to hold copyright in the AI artwork.^[113] At the time this Note was first drafted, the outcome of this case was unknown. Nonetheless, it seemed likely that the district court would continue the tradition of prohibiting nonhuman authors from holding copyrights—and the court did.^[114] So, this leaves a gap in copyright jurisprudence; the term “author” remains a loose concept but is constricted enough to prohibit the advancement of the arts by failing to permit the registrability of AI art.

Just as the United States looked to England for reference to the Statue of Anne as a basis for the Copyright Clause,^[115] perhaps the United States should look to other nations to determine the solution to this issue. In fact, many other nations recognize the copyrightability of AI art in some form, including the United Kingdom,^[116] India,^[117] New Zealand,^[118] Hong Kong (SAR),^[119] and Ireland.^[120] However, unlike these nations, it does not seem that the United States is ready to accept AI art in a statutory sense. Instead, expanding the current understanding of the authorship requirement stated in section 306 of U.S. Copyright Office Compendium would provide a baseline to integrate AI art authorship into U.S. copyright laws.

Assuming this infusion of humanity is proper, human authorship should not preclude, as it currently does, the copyright protections and registrability of AI art. Arguably, AI art is not all machine.^[121] Like recognized artists such as Jackson Pollock,^[122] humans who are end users creating AI art have the ideas and exercise control over the output of those ideas. The copyright author is not merely the mechanical executor of the work; instead the author contemplates the work and controls the execution of it.^[123] Further, when a machine is involved, such as a camera, the machine assists in the creation of work, such that the created work would not exists without it.^[124] This Section discussed what an author is not through an examination of the authorship requirements and the nonhuman authors who failed to meet them. In the following Section, this Note considers what an author is through the lens of philosophical theories of authorship.

III. What Is an “Author”?

Interestingly, authorship seems to be a modern problem that came along with intellectual property rights.^[125] Perhaps modern problems require historical solutions. The question of what an author is has been asked by many, including great philosophical minds.^[126] Leaning on philosophy is not a new idea; philosophy is intertwined with the law, from criminal theories of justice to what constitutes a living human being. Using philosophical understandings of authorship, then, is not so out of place. Integrating philosophy into this discussion continues a tradition of looking to disciplines outside of the law, such as philosophy, to examine and define the question of authorship. ^[127]

Copyright law seems to have taken the approach that the intent of the author is of no moment to the question of copyrightability.^[128] If that is so, why pay mind to who (or what) the author is? The law seems to embrace, at least in part, the dismissal of the author (similar to what Barthes and Foucault do with their theories of authorship). The only requirement is that a human author exist (and maybe also create the work).^[129] Barthes and Foucault suggest that the reader/viewer of the work analyze and understand the work absent from what the author of that work intended and absent from the author altogether.^[130] Barthes argues that (1) authorship is a social construct created by the work’s reader/viewer;^[131] (2) the meaning of a work is never fixed and instead is created by the reader/viewer of the work through their reading/viewing;^[132] and (3) the author’s person and intentions have no relevance to the meaning of the text.^[133] Hence, the author is dead and only real via a social construct.^[134] Similarly, Foucault argues that authorship is continuously evolving.^[135] Distinctly, however, the individual viewer does not create the “author”; rather the “author” is created by the power relations of society.^[136] While interesting, these views have not gained much traction in U.S. copyright theory. However, the works of John Locke, a British philosopher^[137] who greatly influenced the founders of the United States,^[138] and Immanuel Kant and Hegel are recognized for their theories of property.^[139]

A. John Locke—Labor

Locke’s property theory argues laborers have rights to the fruits they produce because of their labor.^[140] While there are many different interpretations of Lockean property theory,^[141] each interpretation relies on this foundational idea of labor.^[142] The Supreme Court seems to have moved away from a Lockean labor theory of copyright, stating “that[,] originality, not ‘sweat of the brow,’ is the touchstone of copyright protection.”^[143] Copyright cares about the product, not the process nor the labor put into the product.^[144] Of course, this is only in terms of the idea of originality. The question of authorship is a question of who will own the copyright in the work, an idea more in line with traditional theories of property, like Locke’s.^[145] Scholars often mention a “workmanship interpretation” of Lockean property theory.^[146] This interpretation argues that property rights are the laborer’s because they created the property.^[147] In order to own the copyright, one must create the work—that is, put the labor into creating it. However, the creationist interpretation of Lockean labor fails to properly address the reality of creation itself. The key foundation of the creationist theory is that the creator’s labor as an “ex nihilo activity that does not depend on what comes before it” based on the separate (meaning entirely separate from all other’s ideas) ideas of the individual.^[148] No idea is truly without contribution from outside the individual’s own mind. Creators are influenced by the work others have done, finding new ways to build on it—hence the existence of defenses to copyright infringement claims, such as fair use^[149] and independent creation.^[150]

