Introduction<\/p>\n
A collision is on the horizon. The collision is between a strict
\ninterpretation of the human authorship requirement under U.S.
\ncopyright law, and the ascendence of generative artificial
\nintelligence (Gen AI) as an essential element of modern creativity.
\nFrom publishing and advertising to music, film, media, and software
\ndevelopment, AI systems are reshaping workflows and redefining
\nauthorship. Yet, as this technological revolution accelerates, U.S.
\ncopyright law remains anchored to an interpretation that feels
\nincreasingly out of step: that Gen AI outputs are all ineligible
\nfor copyright and, thus, automatically ejected into the public
\ndomain.<\/p>\n
The U.S. Copyright Office has taken a firm position on this
\npoint. Unfortunately, this stance, while perhaps doctrinally
\nconsistent, is misaligned with economic reality. Our
\ncopyright-reliant industries \u2013 arguably some of the crown
\njewels of the U.S. economy \u2014 cannot tolerate prolonged
\nuncertainty about copyright ownership, protection, registration and
\nenforceability. If copyright law and practice fail to adapt, there
\ncould be economic consequences for U.S. competitiveness. And, even
\nat the individual level, there is a growing sense that the law
\nshould not deprive authors of the fruits of their genuine
\nauthorship, even when intertwined with Gen AI.<\/p>\n
Consideration should be given to the ways in which copyright
\ncould evolve to provide greater certainty to our creative
\nindustries, while remaining aligned with the foundational purposes
\nof copyright law. This article \u2014 in an
\nall-too-cursory-but-constrained-to-blog-post manner \u2014
\nexamines three possible paths forward, namely: (1) amending the
\nCopyright Act to construe Gen AI outputs as works-made-for-hire,
\n(2) amending the Copyright Act to allow Gen AI outputs to be viewed
\nas joint works with human users, and (3) allowing the law to
\ncontinue to evolve organically through the courts and agency
\nguidance. Each approach carries implications for creators,
\nbusinesses, and the integrity of copyright law.<\/p>\n
The Human Authorship Requirement: A Doctrinal Anchor<\/p>\n
Copyright law in the United States rests on the premise that the
\ncopyright monopoly incentivizes human creativity. Section 102 of
\nthe Copyright Act protects “original works of
\nauthorship,” and courts have consistently interpreted
\n“authorship” as requiring a human creator. This principle
\nwas reaffirmed in Thaler v. Perlmutter, where the DC
\nCircuit held recently that a Gen AI system cannot be an author
\nunder current law, and that Stephen Thaler, the system creator and
\nuser in the case, was also not the author of the work.<\/p>\n
The Copyright Office’s guidance echoes this view: works
\n“produced by a machine or mere mechanical process that
\noperates randomly or automatically without any creative input or
\nintervention from a human author” are not registrable. The
\nrationale is understandable \u2014 copyright exists to reward
\nhuman creativity.<\/p>\n
Yet, the boundary between tool and creator is blurring. When a
\nhuman conceptualizes the work and provides prompts, but the Gen AI
\ndetermines composition and detail, is this human authorship? If a
\ndesigner uses Gen AI to generate hundreds of variations and selects
\none for refinement, does that selection, together with the
\nrefinement, constitute sufficient creative input or control? At
\nwhat point does a combination of preliminary concepts, prompts,
\nselection and refinement cross over from non-authorship to
\nauthentic, protectable human authorship? These questions expose the
\ninadequacy of a rigid test in an era of collaborative
\ncreativity.<\/p>\n
Specific Areas of Concern<\/p>\n
The creative community appears to be coalescing around at least
\nthree key areas of concern with the present state of play:<\/p>\n
\nExclusion of Authentic Authorship. There is a
\nsense that the Copyright Office’s decisions to date are
\nyielding inequitable denials of copyright to works that in fact do
\nconstitute authentic human authorship. As now well-known, Jason
\nAllen undertook 624 prompts and significant editing and refinement
\nbefore finalizing his award-winning Th\u00e9\u00e2tre
\nD’op\u00e9ra Spatial; Kris Kashtanova developed 18 pages
\nof images displaying consistent tonal and thematic elements,
\nincluding significant editing and refinement by Kashtanova, in
\ncreating their Zarya of the Dawn comic; and Stephen Thaler
\nprogrammed and trained his own Gen AI tool, and then directed its
\noutput of A Recent Entrance to Paradise. Are these artists
\nreally undeserving of copyright protection for these works? Do the
\nworks not embody sufficient origination, conceptualization,
\norganization, selection, control, creativity, execution,
\nrefinement, superintendence and stewardship to constitute human
\nauthorship? Does denial of copyright protection for these works
\nserve the objective of promoting artistic progress? For many, the
\nanswer to these questions is no.
