Introduction

A collision is on the horizon. The collision is between a strict interpretation of the human authorship requirement under U.S. copyright law, and the ascendence of generative artificial intelligence (Gen AI) as an essential element of modern creativity. From publishing and advertising to music, film, media, and software development, AI systems are reshaping workflows and redefining authorship. Yet, as this technological revolution accelerates, U.S. copyright law remains anchored to an interpretation that feels increasingly out of step: that Gen AI outputs are all ineligible for copyright and, thus, automatically ejected into the public domain.

The U.S. Copyright Office has taken a firm position on this point. Unfortunately, this stance, while perhaps doctrinally consistent, is misaligned with economic reality. Our copyright-reliant industries – arguably some of the crown jewels of the U.S. economy — cannot tolerate prolonged uncertainty about copyright ownership, protection, registration and enforceability. If copyright law and practice fail to adapt, there could be economic consequences for U.S. competitiveness. And, even at the individual level, there is a growing sense that the law should not deprive authors of the fruits of their genuine authorship, even when intertwined with Gen AI.

Consideration should be given to the ways in which copyright could evolve to provide greater certainty to our creative industries, while remaining aligned with the foundational purposes of copyright law. This article — in an all-too-cursory-but-constrained-to-blog-post manner — examines three possible paths forward, namely: (1) amending the Copyright Act to construe Gen AI outputs as works-made-for-hire, (2) amending the Copyright Act to allow Gen AI outputs to be viewed as joint works with human users, and (3) allowing the law to continue to evolve organically through the courts and agency guidance. Each approach carries implications for creators, businesses, and the integrity of copyright law.

The Human Authorship Requirement: A Doctrinal Anchor

Copyright law in the United States rests on the premise that the copyright monopoly incentivizes human creativity. Section 102 of the Copyright Act protects “original works of authorship,” and courts have consistently interpreted “authorship” as requiring a human creator. This principle was reaffirmed in Thaler v. Perlmutter, where the DC Circuit held recently that a Gen AI system cannot be an author under current law, and that Stephen Thaler, the system creator and user in the case, was also not the author of the work.

The Copyright Office’s guidance echoes this view: works “produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author” are not registrable. The rationale is understandable — copyright exists to reward human creativity.

Yet, the boundary between tool and creator is blurring. When a human conceptualizes the work and provides prompts, but the Gen AI determines composition and detail, is this human authorship? If a designer uses Gen AI to generate hundreds of variations and selects one for refinement, does that selection, together with the refinement, constitute sufficient creative input or control? At what point does a combination of preliminary concepts, prompts, selection and refinement cross over from non-authorship to authentic, protectable human authorship? These questions expose the inadequacy of a rigid test in an era of collaborative creativity.

Specific Areas of Concern

The creative community appears to be coalescing around at least three key areas of concern with the present state of play:

1. Exclusion of Authentic Authorship.  There is a sense that the Copyright Office’s decisions to date are yielding inequitable denials of copyright to works that in fact do constitute authentic human authorship. As now well-known, Jason Allen undertook 624 prompts and significant editing and refinement before finalizing his award-winning Théâtre D’opéra Spatial; Kris Kashtanova developed 18 pages of images displaying consistent tonal and thematic elements, including significant editing and refinement by Kashtanova, in creating their Zarya of the Dawn comic; and Stephen Thaler programmed and trained his own Gen AI tool, and then directed its output of A Recent Entrance to Paradise. Are these artists really undeserving of copyright protection for these works? Do the works not embody sufficient origination, conceptualization, organization, selection, control, creativity, execution, refinement, superintendence and stewardship to constitute human authorship? Does denial of copyright protection for these works serve the objective of promoting artistic progress? For many, the answer to these questions is no.
2. Burdensome/Potentially Impossible Copyright Registration Procedure.  Another area of concern involves the Copyright Office’s new requirement that applicants describe and disclaim any more-than-de-minimis contributions to the work created using Gen AI. Per the Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence, applicants are asked to separate portions created by the human author from portions created by the AI. In many cases, this is an extremely burdensome and even potentially impossible task – works often go through an iterative process where dozens to hundreds of drafts are prepared, revised, further revised and so forth, in order to achieve the final product. Under those circumstances, can an applicant always accurately separate which portions were created by AI and which by the human? And if they get it wrong, is the registration subject to invalidation (thus putting statutory damages, attorneys fees, and the other benefits of registration at risk)?
3. Litigation Uncertainty.  A final concern involves copyright infringement litigation. Under current law, virtually every defendant in every copyright infringement lawsuit going forward will be motivated to challenge the validity of the underlying copyright based on the potential presence of Gen AI portions. Even if a plaintiff has complied with the law and only claimed the human-authored portion of a work in the work’s registration, the copyright could still be challenged. And, if the plaintiff is unable to produce detailed records proving the exact portions contributed by the human author versus the Gen AI, the copyright could ultimately be found unenforceable. This creates a tremendous burden and great uncertainty for the future for all copyright infringement plaintiffs, and seems ultimately untenable.