Outside of the workmanship model or creation interpretation, the intellectualist theory is connected to “personhood” instead of physical labor.^[151] Under this interpretation, there is a mix of labor with the creation such that the author is one who put labor into intertwining their person with the work.^[152] Thus, so long as another’s right to self-government is not violated in the creation of a work, the labor of an individual cannot be separated from the work without thereby violating the individual’s right to self-govern and diminishing their labor.^[153]

Legal scholars have recognized the issues created by the law’s use of a romantic notion of authorship,^[154] suggesting a partition between the romantic notion of authorship supported by a creationist theory of Lockean authorship and copyright authorship.^[155] Unlike a romantic author, Locke’s intellectualist labor theory looks to the author’s judgment—their “intellectual input” as shown by their choices, instead of solely to the creativity of the work.^[156] This not only conforms to the ideas of authorship suggested in cases such as Burrow-Giles, but also provides concrete inquiries into authorship for AI art purposes.

B. Kant and Hegel—Personality and Communication

To Immanuel Kant, communication is an act—a speech by the individual rather than the encapsulation of an author’s ideas.^[157] Kant saw the artist as “governed by [the artist’s] choice[s].”^[158] The author’s art then is not a physical thing; it is a “communication with the public” made by the author.^[159] Thus, integral to this idea is the autonomy of the author.^[160] Kantian authorship then broadens copyright’s understanding of authorship beyond the creator’s role in the work’s production and beyond the fixed work itself.^[161] While Kant famously denied the existence of intellectual property,^[162] he acknowledged authorial property rights in an author’s thoughts.^[163] As such, Kant’s theory of intellectual property contrasts directly with Lockean property theories, which focus on the physical labor of the author. For example, Kant discussed the unlawful reprinting of books.^[164] His argument is based on morality—an author’s property is not violated by a reproduction of the work; instead, unauthorized reproduction of an author’s work violates the author’s right to dictate the dissemination of his work’s ideas in connection with the true source of those ideas.^[165] By disseminating the work “in the name of another person” the moral rights of the true author are violated, as the personality of the author has been violated.^[166] Kant’s rights attach to the destiny of the work to protect the author’s personality, not the work itself.^[167] In terms of AI art, no matter if one views the developer, the AI, or the end user as the true author,^[168] the prohibiting registration of AI art perpetuates the violation of the author’s personality. By entering the public domain, AI art authors face unauthorized reprints of their work—a violation of the moral rights of the author.^[169]

Similar to Kant, Hegel’s approach to authorship is a personality theory of authorship.^[170] To Hegel, an individual’s personality is inalienable and through “acts of expression,” one’s inner personality would be transformed and fixed in a physical alienable form.^[171] This transformation of intellect and being into a product is essential to the development of one’s personality and is to be built on by the common pool of thinking.^[172] Hegel does not reject the idea of private intellectual property unlike Kant.^[173] Instead, an individual’s will drives property; so, property rights allow for freedom to attain individual satisfaction though a development of individual personality.^[174] Indeed, the law captures this idea.^[175] Hegel suggests that protecting an author’s work is the “best way of progressing science and arts,”^[176] as the purpose of intellectual property is to facilitate the recognition of an author’s property by others.^[177] Similar to Locke, Hegel values the labor put into the author’s work; for Hegel, it is “realisation of spirit.”^[178]

In combination with a denial of moral rights under Kant, the denial of property rights in AI art is the denial of personality development based on the ability and desire of an author to create. Interestingly, in the case of “accidental authorship,” an author’s adoption of the accidental effect constitutes an act of will.^[179] In AI art, the conscious choices of an author represent the author’s will, spirit, and personality to which moral, labor, and personality rights should be attached. The failure to recognize these rights inhibits the growth of copyright by disincentivizing the creation of new works. Looking at these three views together, authorship is composed of an individual’s communication through the choices the individual makes in the creation of the work, and that product should be protected from immoral modifications. AI authorship, then, consists of labor, personality, and communication.

