Niraj Yawalkar · Texas A&M Undergraduate Journal of Law and Society · January 2025 Article Release

2I. Introduction

Generative artificial intelligence (AI) systems can now produce text, images, music, and other creative content that closely resemble human-created works. This rapid advancement has raised novel legal questions of authorship and ownership: Can AI-generated material be copyrighted, and if so, under what conditions? U.S. law has long held that copyright protects only creations of human intellect, and recent developments confirms that principle in the AI context. In the past two years, the U.S. Copyright Office and courts have grappled with how to apply existing copyright law to AI-generated works. The U.S. Copyright Office launched an initiative in 2023 to study AI and copyright, gathering public comments and holding hearings. On January 29, 2025, the Office released Part II of a major report on Copyright and Artificial Intelligence, addressing when AI-generated outputs are eligible for copyright protection. At the same time, lawmakers have begun proposing legislation to tackle related issues––from "deepfake" AI impersonations to the use of copyrighted data to train AI models. This article reviews the latest Copyright Office guidance (with a focus on the January 2025 report) and recent legislative activity, against the backdrop of ongoing legal cases like Thaler v. Perlmutter, to illustrate how the law is adapting to new technologies and the challenges of defining creativity and ownership in the age of AI.

3II. Background: AI and the Human Authorship Requirement

Human authorship as a baseline: U.S. copyright law has consistently required a human creator for a work to be protected. The Constitution's Copyright Clause and the Copyright Act were intended to incentivize human creativity. Courts have repeatedly affirmed that non-humans (whether animals or machines) cannot be "authors" under the law. For example, in the famous "monkey selfie" case, a federal court held that a monkey (being distinctly non-human) lacked standing to claim copyright. In August 2023, this principle was tested again in Thaler v. Perlmutter, where the creator of an AI system sought to register a visual artwork that his AI had generated autonomously, with no human involvement. The Copyright Office rejected the application on the ground that copyright law "extends only to works created by human beings," and the federal court agreed, emphatically holding that "human authorship is an essential part of a valid copyright claim."

AI-assisted creativity: However, most real-world uses of AI in art and writing involve some human input. The legal gray area lies in works produced with assistance from AI. If a person provides ideas, prompts, or edits to an AI-generated draft, is the result a human-authored work or a machine-authored one? Prior to the recent report, the Copyright Office had already begun addressing this question. In March 2023, the Office issued guidance on registering works that contain AI-generated material. This guidance, prompted in part by cases like Zarya of the Dawn (a graphic novel with AI-generated images), made clear that applicants must disclose AI-generated content in a work and that copyright will only cover the portions with human creative input. The Office emphasized the need for a case-by-case analysis: the key inquiry is whether the AI contributions are the product of "mechanical reproduction" or of the author's own original mental conception. In essence, copyright protects only what a human being created, even if an AI tool was used in the process. This interim guidance set the stage for the more comprehensive analysis in the Copyright Office's January 2025 report.

Defining human authorship in AI outputs: In Part II of its Copyright and Artificial Intelligence report (issued Jan. 29, 2025), the Copyright Office tackled the copyrightability of AI-generated works in depth. The report reaffirms that human creativity is the crux of determining copyright protection. Crucially, the Office drew a line between using AI as a tool versus using it as the actual creator. It concluded that "the outputs of generative AI can be protected by copyright only where a human author has determined sufficient expressive elements" of the work. In practical terms, this means a person must exercise creative control––for instance, by selecting, arranging, editing, or contributing original material to the AI-generated content––such that the person's own creativity is perceptible in the result. On the other hand, simply instructing an AI with a text prompt and accepting whatever it generates is not enough to make the prompter the "author." The Office explicitly stated that the mere provision of prompts to an AI is insufficient for copyright, given that the AI's process is often unpredictable and not guided by human creativity in detail. Each case will depend on its facts, but generally "AI-generated material cannot be copyrighted unless a human author has determined the expressive elements of the work."

