Assessing Human Authorship in AI-Generated Works
7.14.2026
A banana duct-taped to a wall sold for $6.2 million. Maurizio Cattelan’s work, “Comedian,” has been deemed one of the most “talked about works of the century,”[1] though much of the conversation has been sharply critical.[2] Indeed, when art is produced with minimal skill and creativity, inevitably, it garners scrutiny. Perhaps it is no surprise that art generated by artificial intelligence has been labeled “the banana taped to the wall of the digital world”[3] for much the same reason. Criticism is a natural companion to change, after all.
Many resist the artist who merely “presses a few buttons” to make a digital art piece.[4] Copyright law shares that hesitation. Indeed, human authorship is necessary to receive copyright protection, something entirely AI-generated works cannot satisfy.[5] However, some AI-generated works demonstrate a greater degree of human involvement and creativity, beyond a simple exercise of pressing buttons. Further, as of late, copyright law has moved toward recognizing some of those works as being worthy of protection, too.[6] At the same time, the question of how much human involvement is enough to meet the authorship standard is left open for interpretation.[7]
The human authorship assessment has been left to be dealt with on a case-by-case basis.[8] Certainly, a degree of flexibility is warranted, and a one-size-fits-all approach is unfavorable. However, without a guiding framework for assessing human authorship, the existing approach remains riddled with uncertainty. Uncertainty creates risk, and there is much at stake. Some creators may hide their use of AI to improve their chances of obtaining copyright protection, while others may avoid experimenting with new technologies altogether. Yet copyright law is designed to “secure a fair return for an author’s creative labor,” with the ultimate aim of “stimulat[ing] artistic creativity for the general public good.”[9] Clarifying the human authorship standard, then, is not only consistent with the copyright regime, but integral to its purpose.
The following research examines guidance from the U.S. Copyright Office on the copyrightability of AI-generated works and proposes a principled framework for evaluating human authorship.
The ‘Conclusions’ on AI-Generated Works
Copyright protects the fruits of intellectual labor founded in the creative powers of the mind.9 Only material “that is the product of human creativity”[10] will be accorded copyright protection.[11] Perhaps unsurprisingly then, where the “traditional elements of authorship were produced by a machine, the work lacks human authorship.”[12] Purely AI-generated material, without any human contribution, is therefore not copyrightable.[13] Yet the door is not entirely foreclosed on registering AI-generated works. In some cases, a “work containing AI-generated material will also contain sufficient human authorship to support a copyright claim.”[14] The lingering question, again, is not whether human authorship is required; that has been answered in the affirmative. Now, the question is how much.
In 2025, the U.S. Copyright Office released its highly anticipated Copyright and Artificial Intelligence Guidance Report. which included a set of “conclusions” on the copyrightability of AI-generated works:[15]
- Questions of copyrightability and AI can be resolved pursuant to existing law, without the need for legislative change. (Conclusion #1)
- The use of AI tools to assist rather than stand in for human creativity does not affect the availability of copyright protection for the output. (Conclusion #2)
- Copyright protects the original expression in a work created by a human author, even if the work also includes AI-generated material. (Conclusion #3)
- Copyright does not extend to purely AI-generated material or material where there is insufficient human control over the expressive elements. (Conclusion #4)
- Whether human contributions to AI-generated outputs are sufficient to constitute authorship must be analyzed on a case-by-case basis. (Conclusion #5)
- Based on the functioning of current generally available technology, prompts do not alone provide sufficient control. (Conclusion #6)
- Human authors are entitled to copyright in their works of authorship that are perceptible in AI-generated outputs, as well as the creative selection, coordination, or arrangement of material in the outputs, or creative modifications of the outputs. (Conclusion #7)
- The case has not been made for additional copyright or sui generis protection for AI-generated content. (Conclusion #8)
To Register or Not To Register? That Is the Question
The following section examines six key copyright decisions, listed chronologically, in which the U.S. Copyright Office has either granted or denied copyright registration for AI-generated works. This section interprets these decisions in light of the office’s report conclusions above. Taken together, the guidance provides the basis for the assessment framework proposed in the conclusion.