IV. The Three Possible Authors of AI Art

According to a wealth of legal scholarship, there are three possible “authors” of AI art: (1) the AI’s developer; (2) the AI itself; and (3) the AI’s end user.^[180] Academics have also introduced the idea of joint authorship^[181] or amending the work for hire statute.^[182] However, authorship is the current focus of the Copyright Office^[183] and the courts,^[184] not a revamp of the statute and case law. While such a revamp is certainly a solution, this may be more difficult than expanding the law’s understanding of authorship. Further, it may not fully encompass the needs of the future, just as the current statute does not respond to current AI art.

A. The Developer as “Author”

The primary foundation for the idea that the developer of the AI is an author of the AI art is the “work made for hire” doctrine.^[185] However, this would call for the expansion of the meaning of “employment,” which the Supreme Court has declined to do.^[186] Nonetheless, a tenet of this view is that developers infuse control and creativity by designing AI to think, and do so creatively, by controlling its inputs.^[187] Further, developers of AI are already afforded legal protections; they are eligible for copyright protections on the AI software.^[188] Affording them copyright protection of the end user’s artistic works created with the developer’s AI poses a “double dipping” concern.^[189] While developers might be seen as the creative geniuses behind the AI, that is not authorship of the AI art itself. Treating the developers as authors of AI art—when the developers themselves are not end users of the AI—does little to advance copyright incentives. Similarly, providing developers the authorial rights of AI art results in a weakened incentive for end users to utilize AI and further the introduction of new ideas and works into the public.^[190]

B. The AI as “Author”

Some have argued for vesting authorship in the AI itself. Ryan Abbott takes this view in the context of an inventor in patent law while analogizing to the issue of authorship in copyright law.^[191] Abbott argues that AI should be considered the legal inventor because the human instructing the AI does not significantly contribute to the AI’s innovative process.^[192] However, he notes that “[c]reative computers invent because they are instructed to invent.”^[193] In the context of patent law, where human expression is not at issue,^[194] AI authors might make some sense. Patents protect the invention which is mechanical in nature and void of human expression.^[195] Contrast this with copyright;^[196] copyright protects the human expression in the art.^[197] The AI itself does not communicate, create, or express, unlike the end user’s role in AI art. Further, as a non-human entity, AI is incapable of qualifying as an author—just as the monkey in Naruto.^[198] It seems clear that AI should not be the legal author of copyright held in AI art.

C. The End User as “Author”

Extending copyright to AI art held by the end user is the most practical solution.^[199] Further, due to the unique relationship between the end user and the developer, the developer depends on the end user’s interest in the developer’s work.^[200] By furthering the copyrights of end users in AI art, the developers’ copyrights in AI software are also furthered—avoiding the “double dipping” concerns and promoting the purpose of copyright.^[201] Additionally, although the creative process of the artist on the one hand is irrelevant to the question of copyrightability in the originality sense,^[202] it can illuminate the authorship requirement of end users. Authorship is inherently creative, and to exclude the creativity of a person from the authorship of work is to exclude the author from the work, thereby undermining the idea of copyright.

AI authorship of AI works necessitates thinking about causation.^[203] The cause of AI work, in a but-for manner, suggests that the end user is an author of AI art.^[204] In this regard, even famed non-AI artist Jackson Pollock’s artwork is a challenge to authorship.^[205] If human touch must be explicitly involved throughout the creation of AI art, why not view Pollock’s work as ineligible for copyright? Continuing with this thinking, perhaps the paint made the art, not Pollock—Pollock just put the paint into motion. Or conversely, the only reason why the art was made was because of Jackson Pollock’s (a human) contribution—choosing the paint and splattering it.

