Introduction
Generative AI has brought a longstanding copyright issue to a head: when does an individual who uses a new technology remain the author of the expression produced, and when does the technology itself become the source of expression that copyright does not protect? The answer lies in the principle of “human authorship” in United States copyright law. The challenge is applying that principle to modern generative systems, whose output may be shaped by prompting, reference material, iterative choice, editing, arrangement and post-generation modification.1
The U.S. Copyright Office has responded in several ways, including through registration guidance and its Copyright and Artificial Intelligence initiative, which produced Part 2: Copyrightability. The Office’s approach is not hostile to the use of AI in creative production. Rather, it asks which expressive features are sufficiently the work of a human creator, and it protects only those. This distinction determines where United States copyright doctrine draws the line between AI assistance and AI substitution.2
The challenge is to determine which elements of an AI-assisted work are unprotectable ideas and instructions and which are protectable expression originating from a human being, and whether the existing registration and case-by-case processes offer adequate clarity.3 This paper asks how United States copyright law applies the human-authorship requirement to generative-AI output. It is a doctrinal study of the U.S. Copyright Office’s Part 2 report and its 2023 registration guidance, read with the constitutional and statutory principles and judicial rulings on human authorship and originality, and it is confined to copyrightability, leaving aside the larger questions of model training and liability.
Literature review
Edward Lee argues that the Copyright Office’s 2023 registration guidance, which excludes AI-generated material from copyright, is wrong: in his view its insistence on the so-called “traditional elements of authorship” understates the creative control that human users can exercise over prompt-based outputs, and he proposes instead a “bare minimum” standard of authorship that many, though not all, prompt-engineered works would meet. Lee observes that Professor Pamela Samuelson advanced a comparable approach in a joint comment with Christopher Jon Sprigman and Matthew Sag, submitted in the Office’s public consultation. That comment, which argued that sophisticated prompts specifying the details of an image should be sufficient to meet the human-authorship requirement, is among the submissions the Report engages with directly. Such consultations have become increasingly common, and they give scholarship a channel through which to respond to the Office’s doctrinal positions.4
The literature reveals a central debate about the extent of user control in generative-AI interfaces. Critics argue that no categorical rule is possible, because sufficiently specific instructions can embody creative conception, while the Copyright Office has stressed the difference between prompting a system and actually generating the expressive elements of the resulting work. This debate frames the paper’s discussion of existing doctrine and registration practice.5
Although the human-authorship principle itself is well established, its application to current generative workflows is highly fact-dependent. Existing authorities do not reduce the human role to a measurable variable, and neither the length of a prompt nor the number of generations is a reliable indicator of authorship.6
The U.S. Copyright Office’s Part 2 report on copyrightability
A. Nature and structure of the report
The U.S. Copyright Office’s report Copyright and Artificial Intelligence, Part 2: Copyrightability, issued in January 2025, is a sustained inquiry into whether, and under what conditions, AI-generated output is copyrightable. It proceeds from governing copyright principles and then examines different forms of human contribution, including prompting, human-authored expressive inputs, selection and arrangement, modification, and assistive uses of AI. The Report repeatedly emphasises that copyrightability remains a case-by-case inquiry rather than a predetermined rule based solely on the use of generative technology.7
The Report responds to the Office’s Notice of Inquiry, which drew more than 10,000 comments, approximately half of which addressed copyrightability. Commenters differed sharply on what types and amounts of human contribution to a generative-AI output could qualify for copyright protection under current law, and, as a matter of policy, some argued that extending protection to AI-generated material would encourage the creation of more works and so further copyright’s constitutional mission of fostering progress in culture and knowledge.8 Recognising that copyrightability is determined case by case, the Office sets out the governing legal principles throughout the Report and assesses their application to AI-generated content.
