Man vs. Machine: The Copyright Office’s 2025 AI Policy and Stance on AI-Generated Art

Credit: VentureBeat made with Midjourney

The rapid advancement of authorship and AI has sparked a new era of generative art, challenging traditional notions of authorship and copyright. In the constantly changing landscape of creativity, AI-powered tools, such as DALL-E, Midjourney, Deep Dream Generator, and Gemini, are reshaping artistic expression. AI has moved beyond its traditional limits, now generating intricate and visually stunning artworks through these innovative tools. As AI systems become increasingly capable of generating content, the issue of ownership and rights over these creations becomes more complex. 

The recent 2025 Report from the U.S. Copyright Office emphasizes the rapidly shifting landscape of AI and copyright law. An expansive interpretation of the Copyright Act that accommodates emerging technologies could provide clarity and grant copyright to AI-assisted works. However, such an approach could undermine the value of human authorship, negatively impacting artists and forcing them to compete against advanced and realistic generators. Policymakers face the challenge of balancing technological advances with incentives for human creativity, ensuring that artists and creators can continue flourishing while AI advances.

A Brief History of Human Authorship 

U.S. copyright law has historically relied on the principle of human authorship. Section 102 of the Copyright Act protects “original works of authorship fixed in a tangible medium of expression,” a standard courts have consistently interpreted as requiring a human creator. The Supreme Court reinforced this requirement in the landmark case Feist Publications, Inc. v. Rural Telephone Service Co. (1991), ruling that originality requires independent creation and a minimum level of creativity. In Feist, the Court determined that a phonebook directory lacked originality because it merely compiled factual information. However, creativity emerges when an author makes expressive choices in selection, arrangement, or presentation. 

Photo by Library of Congress Newsroom

The 2025 Copyright Office Report: The Copyright Office’s Current Stance on AI Art 

On January 29, 2025, the U.S. Copyright Office released the second part of its Report addressing the copyrightability of AI-generated materials. The Report explores the nature and extent of human involvement required for these outputs to qualify for copyright protection and delineates between “AI-assisted” and “AI-created” works, providing guidelines for what and what does not qualify for copyright. The Copyright Office emphasized that the copyrightability of AI-generated works must be determined on a case-by-case basis, considering the specific circumstances of the creation. While AI assistance does not inherently affect copyrightability, works that are entirely generated by AI are not eligible for protection as they fail the Feist standard of human creativity. Further, if an artwork is produced entirely by AI, inputting a prompt into an AI system does not provide enough human control to establish authorship, as prompts merely function as instructions rather than expressions of creative authorship. However, suppose an artist employs AI to generate elements of a piece but maintains creative control over the selection, arrangement, and modification of these elements. In that case, the work is considered “AI-assisted” and may be eligible for copyright. When AI tools assist in a human-created, copyrightable work, the human author retains copyright over the unaltered, perceptible portions of their original authorship as well as the creative selection or creative modifications of the outputs.

The Copyright Office’s 2025 Report reinforces the approach applied in the 2022 comic book Zarya of the Dawn, created by Kristina Kashtanova. Zarya includes a human-written narrative and images generated by the AI tool Midjouney. The U.S. Copyright Office initially granted Kashtanova a copyright registration for the entire work. However, upon discovering that AI produced the images, the Office reexamined the registration and partially revoked the approval in February 2023. While affirming protection for Kashtanvoa’s text and her selection and arrangement of the images, the Office denied protection for the individual AI-generated images. The Office concluded that Kashtanova did not exercise sufficient control over those images’ visual details, concluding that an AI process, rather than human creativity, shaped the comic’s final form. 

The Wall Street Journal 

By formally citing Zarya of the Dawn, the 2025 Report provides more explicit guidelines for distinguishing “AI-assisted” from “AI-created” works. The Office explains that text prompts alone do not make users the “author” of AI-generated output. Instead, true authorship requires meaningful human control over the expressive details rather than a simple prompt or acceptance of whatever the system produces. The report emphasizes that “prompts alone do not provide sufficient human control to make users of an AI system the authors of the output.” Under the 2025 framework, Zarya’s Midjourney-created images remain “not protectable” because Kashtanova’s text prompts did not result in sufficient control over “determining the expressive elements” in the output.

Although the 2025 Report focuses on whether a final AI-generated work meets the human authorship standard, Zhang v. Google, a recently filed proposed class action in the U.S. District Court for the Northern District of California, raises a separate but interconnected question of whether training an AI system on copyrighted images constitutes infringement of those underlying works. Plaintiffs allege Google’s Imagen text-to-image AI “ingested” millions of photos without permission, potentially copying entire works to build its training database. Regardless of how training data is acquired, the Copyright Office’s stance remains that purely AI-driven outputs lack human authorship under Feist, which requires a human mind behind creative expression. Only when a user meaningfully directs or modifies the output may there be sufficient authorship to secure copyright, which also presupposes that the initial training process was itself lawful. 

Exhibit B from Zhang v. Google 

While Zhang v. Google does not explicitly address the copyrightability of AI-generated images, it nonetheless raises a fundamental question at the heart of Feist: what qualifies for “original work of authorship?” Although Imagen’s outputs may exhibit significant creativity without directly replicating training images, copyright law still requires the crucial element of human creativity. Courts, including those in Thaler, have consistently held that if the creativity of a work is entirely derived from a machine, the authorship prong of Feist is not met. Therefore, even if an AI generator produces unique and non-duplicative images, the absence of human autonomy in generating the expressive details renders the images unprotectable. 

Beyond the lawfulness of AI training processes, Zhang also touches on whether instructing an AI model to generate images in a particular style meets the threshold for human authorship. Addressing this complexity, Edward Lee, an expert and professor at Santa Clara University School of Law, wrote a letter to the U.S. Copyright Office proposing a nuanced approach. Professor Lee advocates for a “bare minimum” test of authorship, emphasizing that a human’s selection or arrangement of elements in a work, no matter how minimal, can qualify for copyright protection under Feist. However, works that meet only this minimal threshold would receive only “thin” copyright protection, meaning they would only be protected against identical copying, not derivative uses or adaptations. This framework ensures that AI-assisted works receive protection only when a human’s creative intent and choices are demonstrable. Further, this approach avoids granting full copyright to AI-generated works, which could lead to monopolization and hinder progress, while still giving human creators legal rights when they play a clear role in the creative process. This “bare minimum” standard provides a clear, predictable test for courts and copyright offices while allowing the legal system to adapt as AI technology evolves.

Looking ahead, as AI technology continues to evolve and play a greater role in content creation, legal frameworks will need to adapt to address the complexities of authorship and copyrightability. Zarya of the Dawn and Zhang v. Google exemplify the growing challenges of determining how much human input is necessary for copyright protection and whether the methods used to train AI systems may themselves be infringing. By distinguishing AI-assisted from entirely AI-generated content, the 2025 Report offers some clarity, but this is not enough. Still, courts, policymakers, and industry leaders must explore solutions that balance innovation with the fundamental principles of copyright law, ensuring that human creativity remains in the course of intellectual property protection. Further, artists and creators must navigate this landscape, balancing the use of AI as a tool with the imperative to maintain sufficient creative control to secure their rights. As AI advances, the challenge lies in harnessing its potential while preserving the incentives and protections underpinning human creativity. 

*The views expressed in this article do not represent the views of Santa Clara University.