This concept was at play in Burrow-Giles; the actions of the individual behind the camera resulted in a work of authorship.^[206] The camera, invented in 1816,^[207] was a new technology used well before the famous Burrow-Giles case that solidified photography’s place in U.S. copyright law.^[208] Had the Court not recognized the person behind the camera as the author, a similar debate to that of AI authors might have taken place. The main arguments against authorship of photography were that a photo is not creative and the camera itself, not the man behind the camera, created the image.^[209] Drawing from arguments for the developer or AI to be named authors of AI art,^[210] for every photo this camera captures, would then the camera be the author of the work or the camera’s manufacturer? No, this seems absurd; the author is the camera’s user. The user fixed the idea of capturing the image by pressing a button on the camera, by determining which of the many photos the camera took best illustrates the intended result, and by tweaking the inputs to the camera.^[211] It is the same with AI. The end user chooses the words to inspire the painting (choosing the paint or photograph’s subject) and then indicates to the AI that it should process that selection (the splattering or snapping of an image).^[212] Within this process of choosing the prompt, the end user makes a host of decisions.^[213] The issue is that the human is less involved with the actuation (the splattering) than Pollock. Pollock made choices such as how fast to move, how much paint and so forth.^[214] Here, arguably, the AI makes these decisions. However, this does not mean there is a lack of creativity from the end user. The standard is a “modicum of creativity”—not as much creativity as Vincent Van Gogh.^[215] The creation of AI art is not devoid of human touch, similar to the creation of a photograph. Further, the creation of the art is not finalized yet. The end user may not find that the AI images capture what they were going for—here there are choices.^[216] Such choices throughout the process of creating AI art speak to the Lockean autonomy, the infusion of personality under Hegel, and the Kantian communicative act.^[217]

The AI is the mechanical creator of the work, whereas the end user is the intellectual creator of the work. AI art presents an infusion of creativity by the end user author akin to Hegel and Locke’s theories—the author is the end user because they fused their creativity with the final outcome. Thus, when the end user develops a string of words to enter as a text prompt, sifts through the outputs and rests on the output that captures their creative idea, they have labored. Of course, the amount of labor is not the metric—just the fact that a human has labored to create it.^[218] It is not as if the AI is creating this art on its own; there is human interaction (labor). With this creativity comes personality. If the human interacting with the AI is not the author, the end user has no agency, thus no autonomy.

V. Solution

A. Possible and Proposed Solutions

Many solutions have been proposed to the copyright problems raised by AI art, with some of the more prominent including joint authorship^[219] and modifications of the work-for-hire doctrine.^[220] One might even briefly consider affording copyright protections to the prompt an end user employs to create the AI art; however, this is likely the attempted copyright of the idea and not the expression.^[221] Similar to the above solutions, in that they attempt to work with the pre-existing structure of U.S. copyright law, a redefinition of “author” does the same by stretching the law’s understanding of “author” to encompass the reality of the AI art landscape.^[222] Of course, any solution in this area would likely implicate patent law, as many others have written on.^[223]

B. A New Solution—AI Art Authorship

Broadening the term “author” based on ideas of labor, personality, and communication expands on the but-for causation authorship suggested by the Burrow-Giles court.^[224] It adds a “proximate cause” element of authorship based on expanded ideas of who (or what) is an “author.” The formula for authorship is a combination of labor, personality, and communication.

This formula is an evolution, as opposed to the death,^[225] of the copyright author. It solves the AI author issue and demonstrates that AI art is worthy of copyright registration. Authorship is “a culturally, politically, economically, and socially constructed category rather than a real or natural one.”^[226] If AI art is socially, economically, and culturally accepted, so too should its authorship. Evidence of this creation appears through case law and statutes. Denicola interprets section 201(b) of the Copyright Act to mean that works made for hire afford authorship to the employer “who may have played no role at all in the actual creation of the [art] work.”^[227] Further, in Goldstein v. California the Court stated that “in its constitutional sense, [author] has been construed to mean an ‘originator,’ ‘he to whom anything owes its origin.’”^[228] In Urantia Foundation v. Maaherra, “some element of human creativity must have occurred in order for the Book to be copyrightable.”^[229] But unlike the cases of Urantia Foundation and Penguin Books, in the case of AI art, the divine being is the human, not the machine.^[230] The end user communicates to the AI and through that communication, labor, and infusion of personality creates a work.

Harkening back to Justice Holmes, creativity is subjective.^[231] Judging what is considered art involves often making a value judgment about the amount of creativity put into a work.^[232] The more creativity (and usually the more effort), the more likely the creation is called “art.” However, judges, lawmakers, and the like are not art critics. Justice Holmes recognized this.^[233] The law should not allow judges and lawmakers to deem certain arts valid and other arts invalid by way of allowing copyright registration to some but not others. Some believe that AI art is not “art” because a phrase or paragraph of text prompt inputted into AI art generators leads to a visually interesting image to appear on the user’s screen.^[234] The law seems to agree with this view—the only creator is the AI itself developing the image, and so it is the author.^[235] But an AI author cannot hold copyright;^[236] thus, AI art is ineligible for copyright registration.^[237] However, the end user is the creative and so, the end user is the author. The end user, as the author of AI art, should be given copyright protection over the AI art which the end user creates. This expansion of copyright authorship rides on copyright law’s eventual evolution to cover new subject matters, which is premised on “protecting all forms of original authorship.”^[238]