The Report is organised in four substantive sections, followed by a conclusion. Section I identifies the copyrightability issues raised by AI technologies, places them in the history of copyright’s adaptation to earlier technological advances, and describes the Office’s broader AI initiative. Section II summarises the relevant technology and the existing legal framework, including the human-authorship requirement, the idea-expression dichotomy and the originality standard, and then examines how those principles apply to assistive uses of AI and to different kinds of human contribution: prompting, human-authored expressive inputs, and the modification or arrangement of AI-generated output. Section III surveys the international landscape, describing how other jurisdictions are approaching the same copyrightability questions under their own law. Section IV addresses the policy implications of granting additional legal protection to AI-generated material and evaluates the arguments for and against legislative change.9
B. The Office’s conclusions
The Report’s conclusions form the core of this part of the analysis. First, questions of copyrightability and AI can be resolved under existing law, without legislative change. Second, 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. Third, copyright protects the original expression in a work created by a human author, even if the work also includes AI-generated material. Fourth, copyright does not extend to purely AI-generated material, or to material over whose expressive elements there was insufficient human control. Fifth, whether human contributions to an AI-generated output are sufficient to constitute authorship must be analysed case by case. Sixth, given the functioning of current generally available technology, prompts alone do not provide sufficient control. Seventh, human authors are entitled to copyright in their works of authorship that are perceptible in AI-generated outputs, as well as in the creative selection, coordination or arrangement of material in the outputs, or in creative modifications of the outputs. Eighth, and finally, the case has not been made for additional copyright or sui generis protection for AI-generated content.10
The overarching question the Report considers is whether AI outputs, which increasingly resemble traditionally protected creative works, can themselves be copyrightable, both on their own and when incorporated into a larger work, and, crucially, how much human contribution is required. The Office’s answer is that this turns on the nature and extent of the human contribution and on whether it qualifies as authorship of the expressive elements contained in the output.11
C. Human authorship, originality, and precedent
Copyright rests on the principle that a human must be involved in creating the work. That principle is grounded in the Copyright Clause of the U.S. Constitution, which empowers Congress to secure to “Authors” the exclusive right to their “Writings”, and in the Copyright Act, which protects “original works of authorship”. The Supreme Court has explained that the author is “the person who translates an idea into a fixed, tangible expression entitled to copyright protection”, and no United States court has recognised copyright in material created by a non-human. The requirement goes back at least as far as Burrow-Giles Lithographic Co. v. Sarony, where the Supreme Court, faced with a constitutional challenge to Congress’s extension of copyright protection to the then-new technology of photography, defined an “author” as “he to whom anything owes its origin; originator; maker” and described copyright as “the exclusive right of a man to the production of his own genius or intellect”. The camera did not eliminate the need for a human author; the Court asked instead whether the photograph reflected the photographer’s own creative choices. The same question must now be asked of generative AI systems.12
The Report also draws on the law of joint authorship as a helpful, though not directly applicable, analogy for deciding whether a party contributed enough expression to be regarded as the author of an AI-assisted work. In Community for Creative Non-Violence v. Reid, the Supreme Court held that the sculptor, who sketched the design and executed his creative vision in tangible form, was an author, and in the same litigation the D.C. Circuit, while recognising that copyright does not protect ideas, treated the commissioning organisation, which had conceived the sculpture in detail, guided its execution and contributed its steam-grate pedestal, as a possible joint author and remanded that question; a person who merely describes what a commissioned work should do or look like, however, is not a joint author. By contrast, in Andrien v. Southern Ocean County Chamber of Commerce, the Third Circuit held that a plaintiff who had expressly directed a printer, in specific detail, in preparing a compilation of maps was its author, because the printer did not change the substance of his original expression; where one person authorises another to embody the expression, “that process must be rote or mechanical transcription that does not require intellectual modification or highly technical enhancement”.13
