by Emma Wozniak, Associate Member, University of Cincinnati Law Review Vol. 93
I. Introduction
Art is the manifestation of human imagination. It can forge emotional connections between the creator and the viewer, allowing communication to transcend time and space. Art and other creative works have long been revered, protected, and displayed in museums and libraries worldwide. Our society recognizes the importance of protecting creators’ rights and encouraging them to continue producing new works in the form of intellectual property.[1]
The term “intellectual property” refers to creations of the mind like signs, symbols, artistic expressions, designs, and inventions.[2] The United States government grants intellectual property rights to creators, allowing them to control the use of their designs or inventions by others.[3] However, an emerging issue in this field is the creation of artificial intelligence (“AI”). AI has been developed with the capability to generate images, music, and videos.[4] With AI as a tool, anyone can create an artistic masterpiece. Recent developments in image generators blur the lines between human-made and AI art.[5]
While AI has made creating art more accessible to the masses, the products of AI tools lack the same protection provided to human-made pieces.[6] This article examines the implications of current copyright law on AI-generated art. Part II of this article will provide background on what AI is and how image generators work. Part II will also delve into current copyright laws and cases that address the issue of whether AI-generated artwork can receive copyright protection. Part III of this article will discuss the consequences of the current state of regulation and provide suggestions on how AI artists could gain protection for their works. Finally, Part IV concludes with predictions on the future of copyright protection on AI-generated artwork.
II. Background
AI can be defined in various ways, but the United States Copyright Office’s Notice of Intent broadly defines AI as a category of systems that are automated to perform tasks typically associated with cognitive function or human intelligence.[7] This includes activities such as learning, reasoning, problem-solving, and language processing.[8] Specifically, the Copyright Office defines AI systems as software services or products developed for use by end-users, that substantially incorporate AI models or combinations of computer code and numerical values designed to accomplish specific tasks.[9] Essentially, AI is software trained on large datasets and designed to generate intelligent outputs based on user-provided inputs.[10] These systems can range from generative tools that produce texts, images, or music, to decision-making machines used in data analysis and automation.[11] As AI continues to evolve and expand across industries, clear definitions are crucial for informing legal and regulatory frameworks.
AI encompasses a wide range of types and subsets, each designed to perform different functions. However, this article focuses specifically on Generative Artificial Intelligence (“Generative AI”). Generative AI refers to a class of AI applications that can take commands, or prompts, from a user to generate expressive material such as images, audio, video, and text.[12] Generative AI differs from other forms of AI. Generative AI can create new data, rather than merely making predictions based on inputted data.[13] Generative AI is inherently creative in nature – a trait that sets it apart from the automated decision-making of traditional AI.[14] The creative capabilities of Generative AI have significant implications across a variety of industries, including design, entertainment, and law.
Generative AI models require a massive amount of data, or information, to function effectively.[15] To train an accurate AI model, the model must be trained on high volumes of high-quality data, which helps the AI model understand patterns, relationships, and nuances within the data.[16] Training data can be sourced in a few different ways. Due to the enormous quantity of data required for a successful generative AI model, a popular way for developers to source data is through a process known as web scraping.[17] Web scraping involves extracting a large quantity of information from websites.[18] While this process provides developers with a vast reservoir of data, not all of this data is free to use, as some websites prohibit scraping and some scraped content has copyright protections.[19]
To learn from this collected data, Generative AI models rely on deep learning, a subtype of machine learning.[20] Deep learning employs “neural networks,” which is a concept inspired by the human brain.[21] These networks consist of multiple layers of interconnected nodes that process and pass information, allowing the AI system to identify complex patterns and generate outputs.[22] This structure enables Generative AI to perform creative tasks, like generating text, images, and music, by learning about each of these structures from the data the model is trained on.[23]
Once properly trained, Generative AI can produce complex and remarkable outputs. While Generative AI is widely discussed, it is not a new technology. In 1948, Alan Turing, a British mathematician, wrote in an unpublished paper titled “Intelligent Machinery” about many central concepts of AI including the idea of training networks of artificial neurons to perform specified tasks.[24] A few years later, Arthur Samuel developed the first machine learning game, a checkers-playing program.[25] The first instance of AI-generating art began with Harold Cohen.[26] In 1973, Cohen developed a program named AARON which used algorithms to generate new and unique art.[27] The early version of AARON could only create abstract drawings limited to imagery like plants and rocks, however, later developments of the program allowed for color and the addition of more representational figures.[28]