Outside of the academic thought exercises and the legislative rulings, there is an uproar from some artists that their work is being used to train AI models.^[239] By using their art in the training data set, the AI can later produce AI art in their style.^[240] These artists suggest that such use of their copyrighted work takes away the profitability of their work.^[241] By extending copyright protections to art generated by AI, AI art is made more ethical. Developers would potentially be more conscientious of the images they use to train the AI, thereby restoring the creative integrity and market share that these artists claim is taken from them. Further, the new art would arguably have to be more creative and push the boundaries of art, which “promote[s] the [p]rogress of . . . [the] useful [a]rts.”^[242] The continued release of AI art images into the public domain results in those images being “recycled” into the AI, thereby producing works of a lesser creative degree. While copyright law purports not to care about the creativity of the art, the overall policy goal of copyright law always has.^[243] Still, the public domain serves an important purpose.^[244] While AI art authors should be afforded copyright protection and registrability, providing among many rights, the right to exclude, this exclusion must not last forever. Instead, AI art author’s right of exclusivity should be only a limited period of protection. This period of protection must be enough to advance the goals of the Copyright Clause and offer authors protection but not so long as to diminish the use of the public domain.^[245]

1. Application to the Thaler Case

As explained above, the Thaler case is one of the most recent developments in the issue of AI art and copyright.^[246] In this case, the developer and the end user are the same human being.^[247] Here, it is easy to see that Thaler, as both the developer and the end user, is the author of his AI artwork A Recent Entrance to Paradise.^[248] If the facts are changed slightly, where Thaler no longer simultaneously plays the role of developer and end user but instead, he is only an end user and does not own the AI he used to create the artwork—is Thaler still an author? Based on the reformulated understanding of authorship, yes. Had Thaler never inputted a text prompt into the AI art generator, A Recent Entrance to Paradise^[249] would not have existed. Thaler likely had to develop and tweak the inputs, like the photographer in Burrow-Giles, which meets the labor element under a Lockean intellectual labor theory.^[250] Regarding personality, according to Hegel, it is woven into the work.^[251] By suggesting prompts, altering them, and examining the AI outputs, Thaler makes the choice and exercises the control indicative of personality. Indeed, Thaler likely landed on the output that best captured what he hoped to capture. The artistic process similarly fulfills the Kantian communication element—these acts and the images produced by them are a speech from Thaler, which Thaler has a right to control. However, Thaler’s protection would only last a limited number of years, as previously suggested.

It seems the U.S. Copyright Office and Thaler agree that A Recent Entrance to Paradise satisfies the copyright requirements of fixation and originality.^[252] Perhaps this revamped authorial understanding will allow the parties and the court to respect the foundations of copyright while recognizing the much-needed advancement of it to match the technological and social changes.

VI. Conclusion

René Descartes argued that machines are unable to communicate as humans do because a machine is incapable of putting “words in different orders to correspond to the meaning of things said in its presence, as even the most dull-witted of men can do.”^[253] As any user of ChatGPT can tell you, this is false.^[254] Technology is evolving and providing humans different ways to communicate, create, and express. Limited methods of creativity at the time of writing the statutes need not limit available copyright protections, as the Court suggested in Burrow-Giles.^[255] This stifles creativity and is against the very thing copyright seeks to promote.

Clearly, authorship exists in AI artworks. The developer already has protections and does not have a stake in the copyrightability of the artwork unless they create the work with the AI themselves. As a non-human entity, AI cannot hold the copyright. The author of AI artwork is the end user who sets the AI art’s existence into motion, like a Pollock with a paintbrush or a photographer behind a camera. Infused in the end user’s authorship are the personality, labor, and communicative nature of the author, which deserve legal protection. This legal protection in the form of copyright registration may reflect the nature of the artwork by limiting the exclusivity rights of the author to a shorter term of years than traditional copyrights.

No matter the method of solving this issue, the law should not shun technological advances in art because they do not fit into the current structure. Rather, U.S. law should adapt once more and allow AI art to receive copyright protections and register. How can the law “promote the [p]rogress of . . . [the] Arts”^[256] if it does not protect advancement in the arts?

Mackenzie Caldwell