By: Marisa Cefola
Volume X – Issue I – Fall 2024
I. INTRODUCTIONS TO ARTIFICIAL INTELLIGENCE AND COPYRIGHT
There are two primary types of artificial intelligence (AI) prevalent in everyday life. The first is traditional AI, which uses “if-then” programming to guide a system’s decision-making process. For example, a traditional AI system recommends medical diagnoses to doctors after being provided with patients’ reported symptoms (e.g., if a patient presents a certain symptom, then AI may perceive that symptom to be part of a certain disease). In contrast, generative AI (GAI) systems are trained using large data sets and are designed to generate creative content such as artwork or writing. By identifying patterns in the data, these systems can replicate and produce various creative works with a process known as machine learning. [1] This process makes GAI a Large Language Model (LLM), which can comprehend and generate natural, human-like language that allows it to perform a wide array of tasks. [2] The launch of Open AI’s ChatGPT on November 30, 2022, propelled GAI and LLMs into the mainstream, gaining 100 million users within two months on the market. In turn, these technologies have surged in popularity across corporations and entire industries due to their rapid pace of innovation. GAI and its future potential raise ethical and intellectual property concerns in these industries, however. [3] For instance, GAI was part of a central dispute in the 2023 Writers Guild of America strike, as writers demanded more regulations surrounding the systems. Hollywood producers used machine learning to produce scripts, using writers’ work as part of training data sets (in which writers were not compensated for these efforts) as a cheap alternative to human labor. The writers were ultimately successful in prohibiting AI from being considered an original work in the industry. [4] AI being at the forefront of the writer’s strike demonstrates its growing prominence in society, which has brought up new questions regarding AI’s intersection with United States copyright law.
According to the United States Copyright Office, original works “are independently created by a human author and have a minimal degree of creativity.” [5] Owning a copyright for an original work provides the exclusive rights to make, distribute, and sell copies of creative works. [6] This allows creators to control their works’ value and generate profit through exclusivity, which stimulates an incentive to innovate in both the sciences and the arts. Stephen Thaler is one such innovator. Thaler developed a computer system—which he called the “Creativity Machine”—with the ability to generate artwork analogous to a human-created piece of art. The Creativity Machine uses GAI capabilities to do this, which include learning patterns of existing human work data sets and generating work with similar patterns based on a user’s text prompt. [7] Thaler sought to obtain a copyright for the artwork itself and was denied by the United States Copyright Office on the grounds that a human author did not independently create the work.
This paper explores Thaler v. Perlmutter, a case with the potential to uproot copyright laws and greatly alter innovation in the sciences and arts. Thaler is currently being argued in the U.S. Court of Appeals for the District of Columbia and may eventually reach the Supreme Court. Based on the facts of the case, current regulations, and arguments, this piece will predict the case's outcome and the applicable precedents it may set.
II. A HISTORY AND OVERVIEW OF THALER V. PERLMUTTER
Thaler originally filed a lawsuit against Shira Perlmutter, Register of Copyrights and Director of the U.S. Copyright Office, soon after his copyright application for an art piece titled, “A Recent Entrance to Paradise,” generated by the Creativity Machine, was denied in 2019. He claimed that the AI algorithm behind the Creativity Machine authored the artworks, yet asserted he should be the true copyright owner on the basis that he owned the machine and developed the algorithm. This refers to a “work-for-hire,” which awards copyright to a third party rather than the creator. For example, if a corporation hires an artist to design a logo, they would be granted a copyright instead of the artist. [8] This was Thaler’s main argument, but he notably admitted that “A Recent Entrance to Paradise” was completely made autonomously, which commentators later address. The case’s legal issue asks whether or not work created solely by an AI system could be copyrightable based on the Copyright Office’s definition of an “author.” [9] Ultimately, the question before the court was whether a work-for-hire is eligible for copyright protections when the true author is a GAI algorithm.
i. The U.S. District Court for the District of Columbia
After the initial copyright application was denied, appealed, and rejected again, Thaler was brought to the United States District Court for the District of Columbia. In court, Thaler argued the rejections violated the Administrative Procedure Act (APA) and that the court should be required to invalidate the government action. The Administrative Procedure Act was created to combat the abuse of administrative power and ensure that administered regulations were fair and reasonable. [10] It condemns any act from a government agency that is “arbitrary, capricious, an abuse of discretion, or otherwise not per the law. [11] Therefore, Thaler claims that the USCO decided to reject the copyright in question in an unlawful, non-standardized manner and that it is now up to the court to void the decision since they have exhausted the USCO appeals process. According to the APA, the defendants, namely Perlmutter and the USCO, denied this claim but were burdened with defending their reasoning behind rejecting Thaler’s copyright application. The court can only review an agency’s actions based on the reasons it gave for the action, so a strong defense is imperative for Perlmutter.