AI art includes images, music, videos, and writing that are created, enhanced, or edited with AI technology.[29] AI art has vastly evolved from Cohen’s AARON in the 1970s. The introduction of deep learning and neural networks to AI enables new Generative AI models to produce realistic artwork based on a more complex understanding of the artistic styles and principles it was trained on.[30] Now, AI art is becoming almost indistinguishable from human-made art.[31]
A variety of AI art generators are available to creators. Popular models for AI artists include Leonardo.AI, Microsoft Designer, DALL-E, and Adobe Firefly.[32] Each generator offers a unique experience for AI artists, with some models producing more hyper-realistic imagery or text in images than others.[33] AI-generated art has made a splash in the art world. In 2018, Generative AI was used to create a portrait of a fictional character created by members of the Parisian Obvious collective.[34] The collective used an open-source algorithm and trained it on approximately 15,000 14th to 19th-century portraits.[35] The portrait produced, entitled Edmond de Belamy, was sold at an auction in Christie’s, New York for $432,500.[36]
While Generative AI has demonstrated the ability to produce high-value artwork, creators of AI-generated art in the U.S. face significant challenges in securing copyright protection for their creations. Copyright, a form of intellectual property, protects “original works of authorship” when an author “fixes” the work in a “tangible form of expression.”[37] Works that authors could copyright include paintings, books, poems, illustrations, and movies.[38] A work becomes “fixed in a tangible medium of expression” when an author captures the work in a sufficiently permanent form.[39] This threshold can be satisfied by writing the work on paper, recording it on tape, or sculpting the work out of clay.[40]
Another copyright requirement is that the work must be original, and this requirement has two parts.[41] To be “original” a work must be independently created and display a “spark” or “modicum” degree of creativity.[42] Independent creation is a low bar. Generally, as long as the work is not a copy, it can be considered independently created.[43] As for creativity, in Feist Publications v. Rural Telephone Service Co., the U.S. Supreme Court held that only an extremely low level of creativity is required, even a slight amount of creativity will satisfy this requirement.[44] For example, telephone directory white pages would be deemed to lack minimal creativity since they contain raw data or facts.[45]
While certain AI-generated work may run into complications meeting the criteria discussed above, the most difficult and contested requirement for AI art is the necessity of human authorship.[46] While the U.S. Constitution and Congress do not explicitly require human authorship, the U.S. Copyright Office has taken the position that for a work to receive copyright protection, it must have a human author.[47] In the second part of the Copyright Office’s report Copyright and Artificial Intelligence, the office contemplates whether the outputs of AI art generators should be permitted copyright protection.[48] The Copyright Office posits that the answer rests on the nature and extent of a human’s input to the work, and whether the contribution qualifies as authorship of expressive elements within the output.[49]
Shortly after the Copyright Office released guidance on the availability of copyright protection for AI-generated art, a significant legal ruling emerged in the case of Thaler v. Perlmutter, which further clarified this issue.[50] In Thaler, a computer scientist, named Dr. Stephen Thaler, attributed authorship of an artwork titled “A Recent Entrance to Paradise” to a generative AI model named the “Creativity Machine.”[51] In his copyright application, Dr. Thaler listed the Creativity Machine as the sole author of the artwork, while he claimed ownership of the work for himself.[52] However, the Copyright Office denied this application, finding that it failed to establish the human-authorship requirement because the artwork was autonomously generated by an AI model rather than a human creator.[53] The District Court upheld this decision, affirming the Copyright Office’s stance that all eligible work must be authored by a human being.[54]
The court provided several reasons to justify its ruling and identified why human authorship is necessary. Primarily, it emphasized that copyright is fundamentally a property right, and as such, it cannot be held by non-human entities like AI.[55] The court further explained that the Copyright Act (the “Act”) consistently treats machines as tools rather than authors.[56] One of the key examples is the Act’s definition of “computer program” as a “set of statements or instructions to be used directly or indirectly to bring about a certain result.”[57] This description underscores that machines are seen as instruments used by human authors to achieve a desired outcome, rather than autonomous creators.[58] As a result of this ruling, AI-generated artwork cannot be considered eligible for copyright protection under the current framework, as the law requires a human creator to be the author.[59]
Despite this decision, the door remains open for potential copyright protection for works created with the aid of Generative AI. While Thaler firmly established that AI itself cannot be considered an author under current copyright law, artists and creators who use AI as a tool in their creative process may still have an opportunity to claim copyright for their contributions to the work, provided they can demonstrate sufficient human involvement. This case marks a pivotal moment in the ongoing legal and ethical discussion surrounding the intersection of AI, intellectual property rights, and creative expression. While Thaler offers important clarity on the current legal stance, it is clear that the landscape of copyright law is evolving. As AI technologies continue to advance, there are likely to be many more legal challenges that explore new ways to adapt existing frameworks, potentially creating avenues for protecting AI-assisted artistic endeavors in the future.