In 2023 the U.S. District Court for the District of Columbia, in Thaler v. Perlmutter, became the first court to address specifically the copyrightability of AI-generated output, and it ruled in favour of the Copyright Office; the Report relies on that decision. The plaintiff had sought to register an image described in his application as “autonomously created by a computer algorithm running on a machine”. Affirming the Office’s refusal, the court held that “copyright law protects only works of human creation”, that “[h]uman authorship is a bedrock requirement of copyright”, and that “[c]opyright has never stretched so far . . . as to protect works generated by new forms of technology operating absent any guiding human hand”. The Report also relies on Feist Publications, Inc. v. Rural Telephone Service Co., in which the Supreme Court rejected the theory that “sweat of the brow” (effort or investment of time) alone can sustain a claim to copyright, while stressing that the requisite level of creativity is extremely low: “[t]he vast majority of works make the grade quite easily, as they possess some creative spark”.14
The decision was appealed. In March 2025 the U.S. Court of Appeals for the District of Columbia Circuit affirmed the refusal to register the autonomously generated image, holding that the Copyright Act of 1976 “requires all eligible work to be authored in the first instance by a human being”. The court decided the case on statutory grounds and expressly declined to address the Copyright Office’s argument that the Constitution itself requires human authorship. The Supreme Court denied certiorari in March 2026.15
D. Categories of human contribution to AI-generated output
The Report’s most significant analytical contribution is its careful consideration of the various ways in which a human can participate in producing the output of a generative AI system, and of whether each can satisfy the authorship requirement.16
Prompting alone. Commenters disagreed sharply about whether a sufficiently detailed prompt can amount to authorship. Some argued that if a prompt were detailed enough, the resulting AI-generated work could be eligible for copyright. Professors Pamela Samuelson, Christopher Jon Sprigman and Matthew Sag, for example, submitted that a person who instructs a generative AI system with enough detail, such that the output reflects that person’s original conception of the work, should be regarded as the author of the resulting work. Other commenters, such as the Authors Guild, stressed the unpredictability of the prompt-to-output process, which makes it difficult to show sufficient closeness between conception and execution, while Adobe pointed out that the output a user receives is based solely on the AI’s interpretation of the prompt, and that the AI’s expression of the user’s idea is not copyrightable. Ultimately, the Office concluded that, given current generally available technology, prompts alone do not provide sufficient human control to make users the authors of the output: prompts essentially function as instructions that convey unprotectable ideas, rather than a mechanism by which a user controls a work’s expression. The Report also notes that identical prompts can generate multiple different outputs, which further indicates that the user does not control the expressive output through prompts alone. Repeatedly revising and resubmitting a prompt does not change this analysis, however much work is involved, because copyright protects original authorship, not effort; the question is control over protectable expression, not the amount of work or the number of iterations.17
The Report also addresses, and rejects, a theory that some commenters described as “authorship by adoption”, under which a user’s creative judgment in choosing among several AI-generated outputs would be enough to confer authorship on the chosen output. The Authors Guild likened repetitive prompting to “spinning a roulette wheel with infinite possibilities”, and the Office agreed that providing instructions to a machine and selecting an output does not amount to authorship.18
Selection, coordination, arrangement and modification. By contrast, the Office concluded that human authors can claim copyright in their own contributions to an AI-generated output, such as the creative selection, coordination or arrangement of material, or creative modifications of the output, an exercise the Office compared to the protection of derivative works, in which copyright is limited to the material contributed by the later author. Commenters argued, and the Office accepted, that a person who supplies his or her own expression as the input to an AI system, such as an illustration or a media file, stands on stronger ground than a person who starts from a loose prompt, because a specified starting point constrains the autonomy of the output; where that human-authored input is perceptible in the output, its author is the author of at least that portion of the output. In one concrete example discussed in the Report, the comic book Zarya of the Dawn, the Office found that the selection of the AI-generated images and the placement and arrangement of the images and text on each page reflected sufficient creativity under Feist. The registration was accordingly limited to the human-authored text and to the selection, coordination and arrangement of the work’s written and visual elements, and did not extend to the images generated by Midjourney themselves.19
E. AI as an assistive tool