‘A Recent Entrance to Paradise’ by Steven Thaler (denied, August 2019)
The first major decision addressing the human-authorship standard for AI-generated works is Thaler v. Perlmutter, which confirmed that AI cannot be an “author” of a work.[16]
Steven Thaler attempted to register a piece of AI-generated artwork entitled “A Recent Entrance to Paradise.” In his copyright application, Thaler stated the artwork was generated autonomously by an AI program he owned called Creativity Machine.[17] Thaler sought copyright in “A Recent Entrance to Paradise,” arguing that the AI-generated work should qualify as a “work-for-hire” since he owned Creativity Machine.[18] The Copyright Office denied Thaler’s application because the work “lack[ed] the human authorship necessary to support a copyright claim.”[19] Thaler requested reconsideration of his application. It was again denied, reiterating that the work lacked the requisite human authorship to receive copyright protection. Thaler made a second request for reconsideration, which was denied for the same reason. He then challenged the decision in court, arguing that it was “arbitrary … and not in accordance with the law.”[20]
The U.S. District Court of Columbia heard the challenge and upheld the Copyright Office’s decision,[21] holding that there was no valid copyright “absent human involvement.”[22] The decision was appealed, and again affirmed by the U.S. Court of Appeals for the District of Columbia, where it was unequivocally stated that “[t]he Copyright Act of 1976 requires all eligible work to be authored in the first instance by a human being.”[23] The Supreme Court recently denied review of the decision, a final nail in the coffin. However, the decision is unsurprising. Indeed, “A Recent Entrance to Paradise” “lack[ed] traditional human authorship”[24] and Conclusion #4 already makes clear that “[c]opyright does not extend to purely AI-generated material.”[25]
‘Théâtre D’opéra Spatial’ by Jason M. Allen (denied December 2022)
Next, Jason Allen attempted to obtain copyright protection over “Théâtre D’opéra Spatial,”[26] an AI-generated art piece that won the Colorado State Fair art contest.[27] While the image was entirely generated by Midjourney, Allen explained that he used a “series of prompts … adjust[ed] the scene, select[ed] portions to focus on, and dictate[d] the tone of the images,” and used Adobe Photoshop to edit the image after it was produced.[28] Allen asserted that he experimented with over 600 prompts before arriving at the text prompt that was ultimately used to create the artwork.[29]
Allen sought copyright registration of “Théâtre D’opéra Spatial” as a whole. He also refused to disclaim the portions attributed to AI.[30] The U.S. Copyright Office refused to register the work because the artwork “inextricably merged, inseparable contributions” of Allen and Midjourney.[31] Allen requested a secondary review of the decision and again was denied registration. In reaching this conclusion, the Copyright Office considered the circumstances surrounding the work’s creation, including Allen’s use of over 600 text prompts to refine the final input that generated the image. However, the office still determined that “Mr. Allen’s actions as described do not make him the author of the Midjourney image, because his sole contribution … was inputting the text prompt that produces it.”[32]
The denial of “Théâtre D’opéra Spatial” illustrates that using a text prompt alone will not meet the required threshold for human authorship. This is reiterated in Conclusion #6, which provides that, based on currently available technology, prompting alone does not meet the requisite threshold for human authorship.[33] Importantly, however, Allen did more than just input a prompt into Midjourney. After all, he did use Adobe Photoshop “to remove flaws and create new visual content” after the image was generated.[34] Conclusion #7 provides that human authors are entitled to copyright in their creative modifications of AI outputs.[35] However, the Copyright Office did not decide whether Allen’s modification rose to the level of copyrightability because there was a lack of sufficient evidence to make that determination.[36]
‘Zarya of the Dawn’ by Kristina Kashtanova (granted February 2023)
In 2023, Kristina Kashtanova successfully obtained copyright protection for “Zarya of the Dawn,” a comic book comprised of AI-generated images alongside human-authored text.[37] The Copyright Office initially granted full copyright protection over “Zarya of the Dawn” because Kashtanova did not state that the images were AI-generated in her application.[38] After learning of the AI-generated material, the office amended the registration to recognize Kashtanova as “the author of the Work’s text as well as the selection, coordination, and arrangement of the Work’s written and visual elements.”[39] However, the AI-generated images alone were not copyrightable.[40]
“Zarya of the Dawn” demonstrates that AI-generated content that is part of a larger, human-authored work will not impact the copyrightability of the work as a whole.[41] Conclusion #3 echoes the same sentiment that copyright protects the original, human expression in a work, even if that work includes AI-generated material.[42] This decision also illustrates Conclusion #7, namely, that human authors are entitled to copyright in the creative selection, coordination, or arrangement of material in the outputs.[43]
‘SURYAST’ by Ankit Sahni (denied June 2023)
Following the successful registration of “Zarya of the Dawn,” Ankit Sahni sought copyright protection for his artwork titled “SURYAST,” which he created using an AI painting app called RAGHAV.[44] Sahni listed RAGHAV as one of the authors when he attempted to register the work.[45] To create “SURYAST,” Sahni inputted one of his photographs into RAGHAV and provided Vincent van Gogh’s “The Starry Night” painting as a “style input.”[46] The resulting image, “SURYAST,” differed quite substantially from his original inputted photograph.