III. Discussion
The use of technology does not preclude an artist from earning copyright protection.[60] The relationship between human authorship and technological assistance has long been the subject of inquiry. As early as the 1960s, the Copyright Office began assessing whether material created using emerging technologies should have authorship attributed to human creators or the machines that assisted them.[61] While Generative AI is a far cry from 1960s technology, the underlying legal question remains strikingly similar – does the involvement of a technological tool change the nature of human authorship? The Copyright Office’s response appears to be the same: there is no one-size-fits-all answer.[62] Each case must be assessed based on the extent of human contribution and control over the final work.
The ruling in Thaler provides one clear answer, that neither the Copyright Office nor the courts are prepared to recognize works authored solely by Generative AI as copyright-eligible. However, this ruling does not foreclose the possibility that some AI-assisted works could still qualify for copyright protection. Artists have a range of ways to incorporate generative AI technology in their work. The critical factor for determining copyright eligibility seems to be the degree of human involvement in the final work.[63]
Aligned with the outcome of Thaler, the Copyright Office provides that given the available AI technology, written prompts alone do not provide sufficient human control to make users the authors of an output.[64] Currently, users providing prompts to a Generative AI model do not oversee, direct, or understand the contributions of the model.[65] While the projected future for AI capabilities may open up this avenue, the current state of today’s models does not allow prompts to adequately determine what expressive elements are produced or control how the model translates the prompt into an output. Currently, AI is described as a “black box.”[66] This means that the inner workings of AI models are often opaque. Even experts in AI technology have a limited understanding and ability to predict behavior for specific Generative AI models.[67] AI system outputs can vary across requests, even if the AI system receives identical prompts.[68] This feature demonstrates that AI systems cannot yet translate an idea into a fixed, tangible expression that is entitled to copyright protection.
AI models are capable of accepting more than just written prompts. Some Generative AI models process various types of inputs, including images, video, or audio, and are designed to modify these inputs based on specific instructions provided by the user. The Copyright Office acknowledges that in some situations in which the image input is itself a copyrightable work, there may be an avenue for the output to be copyright-eligible. However, similar to the concept of derivative works, copyright protection would likely only extend to the original human expression found in the output.[69] This means the new creation would be protected only to the extent that it reflects a transformation of the original work through human creativity.
On the other hand, if an AI output undergoes significant modification or enhancement by a human creator, this could also provide a basis for copyright protection. In such cases, human input could be seen as substantial enough to support the argument for authorship, even while AI tools are used as part of the creative process. Essentially, the use of AI in generating content would not automatically negate the copyrightability of the work as a whole, provided that there is clear evidence of significant human authorship and creative involvement in shaping the final product. This is a key distinction in determining how AI-assisted works are treated under existing copyright laws.
Generative AI technology is a new frontier for artists and creators. While the Copyright Office has begun to offer guidance on the copyrightability of current AI capabilities, there are still no clear guidelines. Just like in Thaler, for the time being, creators will likely have to attempt to copyright AI art and rely on court decisions to get bright-line rules. However, this could be an arduous and time-intensive process. In a field where the available technology and its uses are evolving by the day, a faster route for answers may be required.