One of the Report’s most prominent illustrations of the assistive-tool principle concerns AI tools used by creators with disabilities. Commenters argued that extending protection to AI-generated works would empower such creators, but the specific uses they identified involved AI as a tool assisting human creation rather than as a source of expressive content, and the Office confirmed that copyright protection remains available where AI is used in that way. The Report gives the principle concrete form through a sound recording by the country singer Randy Travis, whose speech has been limited since a stroke: the track was created from a recording of a human voice using a special-purpose AI vocal model, and the Office registered it because AI had been used as a tool to assist the human creative process rather than to generate the expression. The example shows in practice the boundary that runs through the whole Report: AI-assisted human creativity can be copyrightable, but AI-generated expression over which humans lack sufficient control is not.20
F. The international comparative landscape
Section III examines how other jurisdictions have dealt with copyrightability, and the comparison sheds light on the international treatment of human authorship. In 2023 South Korea’s Ministry of Culture, Sports and Tourism and the Korea Copyright Commission issued guidance stating that copyright registration for an AI output is impossible unless a human contributed creatively to its expressive form; where a human modifies, adds to or deletes from an AI output, only the changed part is copyrightable, and a creative selection and arrangement of AI output may be registered as a compilation. In May 2024 the Copyright Subdivision of the Cultural Council in Japan published guidance identifying four factors: the amount and content of the instructions and input prompts provided by the user; the number of generation attempts; the user’s selection from multiple outputs; and any subsequent human additions or corrections to the output.21
In the People’s Republic of China, the Beijing Internet Court reached a more permissive result. In November 2023 it held that an image created using Stable Diffusion was protected by copyright and that the person who used the AI system to generate it was its author, finding that his selection of over 150 prompts, combined with subsequent adjustments and modifications, showed that the image reflected his intellectual achievement and personalised expression. The United Kingdom offers an entirely different model, because its copyright statute already provides for works generated by computer in circumstances where there is no human author: the author is taken to be “the person by whom the arrangements necessary for the creation of the work are undertaken”. In 2021 the UK Intellectual Property Office consulted on whether to change that rule in light of generative AI but, given the absence of case law applying it to AI and because it was unclear whether removing the protection would promote or discourage innovation, decided to leave the law in place; it has since opened a further consultation on copyright and AI. In Canada, a 2021 review of the Copyright Act acknowledged a lack of clarity about the authorship of AI-generated works, and in Australia participants in 2024 consultations held by the parliamentary Select Committee on Adopting Artificial Intelligence raised similar concerns about the extent of copyright protection, if any, afforded to works created with the assistance of AI.22
The Office’s overall reading of the international situation is that some consensus on the need for human authorship appears to be emerging, but that views are still forming and there are significant differences in how computer-generated works and the degree of human involvement are treated. The Report is clear, however, that foreign approaches do not determine the domestic outcome: whatever other countries conclude, the United States is bound by its own Constitution and copyright principles. Comparative practice can therefore inform the analysis but cannot substitute for the domestic human-authorship inquiry.23
G. Policy analysis and legislative reform
The Report also considers a policy question: should AI-generated material receive additional legal protection? Section IV, which draws on the incentive rationale at the heart of the Copyright Clause, addresses this issue. Commenters were divided on how the constitutional purpose of promoting progress should be understood in the context of AI: some argued that the Copyright Clause should be read flexibly to accommodate new technologies, while most of those who addressed the issue emphasised that the Clause provides incentives to human authors, not to inanimate AI systems.24
The Office agreed with the latter position. It observed that the developers of AI models and systems already enjoy meaningful incentives under existing law, including patent, copyright and trade-secret protection for the machinery and software, as well as potential funding and first-mover advantages, and that, given the rapid development and adoption of generative AI even without copyright in its raw output, it is not clear that new incentives are needed. The Report further reasoned that protecting purely AI-generated content could prove counterproductive: if a flood of easily and rapidly generated AI content were to drown out human-authored works in the marketplace, the resulting loss of income for human authors could reduce human creative output, so that additional protection would undermine rather than advance the goals of the copyright system. The Report adds that, although AI systems can produce results comparable to human creative work, AI training itself is reportedly reliant on human-generated content, with synthetic training data leading to lower-quality results.25