The U.S. Copyright Office denied registration. Sahni argued that the work contained sufficient human authorship because he used a copyrightable photographic input and directed RAGHAV to change the colors, shapes and style in a particular manner.[47] The office disagreed. The reviewers took issue with the fact that when Sahni entered the style input, RAGHAV controlled where many of the expressive elements would be placed, as opposed to Sahni.[48]
Conclusion #4 is clear: Copyright does not protect works where there is insufficient human control over the expressive elements.[49] Conclusion #2 also provides that using AI tools to assist rather than replace human creativity does not affect the availability of copyright protection for the resulting work.[50] Here, the Copyright Office did not see RAGHAV as an assistive tool, but rather as a substitute for human creativity.
Sahni provided an original photograph as an input in RAGHAV. Conclusion #7 provides that copyright may protect original works of human authorship that are perceptible in AI-generated outputs.[51] If the photograph that Sahni used in RAGHAV was perceptible in the output, he would be entitled to copyright protection, at least for the portions that were clearly perceptible. Yet that was not the case – when RAGHAV applied the style of “The Starry Night” painting, the resulting image differed significantly and Sahni’s photographic contribution could no longer be distinguished in the final work. The Copyright Office found that, given the substantial change to the image in the resulting output, the “human authorship” could not be “distinguished or separated from the final work produced by the computer program.”[52]
‘Rose Enigma’ by Kristina Kashtanova (granted late 2023)
Following the denial of “SURYAST,” Kristina Kashtanova succeeded in obtaining copyright protection for her art work, “Rose Enigma.”[53] To make “Rose Enigma,” Kashtanova provided a copyrightable, hand-drawn illustration as well as a text prompt into Stable Diffusion that read: “a young cyborg woman (((roses))) flowers coming out of her head, photorealism, cinematic lighting, hyperrealism, 8k, hyper-detailed.”[54] The resulting image, “Rose Enigma,” clearly resembled the initial hand-drawn illustration used in the input.[55] Kashtanova properly disclaimed any non-human expression in the final work. Accordingly, the registration was “limited to unaltered human pictorial authorship that is clearly perceptible in the deposit and separable from the non-human expression that is excluded from the claim.”[56]
One of the biggest differentiators between the success of “Rose Enigma” and the preceding denial of “SURYAST” was the perceptibility of human contribution. When creating “Rose Enigma,” Kashtanova used a hand-drawn illustration, and that illustration was clearly recognizable in the final AI-generated image (distinguishing it from “SURYAST”). “Rose Enigma” looked like an enhanced version of the original hand-drawn illustration. The Copyright Office has stated that when a human inputs their own copyrightable work, and “the AI output bears a clear resemblance to an original and otherwise copyrightable work,” the human “will be the author of at least that portion of the output.”[57] This, too, is reiterated in Conclusion #7, which states that human authors are entitled to copyright in their works of authorship if those works are perceptible in the AI-generated output.[58]
‘A Single Piece of American Cheese’ by Kent Kiersey (granted January 2025)
Following the successful copyright registration of “Rose Enigma,” Kent Kiersey succeeded in obtaining copyright protection for his work titled “A Single Piece of American Cheese,” an AI-generated image created with Stable Diffusion.[59] This decision marks the first time the U.S. Copyright Office granted copyright protection for an image created entirely using AI-generated material.[60]
Unlike the works of “Rose Enigma” and “SURYAST,” Kiersey did not feed an original work into Stable Diffusion to generate the work. The image was entirely AI-generated using a textual prompt (which, alone, is insufficient evidence of authorship). However, after the image was created, Kiersey further refined the work by manually erasing, adjusting, and editing the AI-generated portions of the image. Initially, the Copyright Office denied registration of “A Single Piece of American Cheese.” However, the decision was reversed in 2025 (notably around the same time the U.S. Copyright Office report was released) after Kiersey successfully argued that the work met the standard for human selection, coordination and arrangement.[61]