\n
\nBurdensome\/Potentially Impossible Copyright
\nRegistration Procedure. Another area of concern involves
\nthe Copyright Office’s new requirement that applicants describe
\nand disclaim any more-than-de-minimis contributions to the
\nwork created using Gen AI. Per the Copyright Registration
\nGuidance: Works Containing Material Generated by Artificial
\nIntelligence, applicants are asked to separate portions
\ncreated by the human author from portions created by the AI. In
\nmany cases, this is an extremely burdensome and even potentially
\nimpossible task \u2013 works often go through an iterative process
\nwhere dozens to hundreds of drafts are prepared, revised, further
\nrevised and so forth, in order to achieve the final product. Under
\nthose circumstances, can an applicant always accurately separate
\nwhich portions were created by AI and which by the human? And if
\nthey get it wrong, is the registration subject to invalidation
\n(thus putting statutory damages, attorneys fees, and the other
\nbenefits of registration at risk)?
\n
\nLitigation Uncertainty. A final concern
\ninvolves copyright infringement litigation. Under current law,
\nvirtually every defendant in every copyright infringement lawsuit
\ngoing forward will be motivated to challenge the validity of the
\nunderlying copyright based on the potential presence of Gen AI
\nportions. Even if a plaintiff has complied with the law and only
\nclaimed the human-authored portion of a work in the work’s
\nregistration, the copyright could still be challenged. And, if the
\nplaintiff is unable to produce detailed records proving the exact
\nportions contributed by the human author versus the Gen AI, the
\ncopyright could ultimately be found unenforceable. This creates a
\ntremendous burden and great uncertainty for the future for all
\ncopyright infringement plaintiffs, and seems ultimately
\nuntenable.
<\/p>\n
The remainder of this article considers three possible ways in
\nwhich copyright could evolve to meet the above challenges, to
\nensure both the continued robustness of U.S. copyright law as well
\nas creating a welcoming environment for the increasing use of Gen
\nAI to advance innovation, creativity, and productivity in our
\ncreative industries.<\/p>\n
Works-Made-for-Hire?<\/p>\n
One proposed solution would be to amend the Copyright Act to
\nallow AI-generated works initiated by a human user to qualify as
\n“works-made-for-hire.” Under current law,
\nworks-made-for-hire include those created by employees within the
\nscope of employment or certain commissioned works under written
\nagreement. Extending this concept to the Gen AI context would vest
\nownership of Gen AI outputs in the user, providing clarity and
\npredictability.1<\/p>\n
The advantages of this approach are significant. It would offer
\ncreative individuals and businesses copyright certainty, enabling
\nthem to invest confidently in various types of Gen AI-driven
\nprojects and use cases without fear of losing intellectual property
\nrights to the public domain. Questions would be reduced about
\nwhether enough control was exerted by the human user, how
\nauthorship should be identified on copyright applications, and how
\ncopyrights could be enforced in copyright litigation. Any need to
\ndocument clear separation of the contribution of the Gen AI versus
\nthe contribution of the human author would be mitigated.<\/p>\n
This structure also makes some conceptual sense: when an
\nemployer engages an employee to develop a creative solution
\n(whether in text, software, visual art, music, or otherwise), the
\nemployer defines the problem, task, or concept, the employee
\nexecutes creative work to design a solution, and the parties may
\nthen work together to further refine the solution until finished.