Both parties additionally filed a motion for a summary judgment, which a court should grant if “there is no genuine issue of material fact” between the parties. [12] A summary judgment motion is beneficial to file, as it can avoid a jury trial by affirming that there is no factual dispute between the parties that would require a trial. For Thaler, if summary judgment is granted in his favor, it would open up the potential for AI-generated art to be eligible for copyright. On the other hand, if the court grants summary judgment for Perlmutter et al., it would affirm the USCO’s decision to deny copyright applications relating to AI-generated art. The district court decided Thaler in August 2023. In their discussion, the court noted the only genuine legal question is whether the USCO acted arbitrarily or capriciously in making their decision. [13] It declared the copyright in question is not protected by law due to being created by Thaler’s autonomous system rather than Thaler himself; therefore, the USCO did not err in their judgment and acted properly in making their decision, with no evidence to prove otherwise. As the court states, Thaler complicated the case by basing much of his argument on legal theories to connect the artwork to him as an owner such as the work-for-hire doctrine. However, these cannot be considered unless it is determined there is a valid copyright with human authorship. In his argument, Thaler notes the flexibility of copyright and how it is designed to adapt to new technology. He cites Burrow-Giles Lithographic Company v. Sarony (1884), a case the Supreme Court decided that extended valid copyrights to photography. [14] However, one consistent element is the presence of human creativity in these adaptations; the court believes there is no inherent human control or creativity present in the Creativity Machine compared to photography, where a photographer maneuvers a camera and subjects to capture a desired shot. One of the necessary components of copyright is human authorship and without it, a work is not sufficient to be registered.
Because of this, the court denied the plaintiff’s motion for summary judgment and granted the defendant’s cross-motion for summary judgment.
ii. The U.S. Court of Appeals for the D.C. Circuit
Thaler appealed the district court’s decision to the U.S. Court of Appeals for the D.C. Circuit, which is pending a decision. [15] In his reply brief, Thaler raised numerous flaws in the USCO’s reasoning for denying copyright protection to a non-human work. He stated that registrations for works authored by corporations or the U.S. Government are not required to attribute it to any human contribution or human author. According to Thaler, AI algorithms are analogous to this and should receive similar treatment. Thaler also raised alternate theories in his initial copyright application such as the work-for-hire theory and a presumption that he is the only plausible author because no other human could claim ownership. If the Creativity Machine authored the work, Thaler is the “undisputed human originator of the work.” [16] The USCO did not acknowledge these theories, to which Thaler alleges them to be acting arbitrarily and capriciously in a way that violates the APA.
Oral arguments were conducted on September 19, 2024, and, like the district court judge, the three-judge panel remained dubious of Thaler’s theories. Circuit Judge Patricia A. Millett and the district court noted that in APA cases, both courts act as appellate courts. This means they can only evaluate the USCO’s reasoning for denying the copyright and any complaints brought up in Thaler’s appeal filings. Thaler cannot introduce new information if it is not included in his briefs unless there are extraordinary circumstances, which are not present here. The lower court ruled that Thaler had forfeited his right to use these arguments in court because he failed to address them in his district court filing. Millet recognized that Thaler did not challenge this ruling, and, as such, Thaler’s case seemed to have been gutted before judges could consider the issue at hand.
With regards to Thaler’s theory that he owned “A Recent Entrance to Paradise” because he created the AI that generated it, the court separated ownership and authorship. In the oral arguments, Thaler argued that possession is a basis for ownership and Thaler owned the image by being the first to possess it. On the respondent side, Perlmutter maintained that human elements were necessary for authorship on the basis that humans can own property while machines cannot. [17] Perlmutter raised concerns about the implications of the ruling in favor of Thaler, suggesting that Congress would need to revise intellectual property laws in a way the court alone cannot. The nuances of AI go beyond a single case or court ruling, as Perlmutter states.
Intellectual property lawyers and case commentators such as Paul Llewellyn of Arnold & Porter Kaye Scholer LLP believe the court is seemingly avoiding the ambiguity of AI copyright. Llewellyn argues that the court is highlighting procedural flaws while neglecting AI’s rapid growth and emergence in the creative world. If the appeals court rules in favor of the USCO, it would leave questions regarding the degree of creativity required for AI copyright protection. Llewellyn states, “The appellate court seems to be reluctant to reach beyond the record in the case and address some really interesting and intriguing questions that weren’t fleshed out [at the district court].” [18] Attorneys and case reporters generally believe the court will rule in favor of the USCO, but the ruling is still to be determined as of November 19, 2024.