This marks the first instance where the Copyright Office has granted copyright protection to a fully AI-generated image (with significant human involvement at the output stage). The rationale for doing so is articulated in Conclusion #7 – human authors are entitled to copyright in the creative selection, coordination or arrangement of material in the outputs.[62] Additionally, Conclusion #5 provides that human authors are entitled to copyright in their creative modifications of AI-generated outputs.[63]
A Principled Assessment of Human Authorship
The Copyright Office’s decisions to register or deny AI-generated works appear grounded in uniform reasoning. However, the method for assessing human authorship still lacks a transparent, principled approach. To be clear, rigidity is not favorable. The Copyright Office is unequivocal on this point: assessing whether the human contributions are sufficient to support copyright registration is a case-by-case analysis.[64] However, a case-by-case analysis can still be undertaken with a principled and transparent assessment; the two are not mutually exclusive.
While copyright law has always required a degree of flexibility, such flexibility must not come at the expense of clarity. The fair use defense to copyright infringement, for example, is also determined on a “case-by-case analysis rather than bright line rules.”[65] Even so, the fair use analysis is nonetheless guided by a clear set of considerations, including the purpose and character of the use, the nature of the copyrighted work, the amount and substantiality of the portion used, and the effect of the use upon the market.[66] These factors are to be “explored and weighed together.”[67] A similar set of considerations should be clearly articulated and weighed together when assessing human authorship, as well.[68]
Legal scholar Johannes Fritz proposed a four-step test for assessing human authorship in partially AI-generated works that considers who is using the AI system; the type of AI system being used; the extent of subjective human judgment involved; and the degree of human control exercised over the output.[69] Using the Copyright Office’s report’s conclusions and aspects of Fritz’s framework, this research proposes a principled three-prong method for evaluating human authorship, consistent with the current stance of the U.S. Copyright Office. When assessing human authorship, the office should examine, in order, the (a) nature; (b) extent; and (c) perceptibility of the human contribution.
Nature of the Human Contribution
First, the nature of the human contribution should be observed. At this stage, the focus should be on identifying what the human contribution(s) were and when they arose in the creative process. This should be treated as an information-gathering exercise, and the sufficiency of those contributions should not yet be scrutinized.
Human contribution may arise at the input stage, like contributing a hand drawing or otherwise copyrightable image into an AI image generator (like in “Rose Enigma” and “SURYAST”) or using a prompt to instruct the AI to complete a certain task. Note that, while Conclusion #6 states that prompting alone is insufficient human authorship,[70] this is not a determination to be made at this stage; the use of a prompt should nonetheless be noted. The nature of the contribution may also arise at the output stage, such as arranging AI-generated outputs through a manual process (like in “A Single Piece of American Cheese” or “Zarya of the Dawn”); editing an AI-generated output by selecting, manually erasing, or otherwise altering the AI-generated content (like in “A Single Piece of American Cheese”); merging the AI-generated output with a copyrightable image or drawing; or refining the output through new prompting.
Note that human contribution is not mutually exclusive at the input or output stage. Often, there will be human involvement on both ends (like providing a prompt, and then manually adjusting the image output). Once it has been determined at what point(s) the human contributed, the Copyright Office can proceed to the next prong.
