Recent Developments in AI, Art & Copyright: Copyright Office Report & New Registrations

March 4, 2025

By Atreya Mathur

In January 2025, the U.S. Copyright Office released Part 2 of its report, Copyright and Artificial Intelligence: Copyrightability (“the 2025 Report”) providing a detailed legal and policy analysis of how copyright law applies to AI-generated content.^[1] Part 2 builds on foundational principles of copyright law, reaffirming that human authorship remains the cornerstone of copyright protection in the United States.^[2] It provides critical guidance on the conditions under which AI-assisted works may qualify for copyright, clarifying the legal boundaries between human creativity and automated generation.^[3]

In August 2023, in response to the U.S. Copyright Office’s Notice of Inquiry on AI and copyright law, Center for Art Law submitted a public comment where it addressed key concerns such as the use of copyrighted works in AI training, transparency and disclosure requirements, and the legal status of AI-generated outputs. Together with over 10,000 other submissions, the Center stressed best practices for policymakers, lawyers, technology and AI companies, and artists when generating and using AI art.

Practical Challenges in Applying The 2025 Report’s Framework

A key takeaway from The 2025 Report is the Office’s categorical rejection of copyright protection for works generated solely by AI, reinforcing the long-established principle that copyright law protects only “original works of authorship” created by humans.^[4] The 2025 Report reiterates that AI-generated outputs, absent meaningful human creative input, lack the necessary authorship required for protection under the Copyright Act. This conclusion aligns with existing case law and administrative decisions, including recent Copyright Office rulings denying registration for purely AI-generated works.^[5]

Beyond addressing fully AI-generated outputs, the 2025 Report also examines hybrid authorship scenarios, where AI tools assist human creators. The Copyright Office emphasizes that for a work to qualify for protection, creative human involvement must be substantial, demonstrable, and independently copyrightable. The mere use of AI does not preclude copyright eligibility, but the human contribution must extend beyond basic prompts or trivial modifications. The 2025 Report distinguishes between varying levels of human involvement through examples:

• Minimal Human Input: If an artist enters a simple text prompt into an AI system like Midjourney or DALL·E and accepts the resulting image without significant modification, the work does not qualify for copyright protection. The human input in such cases is deemed insufficient to satisfy the originality requirement.
• Substantial Human Creativity: A creator, such as a digital artist who selects, edits, and arranges AI-generated elements in a way that reflects creative judgment, may claim copyright protection over the resulting work. This applies when the human contribution shapes the final product in a meaningful way, demonstrating originality. However, only the selection and arrangement is protected and any individual AI-generated components do not receive protection.
• Collaborative Creation: If AI is used as a tool within a broader creative process, such as an artist generating AI-based sketches and then painting over them or digitally compositing elements into a larger work, the human-authored portions may be eligible for protection. This mirrors traditional digital art techniques where artists combine stock images, digital brushes, or 3D renders with original hand-drawn elements.

The 2025 Report acknowledges the rapid evolution of AI technologies and their increasing role in creative industries, emphasizing the need to balance technological innovation with the fundamental principles of human authorship in copyright law. While existing legal frameworks provide a foundation for addressing AI-generated works, the Copyright Office recognizes that ongoing evaluation and potential legislative refinements may be necessary to address novel challenges posed by AI’s growing capabilities, but for now there is no need for any legislative reforms.^[6]

Legal and Practical Challenges in Applying The 2025 Report’s Framework

While the 2025 Report provides clear guidance on human authorship as the basis for copyright protection, it also raises several legal and practical questions regarding the real-world application of these principles.

Ambiguities in Defining “Sufficient Human Creativity”

One of the 2025 Report’s key takeaways is that AI-assisted works may qualify for copyright protection only if a human’s contribution is substantial and independently copyrightable.^[7] However, the line between trivial modifications and meaningful human authorship remains unclear. The 2025 Report provides examples, but in practice, how much human input is “enough” to satisfy copyrightability? Would slight modifications to an AI-generated text or image—such as adjusting composition, rewording, or adding a personal style—qualify, or must there be a significant transformative effort? This leaves room for inconsistent interpretations in copyright registration and potential litigation and will be done on a case-to-case basis.