\nIn that case, the employee would ordinarily own the copyright in
\nher creative work, but that result is flipped by the
\nwork-made-for-hire doctrine \u2013 with the ownership and
\nauthorship of the work vesting in the employer instead of the
\nemployee. Analogously, a Gen AI user ideates a problem, task, or
\nconcept, prompts the Gen AI system accordingly, the Gen AI executes
\ncreative work to produce a solution, and the user may further
\nrefine the output, either with or without the AI’s assistance,
\nuntil a satisfactory final product is achieved. Instead of the Gen
\nAI being viewed as the author, the ownership and authorship of the
\nwork would vest automatically in the human user.<\/p>\n
However, there are important drawbacks to this approach which
\ncannot be ignored. Gen AI systems are neither employees nor
\ncontractors, and redefining “work-made-for-hire” to
\ninclude non-human entities may distort statutory coherence and
\ncreate other interpretive problems. This approach could also be
\nover-inclusive and dilute the authorship principle, incentivizing
\nminimal human input and undermining the creative threshold that
\ncopyright seeks to protect and encourage. The question would
\ninvariably be asked, do we really want to award a copyright
\nmonopoly to a work fully created by Gen AI with, say, a single
\nthree-word prompt by the human user? Is the objective of promoting
\nprogress in the useful arts still being met with this solution?<\/p>\n
Joint Works?<\/p>\n
Another proposal is to treat Gen AI outputs as joint works, with
\nthe human user as co-author. Joint authorship under current law
\nrequires two or more authors intending their contributions to merge
\ninto a unitary whole, and with each contributing sufficient
\nmaterial to constitute authorship. Extending this concept to Gen AI
\nwould acknowledge the collaborative nature of human-AI
\ncreativity.<\/p>\n
The advantages of this approach lie in its realism. Many
\ncreative processes now involve iterative human-AI interaction, and
\nrecognizing joint authorship would reflect this dynamic. It also
\npreserves human contribution, ensuring users retain rights even
\nwhen Gen AI plays an equal or even dominant role. Moreover, this
\nmodel offers flexibility, adapting to varying degrees of human
\ninput and accommodating diverse creative workflows. It would also
\nstill require non-minimal contribution by the human author, and
\nwould allow for both the Gen AI and human portions of the work to
\nbe safely protected by copyright and not exposed to potential
\nrelegation to the public domain.<\/p>\n
Yet, this solution is also not without challenges. Gen AI cannot
\nown anything, making “joint authorship” a legal fiction.
\nFurther, determining the threshold of human input required for
\nauthorship would introduce ambiguity. This approach would still
\nsuffer from the need for the human to demonstrate a non-minimal
\ncontribution, and courts would struggle to apply consistent
\nstandards. This uncertainty could lead to litigation difficulties
\nthat may be no better than the challenges of the present legal
\nsituation, with disputes over how much human authorship was present
\nand whether this amount was sufficient for the human to be
\nconsidered a joint author of the work. (That said, the minimum
\nthreshold for determining whether a contribution is sufficient for
\nparticipation in a joint work is well-trod legal terrain; in
\ntheory, this would be a significantly better and more certain area
\nto litigate versus the as-yet uncertain area of identifying the
\nboundary between copyrightable human-authored works assisted by Gen
\nAI and Gen AI-produced works ineligible for copyright.)<\/p>\n
Organic Development; Just Let it Be?<\/p>\n
The third path is to let the law evolve organically through
\njudicial decisions and Copyright Office guidance. Under this
\napproach, courts would continue to distinguish works that are
\n“substantially Gen AI” (ineligible) from those with
\n“sufficient human authorship” (eligible), refining
\nstandards case by case in a slow and laborious process.<\/p>\n
This approach offers several advantages. It avoids premature
\nlegislative overhaul, allowing nuanced, fact-specific evolution as
\ntechnology matures. Judicial flexibility ensures that courts can
\nadapt standards to emerging realities without locking in rigid
\nstatutory definitions. Importantly, this path preserves core
\nprinciples, maintaining human authorship as a doctrinal anchor
\nwhile accommodating incremental change.<\/p>\n
However, the drawbacks are considerable. Prolonged uncertainty
\nleaves businesses and individual creators navigating a legal gray
\nzone, exposing them to risk and inconsistent outcomes. This
\nambiguity could chill investment in AI-driven creativity, deterring
\ninnovation in sectors that depend on copyright certainty at a time
\nwhen these industries should be innovating and taking advantage of
\nthe opportunities offered by AI augmentation. Fragmentation is
\nanother concern, as different courts and agencies could adopt
\ndivergent approaches, creating a patchwork of standards that
\ncomplicates compliance.<\/p>\n
Conclusion and Next Steps<\/p>\n
Unlike many other hype-curve developments in recent years, Gen
\nAI truly seems to challenge copyright law at a fundamental level.