Extent of the Human Contribution
Next, the Copyright Office should examine and scrutinize the extent of the human contribution. This prong shifts the focus from identifying the contribution to evaluating the contribution. Fritz suggests that human authorship must be assessed with reference to the type of AI tool used and the degree of human control over the output.[71] The extent of the human contribution should be assessed with reference to both considerations, as well as how the AI tool was used.
The Copyright Office already takes these inquiries into account – copyright will not protect material where there is insufficient control over the expressive elements.[72] What matters is the “extent to which the human had creative control over the work’s expression.”[73] The use of AI tools “to assist rather than stand in for human creativity” still permits copyright protection for the output.[74] All of these considerations are interconnected. To determine whether the system assisted, or stood in, for human creativity, the Copyright Office must examine how much control the human had over the expressive elements. To determine the level of control over the expressive elements, the Copyright Office must also examine how the tool was used and the capabilities of the AI tool itself.
Fritz argues that examining the capabilities of the AI tool can be done by classifying AI tools into three categories – partially, highly, and fully autonomous.[75] Yet technology changes too fast; it cannot and should not be siloed as such. The Copyright Office also rejects characterizing different AI systems by capabilities, contending that the important inquiry is “how the system is being used, not on its inherent characteristics.”[76] While categorizing tools into arbitrary categories is unfavorable, respectfully, the Copyright Office also creates an unnecessary distinction. Examining the capabilities of the AI tool, at least generally, is necessary, and somewhat unavoidable, when assessing the level of human control over the expressive elements. The two are intertwined.
The Copyright Office has already determined, for example, that when an AI system receives a textual prompt and produces “complex written, visual, or musical works in response,” there is insufficient human control over the output, and thus, a prompt alone does not support human authorship.[77] However, it notes that its stance is “based on the functioning of currently generally available technology,” a conclusion that could only be drawn after examining the characteristics of such generally available technology.[78] AI will continue to evolve, and when it does, the Copyright Office can, and must, continuously reassess its capabilities.
At this stage, the Copyright Office should consider the extent of the human contribution, like the “amount and content of the instructions and input prompts by the AI user” and “the number of generation attempts”[79] with reference to the capabilities of the AI system itself. This, in turn, informs the degree of creative control the human had over the final work. The prong also requires the office to examine any human modifications to AI-generated outputs, and whether there was a requisite degree of control over those modifications as well.
Take, for example, “A Single Piece of American Cheese.” Although it was AI-generated, Kiersey also took time to adjust and erase parts of the image. That process, too, must be examined. To what extent did he contribute? Did Kiersey exercise direct control over the editing process, manually modifying the image himself? Or did he simply instruct Stable Diffusion to make the changes on his behalf? The latter scenario would likely fail to meet the threshold under this prong, whereas the former would support registrability. Once the extent of the human contribution is considered, the Copyright Office can move to the final prong.
Perceptibility of the Human Contribution
The final prong should assess the extent to which the human contribution is perceptible in the final work submitted for registration, whether visibly apparent or substantiated through additional evidence. This prong introduces a clear evidentiary burden, one that is already seemingly required but not clearly articulated.
The Copyright Office has stated that human authors are entitled to copyright in their works of authorship that are perceptible in AI-generated outputs.[80] This conclusion refers to the ability to visually perceive, with the naked eye, when a human-authored work is visible in an AI-generated output (like the hand-drawn image in “Rose Enigma”). Yet when the human contribution is not “perceptible” to the naked eye, the Copyright Office should clearly articulate that there is a burden to make the human contribution perceptible by furnishing additional evidence at this stage.
The evidentiary burden already exists. Recall the creation of “Théâtre D’opéra Spatial.” There, Allen AI-generated the image using a prompt, but also stated that he made modifications to the output using Adobe Photoshop. The Copyright Office found that Allen did not provide sufficient evidence of these modifications in his application. Consequently, the Copyright Office could not determine whether those modifications “rose to the level of copyrightability.”[81] Yet the wise learn from others’ missteps. When creating “A Single Piece of American Cheese,” Kiersey used a software called Invoke Provenance Records, which preserved a detailed history of all his digital modifications and used it as evidence of his contribution.[82] Again, the evidentiary burden is not a new imposition. Evidence of human authorship is already required.