On legislative reform, the Office acknowledged the strong interest among stakeholders in greater clarity but concluded that legislation is not necessary at this point. Much of the concern expressed related to the assistive use of AI tools, which the Report confirms does not undermine protection; the courts will provide further guidance on the human-authorship requirement as it applies to specific uses of AI; and the Report’s own analysis can clarify how existing principles apply. Because the copyrightability inquiry requires analysis of each work and the circumstances of its creation, statutory language would in any event be limited in its ability to draw brighter lines. The Office concluded that existing legal doctrines are adequate and appropriate to resolve questions of copyrightability: outputs will often be copyrightable in whole or in part where AI is used as a tool and where a human has been able to determine the expressive elements they contain, but prompts alone are, at this stage, unlikely to satisfy that standard. The Office will continue to monitor technological and legal developments to evaluate any need for a different approach.26
Critical evaluation
The main strength of the U.S. framework is its conceptual clarity. It does not treat the use of AI as disqualifying in itself, nor does it recognise a fictional machine author. Rather, it protects human expression that is discernibly present in the final result, whether as perceptible human-authored material or as qualifying selection, coordination, arrangement or modification, but not AI-generated expression alone. This maintains continuity with established copyright doctrine and prevents originality from being inferred merely because a person has invested time, money and repeated effort in operating a system.27
Its primary weakness is uncertainty at the margins. Generative workflows are iterative, and creative contribution can be spread across prompts, the choice of reference inputs, selection among outputs, compositional choices and subsequent edits. A rule that categorically excluded prompts would be too rigid, while a rule that equated detailed instructions with authorship would risk collapsing the idea-expression distinction. The Office’s case-by-case approach is defensible, and it places considerable weight on documentation of the creative process.28
This uncertainty can be mitigated through sound registration practice that transparently identifies the human-authored components of a work and discloses AI-generated content as required. Such practice provides a workable framework for mixed works, such as a human-made compilation or modification of machine-generated material, in which protection extends to the human-made elements but not to the machine-made ones. Continuing guidance will nonetheless be needed as systems become more controllable and as the technology linking prompts to expression evolves.29
Suggestions
• Preserve the human-authorship requirement: maintain copyright as a safeguard for human-generated expression, and refrain from granting automatic copyright to purely AI-generated content.
• Refine registration guidance for mixed works: give more specific examples of how applicants can identify and disclaim AI-generated content while claiming the text, images, arrangement, selection or modification created by humans.
• Treat prompting as evidence, not as a substitute for authorship: assess prompts as part of the creative process, without treating prompt detail, length or repetition as establishing control over the expressive output.
• Encourage process documentation: where authorship is disputed, evaluate evidence of human contributions, such as multiple drafts, creative choices, edits and changes, that shows control over the expression.
• Protect legitimate assistive uses: distinguish between AI as a tool that facilitates human expression and AI as the source of what is expressed.
• Continue technology-sensitive review: revise guidance to reflect any increase in the determinacy of, or control over, the output of generative systems, without abandoning the case-by-case approach to copyrightability.
Conclusion
The United States has a clear baseline position on the copyrightability of generative-AI content: copyright requires human authorship, and the use of generative AI is not in itself a bar to copyrightability. The critical question is which, if any, of the expressive elements can be attributed to a human author. Purely AI-generated content is not protected by copyright, whereas content written by a person, AI material that a person has creatively selected and arranged or substantially modified, and works made with AI as an assistive tool may be protected.30
The U.S. Copyright Office’s Part 2 report gives this principle practical form. With current technology, prompts alone do not provide sufficient control over the resulting work; copyright may nevertheless subsist in human expression that is perceptible in a mixed work, and in creative decisions that independently meet the originality standard. Thaler v. Perlmutter establishes that a work generated autonomously by an AI system, with no human author, is not copyrightable, confirming that the human-authorship baseline is firmly in place, although the more difficult factual questions raised by genuinely collaborative human-AI processes remain unresolved.31
If the U.S. approach is to endure, it will require continuity of doctrine and greater evidentiary clarity. Registration guidance should therefore be more specific about mixed works and about documenting the creative process, while avoiding formulas that treat the number of prompts or the complexity of the process as a proxy for authorship. Even as technologies continue to evolve, a human-centred, case-specific framework can accommodate technological change without granting copyright to machine-generated expression as such.32