III. RELEVANT INTELLECTUAL PROPERTY AND ARTIFICIAL INTELLIGENCE LAW
Cases such as Thaler highlight potential gaps in copyright law, which has not yet fully caught up to advancements in AI technology. However, these cases also highlight the original purpose of copyright law and why legislation has not yet evolved to reflect changes in the creative market. A foundation of Thaler’s argument is that the Copyright Act of 1976 does not address the requirements of being an “author” nor does it state an author must be human. This is accurate, but 17 U.S. Code § 101 somewhat addresses authors. It states that for a work to be considered “fixed in a tangible medium of expression” — which is crucial for obtaining a copyright—it must be expressed “by or under the authority of the author.” [19] Though it addresses whether the author must be in control of producing the work, an AI could “control” what it generates based on machine learning and user prompts. Therefore, much of this aspect of the law is up to the courts’ interpretations. Historically, courts explicitly constrict who or what is an “author” to humans. Naruto v. Slater (2018) is one example case that reached the U.S. Court of Appeals for the 9th Circuit. It discusses the legalities of Naruto, a monkey, being subject to copyright authorship for several photographs it took of itself with wildlife photographer David Slater’s camera. Naruto’s pictures fit the definition expressed in 17 U.S. Code § 101, as it was in control of the camera and the work was expressed “by or under the authority of the author.” However, the 9th Circuit Court of Appeals ruled that Naruto lacked statutory standing to sue Slater because The Copyright Act does not permit animals from holding authorship. [20] There was no evidence that Naruto was economically harmed if Slater used or communicated the photographs by any means.
The Thaler court was no different. The Oxford English Dictionary and Merriam-Webster Unabridged Dictionary influenced the district court’s definition of author. In their opinion, they stated that an author refers to “one that is the source of some form of intellectual or creative work” and “[t]he creator of an artistic work; a painter, photographer, filmmaker, etc.” [21] Additionally, an author must possess the “capacity for intellectual, creative, or artistic labor”. Considering this and the purpose of the Copyright Act of 1976, the court said an originator must be human. [22] From its inception, copyright laws have aimed to foster innovation in the fields of science and the arts. The United States Constitution Article I section 8, clause 8 states, “The Congress shall have Power ... To 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.” [23] To do so, copyright laws provide an incentive to those who create by rewarding owners with certain property rights. 17 U.S. Code § 106 outlines the exclusive rights copyright holders are entitled to, which include the right to: reproduce the work, create derivative works, distribute copies of the work, and publicly perform or display the work if applicable. [24] Framers drafted these rights with the prediction that authors would want to protect their works against theft and receive some sort of compensation for their creations. To do so, authors would spread their work to the public, which can stimulate the creative economy, promote public welfare, and encourage more transformation in the useful sciences and arts. [25] Many courts such as the courts in Naruto and Thaler likely believe that non-human creators—namely animals and artificial intelligence algorithms—do not respond to incentives in a way that upholds the purpose of copyrights. In other words, these authors are unaware of or indifferent to monetary rewards for innovating. This relationship between the laws and authors seems to be a large reason why courts are hesitant to grant exclusive rights of copyright to a work created entirely by an AI.
Thaler’s work-for-hire theory is intriguing, as it reduces this tension by directing the creator’s incentive toward Thaler himself. If stimulating the useful sciences and arts is the ultimate goal of copyright law, Thaler seems to be achieving both. His algorithm, an innovative technology, is creating unique artwork that arguably promotes public welfare, as seen in “A Recent Entrance to Paradise.” There is also an incentive for other humans to develop algorithms and improve current technology. However, the key factor to consider is the novelty of GAI, particularly how it has not yet exhausted its potential uses. By giving Thaler exclusive rights for the AI-generated art, it limits any creator from creating any art similar to “A Recent Entrance to Paradise,” which establishes a monopoly for Thaler and may disincentivize authors from exploring GAI’s potential uses; rapid research that is happening right now on the subject may plateau. It is important to consider how easy it may be to copy Thaler’s work since it was likely generated through ordinary text prompts. Authors could replicate the art with a different text prompt, which evokes questions about whether the text prompt matters when determining potential copyright infringement. Though the work-for-hire theory seems sound, GAI’s nuances are still raising more questions than answers.