Where sufficient contribution is established under the “extent” analysis in prong 2, that contribution must nonetheless be perceptible under prong 3 – whether visibly apparent or substantiated through additional evidence. The question is: Is the human contribution visible in the final work? If not, can it be made visible through evidence? How much evidence is required will, and already does, differ based on the circumstances. In many cases, the human contribution can be perceived in the final work without much heed; for example, the hand-drawn illustration in “Rose Enigma” was clearly perceptible in the final work and would require limited additional evidence. In contrast, when the work is entirely AI-generated, like “Théâtre D’opéra Spatial” or “A Single Piece of American Cheese,” any human contributions made to the output are unlikely to be perceptible by looking at the final work alone. However, specific evidence of the contribution will cure the defect.
There is, admittedly, a valid concern that requiring artists to track every manual adjustment to AI-generated works could be unduly burdensome. Yes, AI art has been ridiculed as “the banana taped to the wall of the digital world,” but copyright law does not protect the artist who merely “pressed a few buttons to make a digital art piece.”[83] It demands more. Where the human contribution is not clearly perceptible in the final image, it is reasonable to require clear, demonstrable evidence of the contribution, even if that requires a detailed track record.
Fortunately, it is still early enough for artists to adapt. Technology already exists that can track modifications made to AI-generated content.[84] If the challenge of registering AI-assisted art can be resolved through the very tools that created it, perhaps it is not such a bad problem to have after all.
Concluding Thoughts
The threshold for sufficient human authorship remains unclear. Indeed, neither the U.S. Copyright Office nor the court system has provided bright-line guidance on just how much human contribution is required.[85] Perhaps there should be no fixed line in the sand and assessment should remain open-ended. After all, the Copyright Office wants to evaluate authorship on a case-by-case basis, and this research does not seek to change that.
Rather, this article argues for a more transparent, principled framework for assessing human authorship. By harmonizing existing registration decisions with the Copyright Office’s report, this research proposes a more predictable assessment of human authorship, aligned with existing considerations contemplated by the office. This proposal does not seek to entrench the report’s conclusions; they cannot and should not be frozen in time. Rather than using an ill-defined “case-by-case” approach, this research argues for a transparent, principled assessment of human authorship that assesses the (a) nature, (b) extent and (c) perceptibility of the human contribution.
Adopting a principled assessment of human authorship is of paramount importance. Establishing a clear framework can help rebuild confidence in the scope of copyright protection and promote responsible use of emerging technologies. Absent clear guidance, creators may refrain from disclosing AI use when registering their works. Others may resist using AI altogether.
If copyright law is designed to stimulate artistic creativity for the general public good,[86] then establishing a principled framework for assessing human authorship is not just a procedural preference. It is a prerequisite to innovation. And failing to do so may undermine the copyright regime entirely.
This article appears in a recent issue of the EASL Journal, the publication of the Entertainment, Arts and Sports Law Section. For more information, please see nysba.org/easl.
Ambreena Ladhani earned her J.D. from Queen’s University in Canada. She is now part of the Strategic Transactions and Licensing Team at Gunderson Dettmer in New York, where she specializes in licensing and commercializing intellectual property and technology assets.
Endnotes:
[1] Julia Binswanger, Maurizio Cattelan’s Perishable Sculpture Drove Some Critics Bananas. Now, It Could Sell for $1.5 Million, Smithsonian (Oct. 28, 2024), https://www.smithsonianmag.com/smart-news/maurizio-cattelan-perishable-sculpture-drove-some-critics-bananas-now-it-could-sell-for-1-5-million-180985338.
[2] Id.
[3] AI Won an Art Contest, European Commission (Sept. 8, 2022), https://cordis.europa.eu/article/id/442090-trending-science-ai-won-an-art-contest.
[4] Id.