The remainder of this article considers three possible ways in which copyright could evolve to meet the above challenges, to ensure both the continued robustness of U.S. copyright law as well as creating a welcoming environment for the increasing use of Gen AI to advance innovation, creativity, and productivity in our creative industries.

Works-Made-for-Hire?

One proposed solution would be to amend the Copyright Act to allow AI-generated works initiated by a human user to qualify as “works-made-for-hire.” Under current law, works-made-for-hire include those created by employees within the scope of employment or certain commissioned works under written agreement. Extending this concept to the Gen AI context would vest ownership of Gen AI outputs in the user, providing clarity and predictability.[1]

The advantages of this approach are significant. It would offer creative individuals and businesses copyright certainty, enabling them to invest confidently in various types of Gen AI-driven projects and use cases without fear of losing intellectual property rights to the public domain. Questions would be reduced about whether enough control was exerted by the human user, how authorship should be identified on copyright applications, and how copyrights could be enforced in copyright litigation. Any need to document clear separation of the contribution of the Gen AI versus the contribution of the human author would be mitigated.

This structure also makes some conceptual sense: when an employer engages an employee to develop a creative solution (whether in text, software, visual art, music, or otherwise), the employer defines the problem, task, or concept, the employee executes creative work to design a solution, and the parties may then work together to further refine the solution until finished. In that case, the employee would ordinarily own the copyright in her creative work, but that result is flipped by the work-made-for-hire doctrine – with the ownership and authorship of the work vesting in the employer instead of the employee. Analogously, a Gen AI user ideates a problem, task, or concept, prompts the Gen AI system accordingly, the Gen AI executes creative work to produce a solution, and the user may further refine the output, either with or without the AI’s assistance, until a satisfactory final product is achieved. Instead of the Gen AI being viewed as the author, the ownership and authorship of the work would vest automatically in the human user.

However, there are important drawbacks to this approach which cannot be ignored. Gen AI systems are neither employees nor contractors, and redefining “work-made-for-hire” to include non-human entities may distort statutory coherence and create other interpretive problems. This approach could also be over-inclusive and dilute the authorship principle, incentivizing minimal human input and undermining the creative threshold that copyright seeks to protect and encourage. The question would invariably be asked, do we really want to award a copyright monopoly to a work fully created by Gen AI with, say, a single three-word prompt by the human user? Is the objective of promoting progress in the useful arts still being met with this solution? 

Joint Works?

Another proposal is to treat Gen AI outputs as joint works, with the human user as co-author. Joint authorship under current law requires two or more authors intending their contributions to merge into a unitary whole, and with each contributing sufficient material to constitute authorship. Extending this concept to Gen AI would acknowledge the collaborative nature of human-AI creativity.

The advantages of this approach lie in its realism. Many creative processes now involve iterative human-AI interaction, and recognizing joint authorship would reflect this dynamic. It also preserves human contribution, ensuring users retain rights even when Gen AI plays an equal or even dominant role. Moreover, this model offers flexibility, adapting to varying degrees of human input and accommodating diverse creative workflows. It would also still require non-minimal contribution by the human author, and would allow for both the Gen AI and human portions of the work to be safely protected by copyright and not exposed to potential relegation to the public domain.