This article has not addressed a critical aspect of AI thus far: AI has existed and been prevalent for decades- why only in the past few years has it started to generate legal concerns? Over the past decade, society has seen rapid advancements in LLMs and is becoming aware of them with ChatGPT now established as a household name. GAI is integrating into entire industries such as entertainment and healthcare, which was never considered a decade ago despite these models existing in systems such as Apple’s Siri. [26] Economic research predicts that AI will grow at an annual rate of 36.6 percent over the next six years, and by 2030, AI representation in the United States GDP will increase by an estimated 21 percent on net. [27] This establishment of AI as a major player in the country’s economy and universal industries like healthcare proves the contemporary ChatGPT era to be the Gilded Age of AI. Though society is rapidly gaining wealth and prosperity through technological and creative innovation, researchers worry this will come with a widening wealth gap and economic inequality that favors large corporations and the wealthiest individuals. Job displacements, concentration of power in a few AI companies, and dependence on AI are all risks noted to be aware of. [28] Great economic and technological change is often accompanied by political and legal unrest, [29] so it is clear that history is repeating itself in the age of AI.
IV. CASE OUTCOME PREDICTIONS AND IMPLICATIONS
Thaler has the potential to be a monumental case in the world of AI-generated art. However, I do not believe courts are ready to rule in Thaler's favor at this time, as they remain steadfast in the view that copyright laws are designed to favor human authorship. The U.S. Court of Appeals for the D.C. Circuit will likely rule in favor of Perlmutter, expressing that the USCO was right in denying Thaler’s copyright application for “A Recent Entrance to Paradise.” Thaler will almost certainly appeal due to the procedural flaws the appellate court pointed out. I believe Thaler shares a sentiment similar to that of oral argument commentators and will argue that the judges did not flesh out the issue at hand. He may petition for a writ of certiorari to the Supreme Court; if four Supreme Court judges grant the writ, they agree to hear the case and Thaler will be heard before the Supreme Court. [30] It is uncertain whether the Supreme Court will decide to take the case if it becomes applicable. On one hand, they may want to address the question of whether AI-generated art should be subject to copyright and clear up any legal ambiguities for the time being. However, they may agree with the U.S. District Court for the District of Columbia’s granted summary judgment ruling for Perlmutter and assert the USCO did not err in its judgment without the need for a trial. If the case is argued in the Supreme Court, justices will likely reach similar conclusions to the district court.
By granting a copyright to Thaler, the implications would be immense. Thaler’s copyright has the potential to encourage corporate monopolies to form, which would discourage technological or creative growth. Due to AI’s current and eventual impact on the economy, it is risky to grant copyrights, as it sets a precedent that allows large conglomerates to seek out copyrights for works ordinary text prompts generate. These companies can preemptively create a content farm by generating a large amount of art and applying for copyrights with their vast resources. This could result in a concentration of power among major corporations, effectively rendering compensation for creativity improbable if not impossible. Many text prompts would be in the control of a few entities, making infringement easy and common. In turn, artists may be discouraged from exploring AI-generated art due to the potential for using similar prompts and being accused of infringement as a result. Human-created art that resembles AI images such as “A Recent Entry to Paradise” could also be subject to infringement, so allowing copyrights to works with an AI author may put human artists out of work. However, suppose the court chooses to rule in favor of Perlmutter. In that case, it continues to leave questions of how much human input is necessary to allow AI-generated works to be entitled to copyright and if the text prompt is considered part of the copyright. Copyright laws allow individuals to replicate ideas but not the expression of those ideas. In other words, an ordinary prompt is an idea, but the work based on the prompt is an expression of the idea. It will be interesting to monitor the extent to which text will play a role in future AI copyright disputes.
Until Thaler is decided, ambiguities surrounding AI art and copyrights remain.
Endnotes
[1] Ivy, Patrick. “How AI may impact intellectual property rights.” Newburn Law. July 11, 2023. https://www.newburnlaw.com/how-ai-may-impact-intellectual-property-rights/.
[2] IBM. “What are large language models (LLMs)?” IBM. n.d. https://www.ibm.com/topics/large-language-models
[3] Jackson, Amber. “ChatGPT turns one: How AI chatbot has changed the tech world.” Technology Magazine. November 30, 2023. https://technologymagazine.com/articles/chatgpt-turns-one-how-ai-chatbot-has-changed-the-tech-world.
[4] Writers Guild of America. “What We Won.” Writers Guild of America, 2023. https://www.wgacontract2023.org/the-campaign/what-we-won.