\nStrict interpretation of the human authorship requirement to
\nexclude Gen AI outputs seems increasingly impractical in a world
\nwhere creativity is often co-produced by humans and machines. No
\nsolution is perfect. Legislative reform offers clarity but risks
\ndoctrinal distortion; organic development preserves principles but
\nprolongs uncertainty.<\/p>\n
In the short term, the Copyright Office could refine its
\nguidance to provide clearer thresholds for human authorship,
\nreducing ambiguity for creators and businesses. It also could
\nconsider acknowledgement of a broader array of contexts in which
\nhuman authorship may be recognized and potential relaxation of
\napplication requirements. In the long term, Congress will likely
\nneed to act \u2014 whether through a works-made-for-hire model, a
\njoint authorship framework, a hybrid approach or something else
\n\u2014 to ensure that copyright law remains fit for purpose in an
\nAI-driven economy and can continue to incentivize creative
\nexpression.<\/p>\n
This article covers complex topics in insufficient detail.
\nFurther research is warranted in the following areas:<\/p>\n
\nHow do the philosophical foundations of our concept of
\n“authorship” inform our perception of human authorship in
\nthe AI era? Is our “author concept” a relative historical
\nnotion that could be updated to meet current practical reality, or
\nis it necessary for legal continuity?
\n
\nWhat exactly does the Constitution’s mandate to promote
\n“progress” in the useful arts mean? What types of
\nbehaviors and outcomes \u2013 aesthetically, economically,
\nculturally, and phenomenologically \u2013 are we intending to
\npromote with copyright?
\n
\nAre our copyright-reliant creative industries really so
\ncopyright dependent that AI disruption to copyright ownership
\ncertainty poses a serious economic risk? Is it possible that newer
\nbusiness models and technology are sufficient to render uncertainty
\nin copyright not so bad?
\n
\nSome works, like the rejected Allen, Kashtanova and Thaler
\nworks, seem worthy of copyright protection despite the role of Gen
\nAI, while others \u2013 an image, say, resulting from a three-word
\nprompt to Midjourney with nothing more, e.g. \u2013 do not. What
\nprinciples should guide the Copyright Office and law to better
\ndetermine which works are monopoly-worthy and which are not? Would
\nthere be harm in granting at least “thin” protection to
\nusers in connection with Gen AI outputs?
<\/p>\n
Footnote<\/p>\n
1. As outlined in a prior post<\/a>, Dr. Thaler, together with a
The content of this article is intended to provide a general
\ncollection of scholars in an amicus brief<\/a>, have advocated for the construal
\nof Gen AI works as works-made-for-hire, with Dr. Thaler as the
\nputative employer\/commissioner and his Gen AI “Creativity
\nMachine” as the putative employee\/contractor.<\/p>\n
\nguide to the subject matter. Specialist advice should be sought
\nabout your specific circumstances.<\/p>\n","protected":false},"excerpt":{"rendered":"Introduction A collision is on the horizon. The collision is between a strict
interpretation of the human authorship…\n","protected":false},"author":2,"featured_media":141776,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[18],"tags":[23,3,21,19,22,20,25,24],"class_list":["post-314073","post","type-post","status-publish","format-standard","has-post-thumbnail","category-united-states","tag-america","tag-news","tag-united-states","tag-united-states-of-america","tag-unitedstates","tag-unitedstatesofamerica","tag-us","tag-usa"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/posts\/314073","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/comments?post=314073"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/posts\/314073\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/media\/141776"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/media?parent=314073"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/categories?post=314073"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/tags?post=314073"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}