Incorporating AI elements into human works: The report provides helpful scenarios to illustrate these principles. If AI is used assistively, for example, an author uses an AI program to brainstorm ideas or generate rough text that the author then substantially rewrites, the final work can still be copyrighted in the usual way, because the human's creativity dominates. If a human author inputs their own copyrighted material into an AI (say, to have the AI transform or remix it), any output that incorporates the author's original material will be protectable to that extent (similar to how a derivative work is protected only for the new material added). Likewise, if a human selects and arranges AI-generated content in a creatively original way––as one might curate a collage or create a photo album––the compilation or arrangement can be copyrighted, though the individual AI-created elements remain unprotected. The Copyright Office stressed that including AI-generated portions in a larger human-created work does not disqualify the work from protection; the law will simply "filter out" the AI-generated portions as uncopyrightable, while still protecting the original human contributions. In other words, AI-generated content is treated like material from the public domain: it can be part of a copyrighted work, but it receives no protection itself. This framework places the onus on human creators to ensure they add their own creative expression if they wish to claim copyright in works involving AI.

No new rights for AI-only works: Significantly, the Copyright Office concluded that no change in law is presently needed to handle AI-generated works––current doctrines are flexible enough. The report expressly rejected calls for new protections for AI-generated content. The Office reasoned that extending copyright to purely machine-produced material would "undermine rather than further the constitutional goals of copyright." Because copyright exists to encourage human creativity, granting exclusive rights to AI outputs (which cost the human creator little more than a text prompt) could actually discourage human artists and authors. This stance aligns with the views of many experts and creators who submitted comments, and it maintains parity with the historical adaptability of copyright law; From photographs to software, new tools have been accommodated without abandoning the core requirement of human originality.

5IV. Legislative Activity and Future Developments

Proposed legislation (deepfakes): Even as the Copyright Office finds existing law largely sufficient for AI-authored works, lawmakers have been exploring legislative updates related to AI-generated content. Part I of the Copyright Office's AI report (released July 2024) dealt with AI-generated "digital replicas" of real people––essentially deepfake images, videos, or audio that mimic an individual's likeness or voice. The Office observed that current copyright law offers no protection to someone whose likeness is copied by an AI (because copyright protects expression, not personal identity). To fill this gap, the report urged Congress to enact a federal law prohibiting unauthorized, realistic AI-generated replicas of people. Specifically, they recommended a law applying banning knowing distribution of fake likenesses without consent even if done non-commercially, and requiring online platforms to promptly remove such content upon notice. It also suggested strong remedies, including statutory damages and even criminal penalties for egregious violations. In response, several bills were introduced in 2024 to tackle deepfakes and AI impersonation. Prominent examples include the No FAKES Act and the No AI Fraud Act, which would create new legal rights over one's digitally replicated likeness. These proposals seek to treat unauthorized AI replicas as a form of intellectual property or privacy violation, empowering victims to sue and requiring platforms to remove infringing deepfakes. While such bills are still pending, they demonstrate a bipartisan recognition that AI technology is outpacing existing laws in certain areas. Lawmakers have also shown interest in transparency and accountability for AI training data, which foreshadows the next phase of policy activity. As creative industries voiced concern about AI companies––very recently OpenAI––using millions of copyrighted works as training fodder, some members of Congress proposed requiring disclosures or even remuneration. For instance, a bill introduced in late 2024 would authorize copyright owners to demand information about how their works were used to train an AI model, with an eye toward compensation for such use. Although no consensus has yet emerged on this issue, it is a key area to watch.