[5] United States Copyright Office. “What is Copyright?” United States Copyright Office. n.d. https://www.copyright.gov/what-is-copyright/#:~:text=originality%20and%20fixation-,Original%20Works,and%20%E2%80%9Cmodicum%E2%80%9D%20of%20creativity.
[6] Legal Information Institute. “17 U.S. Code § 106 - Exclusive rights in copyrighted works.” Cornell Law School. n.d. https://www.law.cornell.edu/uscode/text/17/106.
[7] University of Massachusetts Global. “Generative Artificial Intelligence (GAI) Resource Guide for Faculty”. University of Massachusetts Global. n.d. https://umassglobal.libguides.com/artificial-intelligence
[8] Legal Information Institute. “work made for hire.” Cornell Law School. n.d. https://www.law.cornell.edu/wex/work_made_for_hire.
[9] Mathur, Atreya. “Case review: Thaler v. Perlmutter (2023).” Center for Art Law. December 11, 2023. https://itsartlaw.org/2023/12/11/case-summary-and-review-thaler-v-perlmutter/.
[10] Thaler v. Perlmutter, Civil Action No. 22-1564 (BAH), 2023 U.S. Dist. LEXIS 145823 (D.D.C. Aug. 18, 2023).
[11] Legal information Institute. “5 U.S. Code § 706 - Scope of review.” Cornell Law School. n.d.
[12] DRS Law. Understanding motions for summary judgment. DRS Law. n.d.
[13] Thaler v. Perlmutter, Civil Action No. 22-1564
[14] Burrow-Giles Lithographic Company v. Sarony, 111 U.S. 53 (1884)
[15] Court Listener. Stephen Thaler v. Shira Perlmutter (23-5233) Court of Appeals for the D.C. Circuit. Free.Law. October 14, 2024. https://www.courtlistener.com/docket/67892225/stephen-thaler-v-shira-perlmutter/
[16] Abbott, Ryan. “Thaler v. Perlmutter (23-5233) Appelant’s reply brief”. Brown Neri Smith & Khan LLP. April, 10, 2024. https://artificialinventor.com/wp-content/uploads/2024/04/Thaler-v-Perlmutter-Reply-Brief.pdf
[17] Jahner, Kyle & Soni, Aruni. “AI art copyright stays doubtful after Appeals Court argument”. Bloomberg Law. September 19, 2024. https://news.bloomberglaw.com/ip-law/ai-art-copyright-remains-doubtful-after-appeals-court-argument
[18] Jahner, Kyle & Soni, Aruni. “AI art appeal’s procedural flaws put broader ruling in doubt”. Bloomberg Law September 23, 2024. https://news.bloomberglaw.com/ip-law/ai-art-copyright-remains-doubtful-after-appeals-court-argument
[19] Definitions. 17 U.S.C. § 101
[20] Naruto v. Slater, No. 16-15469 (9th Cir. 2018)
[21] Thaler v. Perlmutter, Civil Action No. 22-1564
[22] Thaler v. Perlmutter, Civil Action No. 22-1564
[23] United States Patent and Trademark Office. “Copyright basics”. United States Patent and Trademark Office. n.d. https://www.uspto.gov/ip-policy/copyright-policy/copyright-basics
[24 Exclusive rights in copyrighted works. 17 U.S.C. § 106
[25] United States Patent and Trademark Office. “Copyright basics”.
[26] iDeta Groupe Septeo. “How artificial intelligence has evolved over the years”. iDeta Groupe Septeo. n.d. https://www.ideta.io/blog-posts-english/how-artificial-intelligence-has-evolved-over-the-years#:~:text=The%20evolution%20of%20AI%20has,%2C%20and%20voice%2Dactivated%20assistants.
[27] Haan, Katherine. & Holznienkemper, Lauren. “22 Top AI Statistics And Trends In 2024”. October 16, 2024. https://www.forbes.com/advisor/business/ai-statistics/#top_ai_statistics_section
[28] Tobin, James. “Artificial intelligence: Development, risks and regulation”. UK Parliment House of Lords Library. July 18, 2023. https://lordslibrary.parliament.uk/artificial-intelligence-development-risks-and-regulation/#:~:text=Dependence%20on%20AI%2C%20including%20the,for%20the%20future%20of%20humanity.
[29] Newport Mansions. “Newport and the Gilded Age”. Newport Mansions. n.d. https://www.newportmansions.org/gilded-age/
[30] Legal Information Institute. “writ of certiorari”. Cornell Law School. n.d. https://www.law.cornell.edu/wex/writ_of_certiorari