*****
Footnotes
1. Robin Feldman, AI Versus IP: Rewriting Creativity 61 (2025).
2. U.S. Copyright Off., Copyright and Artificial Intelligence, Part 3: Generative AI Training (May 9, 2025) (pre-publication version), https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-3-Generative-AI-Training-Report-Pre-Publication-Version.pdf; see also Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence, 88 Fed. Reg. 16,190 (Mar. 16, 2023) [hereinafter AI Registration Guidance].
3. Christopher T. Zirpoli, Cong. Rsch. Serv., LSB10922, Generative Artificial Intelligence and Copyright Law 1 (Sept. 29, 2023).
4. Edward Lee, Prompting Progress: Authorship in the Age of AI, 76 Fla. L. Rev. 1445, 1445, 1452 n.16 (2024); Comment Letter from Pamela Samuelson, Christopher Jon Sprigman & Matthew Sag to U.S. Copyright Off., Docket No. COLC-2023-0006, at 3 (Oct. 30, 2023), https://downloads.regulations.gov/COLC-2023-0006-8854/attachment_1.pdf.
5. Lee, supra note 4.
6. Ryan Abbott, The Reasonable Robot: Artificial Intelligence and the Law 118 (2020).
7. U.S. Copyright Off., Copyright Office Releases Part 2 of Artificial Intelligence Report, NewsNet Issue No. 1060 (Jan. 29, 2025), https://www.copyright.gov/newsnet/2025/1060.html.
8. U.S. Copyright Off., Copyright and Artificial Intelligence, Part 2: Copyrightability ii, 4 (Jan. 29, 2025) [hereinafter Part 2 Report], https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf.
9. Copyright and Artificial Intelligence, U.S. Copyright Off., https://www.copyright.gov/ai/ (last visited Sept. 15, 2026); Part 2 Report, supra note 8, at i–ii.
10. Part 2 Report, supra note 8, at iii; see also Daniel J. Gervais, The Machine as Author, 105 Iowa L. Rev. 2053 (2020).
11. Part 2 Report, supra note 8, at 1; see also Abbott, supra note 6, at 154.
12. U.S. Const. art. I, § 8, cl. 8; 17 U.S.C. § 102(a); Cmty. for Creative Non-Violence v. Reid, 490 U.S. 730, 737 (1989); Naruto v. Slater, 888 F.3d 418, 426 (9th Cir. 2018) (animals lack statutory standing under the Copyright Act); Burrow-Giles Lithographic Co. v. Sarony, 111 U.S. 53, 57–58, 60 (1884); Part 2 Report, supra note 8, at 7–9.
13. Cmty. for Creative Non-Violence, 490 U.S. at 751–53; Cmty. for Creative Non-Violence v. Reid, 846 F.2d 1485, 1497 (D.C. Cir. 1988), aff’d, 490 U.S. 730 (1989); Andrien v. S. Ocean Cnty. Chamber of Com., 927 F.2d 132, 134–36 (3d Cir. 1991); 17 U.S.C. § 101 (defining “joint work”); Robert A. Gorman et al., Copyright: Cases and Materials 55 (10th ed. 2024); Part 2 Report, supra note 8, at 9–11.
14. Thaler v. Perlmutter, 687 F. Supp. 3d 140, 142–43, 146 (D.D.C. 2023); Feist Publ’ns, Inc. v. Rural Tel. Serv. Co., 499 U.S. 340, 345, 352–61 (1991); Part 2 Report, supra note 8, at 8–9; Gervais, supra note 10, at 2074.