Impact on Creative Industries and AI-Generated Works

The 2025 Report implicitly signals that many AI-generated works will be in the public domain, as they lack human authorship.^[8] While this upholds copyright’s traditional emphasis on human creativity, it also creates uncertainty for businesses and creators who rely on AI tools. For instance, companies using AI for content generation—such as publishers, marketing firms, and game developers—must carefully assess whether their outputs qualify for copyright protection. This may lead to increased reliance on contractual agreements—such as licensing terms that restrict how AI-generated content can be used or redistributed—and trade secrets, where businesses keep proprietary AI models, datasets, or workflows confidential to maintain a competitive edge. For instance, a game developer might use contracts to limit third-party use of AI-generated assets, while a marketing firm might protect AI-generated ad copy by keeping its generation process and datasets confidential rather than seeking copyright protection.

Legislative and Policy Gaps

Although The 2025 Report suggests that current copyright laws are flexible enough to address AI issues, it does not fully resolve certain policy questions: Who owns AI-assisted works that involve multiple contributors? If an AI tool refines a designer’s rough sketches, does the final work belong solely to the designer, or does the AI system’s operator (e.g., a company licensing the AI) have a claim?; What happens when AI-generated works infringe existing copyrights? The 2025 Report does not propose specific enforcement mechanisms for cases where AI-generated content closely mimics human-created works, raising concerns about potential loopholes in infringement claims. However, this will likely be discussed in Part 3 of The 2025 Report which will be published later in 2025.^[9]

Finally, by granting copyright only to the human-authored portions of a hybrid work, the current approach may create loopholes that allow others to replicate AI-generated elements without infringing. Since AI-generated content is effectively in the public domain, a third party could extract, modify, or reassemble these unprotected elements while avoiding direct copying of the protected human contributions. This raises questions about the enforceability of copyright in hybrid works, as the threshold for substantial similarity may be harder to establish.

Registration of Recent Works using Artificial Intelligence

While the Copyright Office has reaffirmed that purely AI-generated works lack the human authorship necessary for copyright protection, its handling of hybrid AI-human creations provides a crucial framework for visual artists navigating AI-assisted tools. The Office has taken a case-by-case approach in determining whether a work contains sufficient human creativity to warrant registration. Analyzing recent registrations granted to works incorporating AI-generated material evaluates creative input in the visual arts and offers a clearer understanding of how the Copyright Office interprets human authorship, the thresholds set for creative input, and the legal uncertainties that remain.

“A Collection of Objects Which Do Not Exist”

In January 2025, the U.S. Copyright Office registered A Collection of Objects Which Do Not Exist, (very aptly titled) as a “collage, selection and arrangement” granting protection to the work as visual art.^[10] The registration specifically excluded “2-D artwork, images generated by artificial intelligence.”^[11] The artist employed AI tools to generate individual components and then exercised creative judgment in curating these elements into a cohesive visual composition. The work received protection as a collage but AI materials were excluded. A similar approach was also taken in the registration of A Single Piece of American Cheese, an AI-generated image registered by Invoke under the work-made-for-hire doctrine, based on selection, coordination, or arrangement of AI-generated elements.^[12] Invoke’s CEO, Kent Keirsey, successfully argued that his hands-on role—specifically, his use of ‘inpainting’ to modify and refine 35 different elements within the image—constituted sufficient creative input.^[13]

This development follows the precedent set by Zarya of the Dawn, a graphic novel created by Kristina Kashtanova using the AI tool Midjourney.^[14] In February 2023, the U.S. Copyright Office granted limited copyright protection to this work, recognizing Kashtanova’s authorship of the text and the selection, coordination, and arrangement of the written and visual elements.^[15] However, the Office denied protection for the individual AI-generated images, citing the lack of human authorship in their creation.^[16] This distinction highlights the Office’s emphasis on human creativity in the selection and arrangement of AI-generated content as a basis for copyright eligibility. The policy clarifies that while AI-generated material alone is not eligible for copyright, works that involve human creativity in selecting or arranging such material can qualify for protection. ​

“Film clip for song Just Like In A Movie (SNEAK PREVIEW)”

In February 2025, the U.S. Copyright Office registered Film clip for song Just Like In A Movie (Sneak Preview) by Andrew John King, an AI-assisted music video featuring AI-generated visuals and music.^[17] While AI tools were used to generate these elements, King contributed original lyrics and played a significant role in editing, arranging, and integrating the AI-generated content into a cohesive final work.^[18] King was granted authorship in the “lyrics and editing of AI-generated footage.”^[19] Under U.S. copyright law, musical works—including composition (melody, harmony, rhythm) and lyrics—are protected as original expressions, rather than mere ideas or concepts.^[20] In this case, the Copyright Office determined that King’s human contributions, particularly in editing AI-generated elements and composing original lyrics, met the threshold for copyrightability. This decision once again emphasizes the Office’s stance that while AI-generated material itself is not eligible for protection, meaningful human authorship—evident through creative input and editorial decisions—remains a prerequisite for registration.