[5] United States Copyright Office, Copyright and Artificial Intelligence, Part 2: Copyrightability (Washington: A Report of the Register of Copyrights, 2025) at Executive Summary iii, https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf (USCO Report).
[6] Id. at Executive Summary iii, bullet 3.
[7] Id. at Executive Summary iii, bullet 5.
[8] Id. at Executive Summary.
[9] Twentieth Century Music Corp. v. Aiken, 422 U.S. 151 (1975).
[10] United States Copyright Office, Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence, 88 Fed. Reg. 16190 (March 16, 2023) (Statement of Policy), 37 C.F.R. pt. 202 at 24 (emphasis added).
[11] See, e.g., Naruto v. Slater, No. 15-cv-04324, 2016 U.S. Dist. LEXIS 11041, at *10 (N.D. Cal. Jan. 28, 2016) (“[Monkey] is not an ‘author’ within the meaning of the Copyright Act”), aff’d, 888 F.3d 418 (9th Cir. 2018) (finding that monkey cannot sue for copyright infringement); USCO Report, supra note 5, at 8.
[12] 88 Fed. Reg. 16190, 16192 (emphasis added).
[13] Id.
[14] Id. at 16191.
[15] USCO Report, supra note 5, at Executive Summary iii, bullets 1–8.
[16] Thaler v. Perlmutter, 687 F. Supp. 3d 140 (D.D.C. 2023).
[17] Id. at 143.
[18] Id.
[19] Id.
[20] Id.
[21] Id. at 151 (“[T]he Copyright Office acted properly in denying copyright registration for a work created absent any human involvement.”).
[22] Id. at 145.
[23] Thaler v. Perlmutter, 130 F.4th 1039, 1041 (D.C. Cir. 2025).
[24] Thaler, supra note 16, at 142.
[25] USCO Report, supra note 5, at Executive Summary iii, bullet #4.
[26] Copyright Review Board, Second Request for Reconsideration for Refusal to Register Théâtre D’opéra Spatial SR # 1-11743923581; Correspondence ID: 1-5T5320R (2023), https://www.copyright.gov/rulings-filings/review-board/docs/Theatre-Dopera-Spatial.pdf.
[27] Kevin Roose, An A.I.-Generated Picture Won an Art Prize. Artists Aren’t Happy, N.Y. Times (Sept. 2, 2022), https://www.nytimes.com/2022/09/02/technology/ai-artificial-intelligence-artists.html.
[28] Théâtre D’opéra Spatial, supra note 26, at 6.
[29] Id. at 6.
[30] Id. at 2.
[31] Id. at 2.
[32] Id. at 6.
[33] USCO Report, supra note 5, at Executive Summary iii, bullet 6.
[34] Théâtre D’opéra Spatial, supra note 26, at 5.
[35] USCO Report, supra note 5, at Executive Summary iii, bullet 7.
[36] Théâtre D’opéra Spatial, supra note 26, at 5. (Holding that “the Board would need more information to know whether Allen’s use of Photoshop rose to the level of copyrightability.”)
[37] Copyright Review Board, Re: Zarya of the Dawn Registration # VAu001480196 (2023), https://www.copyright.gov/docs/zarya-of-the-dawn.pdf.
[38] Id.
[39] Id. at 1.
[40] Id. (“The images in the Work that were generated by the Midjourney technology are not the product of human authorship.”).
[41] USCO Report, supra note 5, at 27.
[42] Id. at Executive Summary iii, bullet 3.
[43] Id. at Executive Summary iii, bullet 7.
[44] Copyright Review Board, Second Request for Reconsideration for Refusal to Register SURYAST SR # 1-11016599571; Correspondence ID: 1-5PR2XKJ, (2023).
[45] Id. at 2.
[46] Id. at 3.
[47] Id. at 3.
[48] Id. at 7.
[49] USCO Report, supra note 5, at Executive Summary iii, bullet 4.
[50] Id. at Executive Summary iii, bullet 2.
[51] Id. at Executive Summary iii, bullet 7.
[52] SURYAST, supra note 44, at 2.
[53] United States Copyright Catalog, Rose Enigma Registration No. VAu001528922 (2023).