15. Thaler v. Perlmutter, 130 F.4th 1039 (D.C. Cir. 2025), cert. denied, No. 25-449 (U.S. Mar. 2, 2026).
16. Part 2 Report, supra note 8, at 11–27; see also Jane C. Ginsburg & Luke Ali Budiardjo, Authors and Machines, 34 Berkeley Tech. L.J. 343 (2019).
17. Samuelson et al., supra note 4, at 3; Part 2 Report, supra note 8, at 12–20.
18. Part 2 Report, supra note 8, at 17–18, 21; see also Francesca Mazzi, Authorship in Artificial Intelligence-Generated Works: Exploring Originality in Text Prompts and Artificial Intelligence Outputs Through Philosophical Foundations of Copyright and Collage Protection, 27 J. World Intell. Prop. 410 (2024), https://doi.org/10.1111/jwip.12310.
19. Part 2 Report, supra note 8, at 22–25; Letter from Robert J. Kasunic, Assoc. Reg. of Copyrights & Dir. of Registration Pol’y & Prac., U.S. Copyright Off., to Van Lindberg, Taylor English Duma LLP, Re: Zarya of the Dawn (Registration # VAu001480196), at 1, 5 (Feb. 21, 2023), https://www.copyright.gov/docs/zarya-of-the-dawn.pdf; see also Ginsburg & Budiardjo, supra note 16, at 365.
20. Part 2 Report, supra note 8, at 11–12, 37–38 (discussing assistive uses and the registration of the Randy Travis sound recording).
21. Part 2 Report, supra note 8, at 28; see also Gorman et al., supra note 13, at 230.
22. Part 2 Report, supra note 8, at 28–31; Copyright, Designs and Patents Act 1988, c. 48, § 9(3) (UK); see also Peter S. Menell et al., Intellectual Property in the New Technological Age, Vol. II: Copyrights, Trademarks and State IP Protections 65 (2025).
23. Part 2 Report, supra note 8, at 31, 39; World Intell. Prop. Org., Generative AI: Navigating Intellectual Property (2024).
24. Part 2 Report, supra note 8, at 32–33; see also Neil Weinstock Netanel, Copyright: What Everyone Needs to Know 84 (2018).
25. Part 2 Report, supra note 8, at 34–37; see also Netanel, supra note 24, at 99.
26. Part 2 Report, supra note 8, at 39–41; see also Menell et al., supra note 22, at 210.
27. Daryl Lim, Generative AI and Copyright: Principles, Priorities and Practicalities, 18 J. Intell. Prop. L. & Prac. 841 (2023), https://doi.org/10.1093/jiplp/jpad081.
28. Mazzi, supra note 18, at 421.
29. Artificial Intelligence Study, U.S. Copyright Off., https://www.copyright.gov/policy/artificial-intelligence/ (last visited Sept. 16, 2026); AI Registration Guidance, supra note 2, at 16,193.
30. Artha Dermawan, AI v Copyright: How Could Public Interest Theory Shift the Discourse?, 19 J. Intell. Prop. L. & Prac. 55 (2024), https://doi.org/10.1093/jiplp/jpad111.
31. U.S. Copyright Off., Copyright Office Releases Part 2 of Artificial Intelligence Report, supra note 7; Thaler v. Perlmutter, 130 F.4th 1039 (D.C. Cir. 2025), cert. denied, No. 25-449 (U.S. Mar. 2, 2026); see, e.g., Complaint, Allen v. Perlmutter, No. 1:24-cv-02665 (D. Colo. Sept. 26, 2024) (pending challenge to the refusal to register Théâtre D’opéra Spatial, an image generated through at least 624 prompts and then edited in Adobe Photoshop); U.S. Copyright Off. Review Bd., Decision Affirming Refusal of Registration of Théâtre D’opéra Spatial (Sept. 5, 2023), https://www.copyright.gov/rulings-filings/review-board/docs/Theatre-Dopera-Spatial.pdf.
32. Dermawan, supra note 30, at 61.