AI-Assisted Software Code

Under U.S. copyright law, computer programs are considered literary works, and the specific code written by a human author is protected as the original expression of an idea.^[21] This protection does not extend to the underlying ideas, algorithms, or functional aspects of the software. In the context of AI-assisted software, the human-authored portions of the code are eligible for copyright protection, while the AI-generated code, lacking human authorship, is excluded.^[22] Several software programs recently registered by the Copyright Office illustrate how it is distinguishing between human and AI-generated contributions.^[23]

IBM’s Watsonx Code Assistant (for VSCode and Eclipse) v. 1.0 received a copyright registration for its human-authored portions while excluding AI-generated source code, prior versions, and third-party contributions.^[24] This decision is consistent with the Copyright Office’s stance that AI-generated material, even when produced using AI as a development tool, does not qualify for copyright protection. However, IBM’s successful registration indicates that the company was able to establish sufficient human authorship—likely through structuring, selecting, and refining the AI-assisted output.

This raises key issues about how much human intervention is necessary to claim authorship over AI-assisted software. The Office’s decision suggests that merely accepting or slightly modifying AI-generated code may not be enough to meet the threshold for protection. Instead, the process of integrating AI-generated elements into a coherent, original work—through substantial revision, organization, and creative structuring—may be the determinative factor. The case also highlights a growing challenge in copyright law: how to evaluate human originality when AI is deeply embedded in the creative and technical process. While IBM’s approach was evidently sufficient, the boundaries of what constitutes “enough” human authorship remain uncertain and will likely require further clarification through case law or additional Copyright Office guidance.

The copyright registration for Adobe GenStudio for Performance Marketing includes only the human-authored components of the software, explicitly excluding AI-generated source code, particularly from GitHub Copilot. This reflects the Copyright Office’s firm stance that AI-generated material, even when used as a routine coding tool, does not qualify for copyright protection. GitHub Copilot, trained on vast repositories of open-source code, generates code snippets in response to user prompts, often integrating seamlessly into a developer’s workflow. The Office’s exclusion of AI-generated code suggests that even when such content is an essential part of a software project, it is still ineligible for protection unless it can be shown that a human author exercised sufficient creative control over its selection, arrangement, or modification.

The Fire Spotter registration follows the same framework as other AI-assisted software registrations, recognizing only the portions of the software written by a human programmer while excluding AI-generated components.^[25] The exclusions of AI-generated source code suggest that:

• The Office is applying a bright-line rule—if a portion of a work is created by AI, it is categorically excluded from protection.
• The role of the human developer in shaping, organizing, and refining AI-assisted work is evaluated on a case-by-case basis, with no clear-cut threshold for when human input is sufficient for copyright protection.
• AI-generated content, even when essential to a functional software product, does not automatically become copyrightable through integration or modification.

As a parallel, for visual artists, this means that merely generating an image through an AI tool does not establish copyright authorship. Artists must demonstrate meaningful creative input—such as editing, refining, composing, or integrating AI-generated visuals into a broader artistic vision—to claim copyright protection.

Part 3 of the U.S. Copyright Office Report, expected in late 2025, is anticipated to examine the legal implications of training AI models using copyrighted works. It will likely address key issues such as licensing requirements, potential liability for copyright infringement, and the broader debate over incorporating copyrighted material into AI training datasets. Part 3 may also be guided by court decisions, such as Thomson Reuters v. ROSS Intelligence, where a federal court ruled that AI-assisted legal research software infringed copyrighted legal materials.^[26] In this case, Thomson Reuters, the parent company of Westlaw, argued that Ross Intelligence unlawfully copied and used Westlaw’s proprietary legal content, specifically its headnotes and key number system to train an AI-powered legal research tool. The court rejected ROSS’s fair use defense, emphasizing that the wholesale copying of the copyrighted materials to develop an AI system was not transformative and harmed the market for the original works.

Conclusion

The Copyright Office’s 2025 Report, alongside recent registration decisions, provides a clearer framework for applying copyright law to AI-assisted works. By reaffirming the necessity of human authorship, the 2025 Report reinforces the long-standing principle that copyright protection extends only to works exhibiting meaningful human creativity. However, recent registrations of AI-assisted creative works—including visual art, music videos, and software—demonstrate how the Office is actively shaping the legal landscape for hybrid AI-human works. These decisions offer valuable guidance for artists, designers, musicians, programmers, and policymakers, clarifying the evolving thresholds for copyright protection in an AI-driven creative economy.