[54] USCO Report, supra note 5, at 23.
[55] Id.
[56] Id.
[57] Id. at 24.
[58] Id. at Executive Summary iii, bullet 7.
[59] Invoke, How We Received the First Copyright for a Single Image Created Entirely With AI-Generated Material, (2025), https://44037860.fs1.hubspotusercontent-na1.net/hubfs/44037860/Invoke-First-Copyright-Image-AI-Generated-Material-Report.pdf (PowerPoint presentation).
[60] Id.
[61] Id.
[62] USCO Report, supra note 5, at Executive Summary iii, bullet 7.
[63] Id.
[64] Id. at Executive Summary iii, bullet 5.
[65] Campbell v. Acuff-Rose Music, Inc., 510 U.S. 569, 577 (1994).
[66] Id.
[67] Id.
[68] One key distinction between the fair use analysis applied by courts and the human authorship standard used by the USCO in registration decisions is that the USCO’s determination to register a particular work carries no precedential value and is not binding in future applications: U.S. Copyright Office, Copyrightable Authorship: What Can Be Registered, Compendium of U.S. Copyright Office Practices, 3d ed, ch. 300 at 11. In contrast, judicial interpretations carry precedential weight. Nonetheless, a principled approach remains valuable in both contexts, as it provides a degree of transparency.
[69] Johannes Fritz, Understanding Authorship in Artificial Intelligence-Assisted Works, 20 J. Intell. Prop. L. & Prac. 5, 354–64, (May 2025), https://doi.org/10.1093/jiplp/jpae119.
[70] USCO Report, supra note 5, at Executive Summary iii, bullet 6.
[71] Fritz, supra note 69, at 6.
[72] USCO Report, supra note 5, at Executive Summary iii, bullet 4.
[73] United States Copyright Office, Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence, 88 Fed. Reg. 16190 (March 16, 2023) (Statement of Policy), 37 C.F.R. Part 202.
[74] USCO Report, supra note 5, at Executive Summary iii, bullet 2.
[75] Fritz, supra note 69.
[76] USCO Report, supra note 5, at 12.
[77] United States Copyright Office, Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence, 88 Fed. Reg. 16190 (March 16, 2023) (Statement of Policy), 37 C.F.R. Part 202.
[78] USCO Report, supra note 5, at Executive Summary iii, bullet #6.
[79] These considerations are also considered in Japan’s human authorship analysis: Legal Subcommittee under the Copyright Subdivision of the Cultural Council, General Understanding on AI and Copyright in Japan, Japan Copyright Office (May 2024), https://www.bunka.go.jp/english/policy/copyright/pdf/94055801_01.pdf; cited in USCO Report, supra note 5, at 28.
[80] USCO Report, supra note 5, at Executive Summary iii, bullet 7.
[81]Théâtre D’opéra Spatial, supra note 26, at 5 (holding that “the Board would need more information to know whether Allen’s use of Photoshop rose to the level of copyrightability.”).
[82] Invoke, How We Made A Single Piece of American Cheese, Vimeo (Feb. 7, 2025) https://vimeo.com/1054656471.
[83] Rachel Metz, AI Won an Art Contest, and Artists Are Furious, CNN Business (Sept. 3, 2022), https://www.cnn.com/2022/09/03/tech/ai-art-fair-winner-controversy/index.html.
[84] See e.g., Invoke, How We Received the First Copyright for a Single Image Created Entirely with AI-Generated Material (2025), https://44037860.fs1.hubspotusercontent-na1.net/hubfs/44037860/Invoke-First-Copyright-Image-AI-Generated-Material-Report.pdf (endorsing the application Invoke Provenance Records, which tracks online image editing).
[85] Richard M. Assmus, Brian W. Nolan & Megan P. Fitzgerald, Supreme Court Denies Cert in AI Authorship Case, Mayer Brown (March 10, 2026), https://www.mayerbrown.com/en/insights/publications/2026/03/supreme-court-denies-review-in-ai-authorship-case.
[86] Twentieth Century Music Corp. v. Aiken, 422 U.S. 151, 156 (1975).