Yet, this solution is also not without challenges. Gen AI cannot own anything, making “joint authorship” a legal fiction. Further, determining the threshold of human input required for authorship would introduce ambiguity. This approach would still suffer from the need for the human to demonstrate a non-minimal contribution, and courts would struggle to apply consistent standards. This uncertainty could lead to litigation difficulties that may be no better than the challenges of the present legal situation, with disputes over how much human authorship was present and whether this amount was sufficient for the human to be considered a joint author of the work. (That said, the minimum threshold for determining whether a contribution is sufficient for participation in a joint work is well-trod legal terrain; in theory, this would be a significantly better and more certain area to litigate versus the as-yet uncertain area of identifying the boundary between copyrightable human-authored works assisted by Gen AI and Gen AI-produced works ineligible for copyright.) 

Organic Development; Just Let it Be?

The third path is to let the law evolve organically through judicial decisions and Copyright Office guidance. Under this approach, courts would continue to distinguish works that are “substantially Gen AI” (ineligible) from those with “sufficient human authorship” (eligible), refining standards case by case in a slow and laborious process.

This approach offers several advantages. It avoids premature legislative overhaul, allowing nuanced, fact-specific evolution as technology matures. Judicial flexibility ensures that courts can adapt standards to emerging realities without locking in rigid statutory definitions. Importantly, this path preserves core principles, maintaining human authorship as a doctrinal anchor while accommodating incremental change.

However, the drawbacks are considerable. Prolonged uncertainty leaves businesses and individual creators navigating a legal gray zone, exposing them to risk and inconsistent outcomes. This ambiguity could chill investment in AI-driven creativity, deterring innovation in sectors that depend on copyright certainty at a time when these industries should be innovating and taking advantage of the opportunities offered by AI augmentation.  Fragmentation is another concern, as different courts and agencies could adopt divergent approaches, creating a patchwork of standards that complicates compliance.

Conclusion and Next Steps

Unlike many other hype-curve developments in recent years, Gen AI truly seems to challenge copyright law at a fundamental level. Strict interpretation of the human authorship requirement to exclude Gen AI outputs seems increasingly impractical in a world where creativity is often co-produced by humans and machines. No solution is perfect. Legislative reform offers clarity but risks doctrinal distortion; organic development preserves principles but prolongs uncertainty.

In the short term, the Copyright Office could refine its guidance to provide clearer thresholds for human authorship, reducing ambiguity for creators and businesses. It also could consider acknowledgement of a broader array of contexts in which human authorship may be recognized and potential relaxation of application requirements. In the long term, Congress will likely need to act — whether through a works-made-for-hire model, a joint authorship framework, a hybrid approach or something else — to ensure that copyright law remains fit for purpose in an AI-driven economy and can continue to incentivize creative expression.  

This article covers complex topics in insufficient detail. Further research is warranted in the following areas:

1. How do the philosophical foundations of our concept of “authorship” inform our perception of human authorship in the AI era? Is our “author concept” a relative historical notion that could be updated to meet current practical reality, or is it necessary for legal continuity? 
2. What exactly does the Constitution’s mandate to promote “progress” in the useful arts mean? What types of behaviors and outcomes – aesthetically, economically, culturally, and phenomenologically – are we intending to promote with copyright?
3. Are our copyright-reliant creative industries really so copyright dependent that AI disruption to copyright ownership certainty poses a serious economic risk? Is it possible that newer business models and technology are sufficient to render uncertainty in copyright not so bad?
4. Some works, like the rejected Allen, Kashtanova and Thaler works, seem worthy of copyright protection despite the role of Gen AI, while others – an image, say, resulting from a three-word prompt to Midjourney with nothing more, e.g. – do not. What principles should guide the Copyright Office and law to better determine which works are monopoly-worthy and which are not? Would there be harm in granting at least “thin” protection to users in connection with Gen AI outputs?

[1] As outlined in a prior post, Dr. Thaler, together with a collection of scholars in an amicus brief, have advocated for the construal of Gen AI works as works-made-for-hire, with Dr. Thaler as the putative employer/commissioner and his Gen AI “Creativity Machine” as the putative employee/contractor.