Congress could provide another avenue for clarification on the human authorship requirement of the Copyright Act. A clear understanding could be reached if Congress defines authorship as either an exclusively human requirement or inclusive of AI-assisted contributions. It is important for courts and legislatures to consider encouraging technological innovation and the protection of the creator’s rights when crafting new rules. Until then, artists should continue to explore new technologies and push the boundaries of creation.
IV. Conclusion
The rapid evolution of AI-generated art offers exciting opportunities for creators. However, as this technology continues to revolutionize the creative landscape, it also presents complex legal challenges, particularly regarding the issue of copyright protection. As AI systems become increasingly capable of generating sophisticated and original artwork, the question of whether and how these creations can be copyrighted remains unresolved.
Currently, the legal stance in the United States emphasizes the necessity of human authorship for copyright protection, meaning that works created entirely by AI without human involvement are ineligible for such protection. However, the rapid development of AI technologies, as well as shifting cultural attitudes toward creativity and authorship, leaves room for potential legislative or judicial interpretations of what constitutes authorship in the context of AI-generated works. Future legal developments may reshape the concept of authorship, particularly as the line between human and machine-generated creativity becomes increasingly blurred.
In the absence of clear and comprehensive legal guidelines, creators who use AI technology to generate artwork must remain vigilant and informed about the copyright implications for their work. It may be beneficial for these creators to consider strategies that ensure human authorship is demonstrated, such as providing substantial input in the creation process or combining AI outputs with their original contributions. Alternatively, some creators may choose to push the boundaries of existing legal frameworks, testing the limits of current copyright law in hopes of advancing the conversation and fostering legal reform.
As AI continues to challenge traditional notions of creativity, originality, and ownership, it is crucial for legal frameworks to evolve in a way that balances the need for innovation with the protection of creators’ rights. The law must adapt to accommodate new technologies while ensuring that all creators can maintain control over their intellectual property.
[1] Copyright Basics, U.S. Pat. and Trademark Office, https://www.uspto.gov/ip-policy/copyright-policy/copyright-basics?utm_source=chatgpt.com (last visited Apr. 24, 2025).
[2] Intellectual Property: Protection and Enforcement, World Trade Org., https://www.wto.org/english/thewto_e/whatis_e/tif_e/agrm7_e.htm (last visited Apr. 24, 2025).
[3] Id.
[4] Robert Sheldon & Sean Micheal Kerner, What is AI Art and How is it Created?, TechTarget (Nov. 2024), https://www.techtarget.com/searchenterpriseai/definition/AI-art-artificial-intelligence-art.
[5] Benji Edwards, OpenAI’s New AI Image Generator is Potent and Bound to Provoke, ArsTechnica (Mar. 27, 2025), https://arstechnica.com/ai/2025/03/openais-new-ai-image-generator-is-potent-and-bound-to-provoke/.
[6] Gil Appel et al, Generative AI Has an Intellectual Property Problem, Harvard Bus. Rev. (Apr. 7, 2023), https://hbr.org/2023/04/generative-ai-has-an-intellectual-property-problem.
[7] Artificial Intelligence and Copyright, 88 Fed. Reg. 59942 (Aug. 30, 2023).
[8] Id.
[9] Id.
[10] Cole Stryker & Eda Kavlakoglu, What is Artificial Intelligence (AI)?, IBM (Aug. 9, 2024), https://www.ibm.com/think/topics/artificial-intelligence.
[11] Id.
[12] Artificial Intelligence and Copyright, 88 Fed. Reg. 59942 (Aug. 30, 2023).
[13] Adam Zewe, Explained: Generative AI, MIT News (Nov. 9, 2023), https://news.mit.edu/2023/explained-generative-ai-1109.
[14] Paramita Ghosh, Generative AI vs. Traditional AI, Dataversity (Aug. 28, 2024), https://www.dataversity.net/generative-ai-vs-traditional-ai/.
[15] What is a Neural Network, IBM (Oct. 6, 2021), https://www.ibm.com/think/topics/neural-networks.
[16] What Data is Used to Train an AI, Where Does it Come From, and Who Owns it?, Potter Clarkson, https://www.potterclarkson.com/insights/what-data-is-used-to-train-an-ai-where-does-it-come-from-and-who-owns-it/ (last visited Apr. 24, 2025).