At the same time, critical questions remain. Where should the line be drawn between sufficient and insufficient human input? How will future AI technologies further blur these distinctions? The 2025 Report suggests that existing laws are adaptable, yet it also acknowledges the need for continued evaluation—and potentially judicial or legislative intervention—as AI capabilities advance. As artists and creators continue to explore AI as a tool for innovation, the balance between human authorship and technological assistance will remain a key issue in the evolving copyright framework.

About the Author

Atreya Mathur is the Director of Legal Research at the Center for Art Law. She was the inaugural Judith Bresler Fellow at the Center (2021-22) and is a Master of Laws Graduate from New York University School of Law where she specialized in Competition, Innovation, and Information Laws, with a focus on copyright, intellectual property, and art law.

References:

1. Part 1 of The 2025 Report on Copyright and Artificial Intelligence released in July 2024, primarily addresses digital replicas, including AI-generated reproductions of a person’s voice, likeness, or other personal attributes. It examines the challenges digital these replicas pose, particularly in the context of copyright, right of publicity, and potential legislative considerations. SeeU.S. COPYRIGHT OFFICE, Copyright and Artificial Intelligence: Part 1: Digital Replicas (2024) available at https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-1-Digital-Replicas-Report.pdf ↑
2. See U.S. COPYRIGHT OFFICE, Copyright and Artificial Intelligence: Part 2: Copyrightability (2025) available at https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf. ↑
3. Id. ↑
4. Id. ↑
5. Id. ↑
6. Id. ↑
7. Id. ↑
8. Id. ↑
9. See Sam Cohen, The Copyright Office’s Latest Guidance on AI and Copyrightability, JD Supra (Feb. 4, 2025) available at https://www.jdsupra.com/legalnews/the-copyright-office-s-latest-guidance-2342492/#:~:text=Looking%20ahead%2C%20Part%203%20of,intersection%20of%20AI%20and%20copyright. ↑
10. A Collection of Objects Which Do Not Exist, VAu001544214 (Oct. 2024). ↑
11. Id. ↑
12. A Single Piece of American Cheese, VAU001543942 (Aug. 2024). ↑
13. Inpainting is a technique of filling in missing regions of images that involves filling in the missing or damaged parts of an image, or removing the undesired object to construct a complete image. See https://help.nightcafe.studio/portal/en/kb/articles/what-is-inpainting-in-ai-and-how-to-use-it. Also see Katelyn Chedraoui, This Company Got a Copyright for an Image Made Entirely With AI. Here’s How, CNET (Feb. 10, 2025) available at https://www.cnet.com/tech/services-and-software/this-company-got-a-copyright-for-an-image-made-entirely-with-ai-heres-how/ ↑
14. Zarya of the Dawn, VAu001480196 (Feb. 21, 2023). ↑
15. Id. ↑
16. Id. ↑
17. Film clip for song Just Like In A Movie (SNEAK PREVIEW), PAU004252052 (Dec. 2024). ↑
18. Id. ↑
19. Id. ↑
20. See 17 U.S.C. § 102(a)(2); also see Newton v. Diamond, 388 F.3d 1189 (9th Cir. 2004) and Williams v. Gaye, 895 F.3d 1106 (9th Cir. 2018). ↑
21. 17 U.S.C. § 102(a)(1); also see Google LLC v. Oracle America, Inc., 141 S. Ct. 1183 (2021). ↑
22. See Thaler v. Perlmutter, 2023 WL 5333236 (D.D.C. Aug. 18, 2023). ↑
23. See U.S. Copyright Office allows registration of computer programs w/ AI-generated source code excluded, Chat GPT Is Eating the World (Feb. 18, 2025) available at https://chatgptiseatingtheworld.com/2025/02/18/u-s-copyright-office-allows-registration-of-computer-programs-w-ai-generated-source-code-excluded/ ↑
24. IBM Watsonx Code Assistant (and Enterprise Java Applications) (extensions for VSCode and Eclipse) v. 1.0, TX0009460181 (Dec. 2024). ↑
25. Fire Spotter, TXU002436023 (Jan. 2024). ↑
26. Thomson Reuters Enter. Ctr. GmbH & W. Publ’g Corp. v. Ross Intelligence Inc., 1:20-cv-613-SB (D. Del. Feb. 11, 2025). ↑