Anticipated developments: The legal status of training AI on copyrighted works is one of the most contentious unresolved questions. Is feeding copyrighted images, books, or songs into a machine learning model a fair use, or does it infringe the reproduction right? Thus far, a number of lawsuits by artists, authors, and media companies against AI developers (for example, Getty Images v. Stability AI, and class actions against OpenAI and others) allege that unlicensed scraping of protected content to train generative AI violates copyright law. The defendants often counter that such intermediate copying is transformative and permissible, but no court has definitively ruled on the merits yet. This gap is exactly what the Copyright Office's forthcoming Part III of the AI report is set to address. According to the Office, Part III will examine the legal implications of AI training: how existing doctrines like fair use might apply, what licensing models could emerge, and who (if anyone) should be liable if an AI model's training infringes copyrights. The report may influence whether Congress decides to step in with legislation or waits for judicial rulings. If courts find that current law does treat broad scraping of works as infringement, there could be pressure to create a new exception or compulsory license to enable AI innovation. Conversely, if AI training is deemed fair use, content creators might lobby Congress for new protections or remuneration schemes. The Copyright Office has indicated it will also consider "any potential liability" and licensing frameworks in this context. Ethically, this debate raises interesting questions about balancing incentives: allowing AI to learn from prior art may spur technological progress, but doing so without credit or compensation to artists could undercut the incentive to create in the long run.

The evolving guidance and proposed laws around AI-generated works highlight a fundamental challenge: how to preserve the legal notion of creativity and ownership in the era of AI. By drawing a bright line at human creativity, the Copyright Office and courts are essentially reaffirming a value judgment––that copyright exists to promote human innovation and expression. This has practical consequences. On one hand, it means that a work produced entirely by an AI (with no human author) is effectively born into the public domain. Anyone may use or reproduce such AI-generated content freely, since no one can claim exclusive rights over it. This could benefit society by expanding the pool of freely available knowledge and art, much as we freely use facts or folk tales. It also avoids the scenario of tech companies or prompt users monopolizing content that they did not truly create. On the other hand, the inability to copyright AI-generated works might disincentivize certain investments––for example, a publisher might be reluctant to invest in an AI-composed novel if it can't be protected against copycats. Moreover, as AI becomes more sophisticated, the philosophical question looms larger: what counts as "creative" when a machine can devise intricate images or text? The law's answer for now is clear––creativity, for legal purposes, requires a human mind at work––but this could be tested as AI systems begin to mimic human-like decision-making in the creative process.

Ethically, the distinctions being drawn force us to consider the role of the human artist in an AI-assisted future. The requirement that a human contribution be "sufficient" or meaningful in an AI-assisted work encourages a model of collaboration where AI is a tool under human direction, rather than an independent creator. It places value on the intentionality and original thought of the human user. Many people have brought concerns about how human "original thought" maintains a sufficient distinction from increasingly advanced AI. As inference models, or even some new innovative model type, get better, the differences between human and AI output continue to decrease. When an AI can replicate human-like thought in a nondifferentiable way, the Copyright office may be forced to completely reevaluate the central basis for copyright protection. The philosophical debates around AI's "intelligence" and likeness to human sentience are far too complex to cover in this article, however they will remain increasingly important as AI progresses, and technological advancements force policymakers to upend entire frameworks.

7VI. Conclusion

As of early 2025, the U.S. approach to AI-generated works can be summarized this way: if a human brain conceived the expressive elements, the work can be protected––if not, no copyright can be applied. The Copyright Office's latest guidance doubles down on that principle, providing more nuance on mixed human/AI works but stopping short of any radical change. Meanwhile, Congress is actively considering how AI technologies should be constrained or regulated when they intersect with personal rights and creative industries. Stakeholders should keep a close eye on the upcoming Part III of the Copyright Office report (on AI training and fair use), as well as legislative proposals on the horizon, such as deepfake bills or data transparency requirements. These developments will further define the legal landscape for AI and copyright. In a broader sense, society is in the midst of negotiating what "authorship" means in a world where machines can generate content. The law is attempting to draw lines that uphold the spirit of copyright––promoting the progress of science and useful arts by rewarding human creativity––while acknowledging the reality that AI is now part of the creative toolkit. By clarifying definitions and responsibilities now, the hope is to provide guidance that will channel AI's capabilities in a way that complements, rather than replaces, human creativity. The conversation is far from over, but the newest Copyright Office report and emerging legislation mark an important step in charting how we define creativity and ownership in the age of AI.