[17] Nayna Jaen, Determining What AI Data you Need and How to Source it, RWS (Apr. 8, 2024), https://www.rws.com/artificial-intelligence/train-ai-data-services/blog/determining-ai-data-needs-and-sourcing/#:~:text=Web%20scraping%20involves%20extracting%20large,various%20web%20sources%20relatively%20quickly.
[18] Id.
[19] Id.
[20] Adam Buick, Copyright and AI Training Data- Transparency to the Rescue?, 20 J. of IP Law & Prac. 182, 184 (2024).
[21] Id. at 184.
[22] Id.
[23] Id.
[24] B.J. Copeland, History of Artificial Intelligence (AI), Britannica (Apr. 19, 2025), https://www.britannica.com/science/history-of-artificial-intelligence.
[25] Matthew, Artificial Intelligence Timeline: The History of AI Art, AI Art Kingdom (Jan. 21, 2024), https://www.aiartkingdom.com/post/artificial-intelligence-timeline#:~:text=The%20journey%20of%20AI%20art,breakthroughs%20in%20technology%20and%20techniques.
[26] Id.
[27] Id.
[28] Is AI-Generated Art Actually Art?, Univ. of Plymouth, https://www.plymouth.ac.uk/discover/is-ai-generated-art-actually-art (last visited Apr. 24, 2025).
[29] Matthew, supra note 25.
[30] Michael Filimowicz, The History and Evolution of AI-Generated Art, Medium (Jun. 4, 2023), https://medium.com/higher-neurons/the-history-and-evolution-of-ai-generated-art-e5ccca5a8e83.
[31] Id.
[32] Nada Shawky, 6 Best AI Art Generators Right Now (I Tried Them All), Medium (Jan. 21, 2025), https://medium.com/@nada.mo.shawky/6-best-ai-art-generators-right-now-i-tried-them-all-acc445c51427.
[33] Id.
[34] Pavitra M, 10 Mind-Blowing Examples of AI-Generated Art, Clickup (Aug. 22, 2024), https://clickup.com/blog/ai-art-examples/#:~:text=Portrait%20of%20Edmond%20de%20Belamy,-Edmond%20de%20Belamy&text=This%20work%20of%20art%20gained,for%20%24432%2C500%20in%20that%20auction.
[35] Id.
[36] Id.
[37] What is a Copyright, Copyright.gov, https://www.copyright.gov/what-is-copyright/ (last visited Apr. 24, 2025).
[38] Id.
[39] Id.
[40] Copyright Basics, Univ. of Michigan Libr. (Jan. 6, 2025), https://guides.lib.umich.edu/copyrightbasics/copyrightability.
[41] What is a Copyright, supra note 37.
[42] Id.
[43] Mackenzie Caldwell, What Is an “Author”?-Copyright Authorship of AI Art Through a Philosophical Lens, 61 Hous. L. Rev. 411 (2023).
[44] Fiest Publications, Inc. v. Rural Tel. Serv. Co., 499 U.S. 340 (1991).
[45] Creativity Requirement, USLEGAL.com, https://copyright.uslegal.com/enumerated-categories-of-copyrightable-works/creativity-requirement/ (last visited Apr. 24, 2025).
[46] Caldwell, supra note 43.
[47] Id.
[48] Shira Perlmutter, Copyright and Artificial Intelligence Part 2:Copyrightability, U.S. Copyright Office (Jan. 2025),
[49] Id.
[50] Thaler v. Perlmutter, 130 F.4th 1039 (D.C. Cir. 2025).
[51] Id. at 2.
[52] Id.
[53] Id. at 3, 7.
[54] Id. at 3.
[55] Id. at 10.
[56] Id. at 12.
[57] Id. at 12.
[58] Id.
[59] Blake Brittain, US Appeals Court Rejects Copyrights for AI-Generated Art Lacking ‘Human’ Creator, Reuters (Mar. 18, 2025), https://www.reuters.com/world/us/us-appeals-court-rejects-copyrights-ai-generated-art-lacking-human-creator-2025-03-18/.
[60] Perlmutter, supra note 48, at 2.
[61] Id.
[62] Id.
[63] Id. at 21-22.
[64] Id. at 18.
[65] Id. at 19.
[66] Id. at 6.
[67] Id.
[68] Id. at 7.
[69] Id. at 24.
Cover Photo generated using ChatGPT.
