Dr. Thaler Is Right, In Part

When Dr. Stephen Thaler asked the U.S. Supreme Court to reconsider the human authorship requirement for copyright protection last month, many observers dismissed the effort. Thaler's claim—that his generative AI system should be recognized as the author of its own outputs—has been consistently rejected by courts and the Copyright Office. On its face, Dr. Thaler's petition for a writ of certiorari looks untenable: under current law, "authorship" remains a human concept.

And yet, Dr. Thaler is right, in part. While his claim that an AI system should itself be treated as a legal author is (currently) unconvincing, his broader complaint—that the Copyright Office's conception of "human authorship" is too narrow—deserves consideration. The law should continue to require human authorship, but it should also recognize that creativity mediated through generative AI can and often does meet that standard.

Dr. Thaler Challenges the Human Authorship Requirement

Thaler v. Perlmutter arises from Dr. Thaler's application to register a visual work generated by his "Creativity Machine" with the U.S. Copyright Office. The Copyright Office denied the application, reasoning that the work lacked the "human authorship" necessary for protection under the Copyright Act because the Creativity Machine is a generative AI system programmed to autonomously produce images. Both the U.S. District Court for the District of Columbia and the D.C. Circuit affirmed the decision, concluding that copyright law protects only works of human creation.

Applicable legal precedent traces back to the nineteenth century. In Burrow-Giles Lithographic Co. v. Sarony (1884), the Supreme Court upheld copyright protection for a photograph of Oscar Wilde, reasoning that it was "a product of the author's intellectual invention" rather than a mere mechanical reproduction. Implicit in the reasoning was that the "author" was human—a creative agent capable of using a machine to generate original expression.

More recently, in the so-called "monkey selfie" case Naruto v. Slater (9th Cir. 2018), the Ninth Circuit held that non-human animals cannot own copyrights and cannot be authors. The court explained that the Copyright Act does not contemplate non-human animals as authors and presupposes a human author. Similarly, in administrative decisions and its Compendium of U.S. Copyright Office Practices (3d ed.), the Copyright Office has reaffirmed that it "will not register works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author."

Dr. Thaler's petition to the Supreme Court challenges this position. He argues that the Creativity Machine—even operating autonomously, as he described—should be recognized as the author of its images, and that he as its owner should be recognized as the owner of the resulting copyrights.¹

On that point, the lower courts were right to reject Dr. Thaler's claim under current law. Our current understanding is that machines are not sentient agents; we believe currently lack consciousness, intent, agency and, thus, the capacity for authorship as the law understands it. Granting authorship to a machine would potentially upend basic tenets of copyright, from ownership to duration to transfer.

Still, Dr. Thaler's case exposes a genuine tension: while the law correctly requires human authorship, the Copyright Office's enforcement of that principle has drifted toward overbreadth.

The Copyright Office's Overcorrection

In its recent guidance on "Works Containing Material Generated by Artificial Intelligence" (March 2023), the Copyright Office took a hard line: if a work's "traditional elements of authorship were produced by a machine," then the work lacks authorship and is not registrable or protectable. Only the portions "conceived and executed" by a human author may be registered.

In practice, this position has resulted in a number of well-publicized denials of registration. When artist Kris Kashtanova registered their graphic novel Zarya of the Dawn, the Office later revoked the registration in part, finding that the images produced using Midjourney lacked human authorship. The decision recognized Kashtanova's creative arrangement of images and text but excluded the images themselves. According to the Office, the images were the visual output of the generative AI technology as opposed to the product of human authorship, despite evidence of Kashtanova's comprehensive involvement in the production and editing of the images. Similarly, the Office refused registration of Jason Allen's Midjourney-generated digital image Théâtre D'opéra Spatial. To create the image, Allen had entered 624 distinct prompts and further edited and upscaled the image using Photoshop and Gigipixel AI. However, the Office found that because "Midjourney does not treat text prompts like direct instructions," the traditional elements of authorship were executed by the machine rather than Allen himself. The Office noted that Allen could have disclaimed the machine-generated portions of the work and sought registration for the human contributions, but Allen did not do so.

The Office's approach, while administratively neat, is too blunt. It treats "AI-generated" as synonymous with "AI-authored," collapsing important distinctions. Most creators using generative tools are not ceding authorship to the machine; they are directing, refining, and curating outputs through a process that remains human.

This is particularly evident in "prompt engineering," an emerging art form in itself. Skilled users like Jason Allen often iterate dozens or hundreds of times, adjusting phrasing, style parameters and seed values to achieve a desired effect. They select among results, modify them, and integrate them into broader works. This process resembles—and can even exceed—the level of creative control exercised by photographers, editors, or art directors in traditional media.

If the photographer in Burrow-Giles was deemed the author of a photograph taken by a camera—a mechanical device that merely recorded the scene the human envisioned—then a human who uses an AI model as a tool of expression should likewise be recognizable as an author, at least in some cases. The tool may be complex, but the authorship often remains human.

Automatic Public Domain Status Is Too Harsh

Under the Copyright Office's current position, any work lacking sufficient human authorship is not protected by copyright and therefore enters the public domain upon creation. That might seem appropriate where a human plays zero creative role, but it is overly punitive where the person's creative contribution is substantial but intertwined with contributions from a generative AI system.

This automatic forfeiture creates perverse incentives. It encourages creators to obscure or minimize their use of AI tools. It also risks undermining investment in AI-assisted creative processes, as the output would be immediately free for anyone to copy.

Moreover, "no protection" is not the only doctrinal option. Courts have long recognized frameworks such as joint authorship, derivative works, works-made-for-hire,² and compilations—one or more of which might accommodate human-machine collaboration without granting authorship to a machine. A balanced solution would recognize that the human directing and curating AI output is the author, while the generative AI platform is simply an assistive instrument.

A More Coherent Approach

A revised framework should include three principles:

1. Human Creativity
   Copyright should attach when a human imposes meaningful human creativity on the expressive elements of the work. Such control or contribution may occur through the design of prompts, iterative selection, post-processing or other involvement—any process through which the human authentically shapes the final expressive result.
2. Instrumental View of AI
   Generative AI systems should in most cases be regarded as instruments, akin to cameras, synthesizers, or design software. The mere use of a generative AI system should not automatically negate authorship in the output, so long as the human user contributes to the expression through substantive choices and direction.
3. More Permissive Registration Process
   The Copyright Office does not need to burden applicants with identifying exactly which parts of their creative works are attributable to generative AI tools, just as it does not require disclosure of other software tools used in the creative process. If a defendant were to wish to challenge the (human) authorship in a particular work, the better place for that type of dissection would be infringement litigation.

These principles align with both the constitutional purpose of copyright—"to promote the progress of science and useful arts"—and the pragmatic need to allow the law to adapt to new creative technologies. The alternative, a blanket exclusion of AI-assisted works, could effectively freeze copyright in an outdated view of authorship and potentially chill the use of AI and stifle its potential.

Dr. Thaler's Role in the Debate

To his credit, Dr. Thaler has escalated this issue. By pressing his claim to the Supreme Court, he has compelled legal and cultural institutions to confront what "authorship" means in an age when machines can contribute to creative processes in novel ways. His arguments may fail in court, but they should succeed in drawing attention to the gaps in current policy.

If the Creativity Machine operates autonomously without human selection, curation, or control, Dr. Thaler's own case may be an extreme example. However, most real-world creators are not operating at that extreme. They are collaborating with and controlling generative AI tools. In that sense, Dr. Thaler may be wrong about his own claim, but right that copyright law needs to evolve to reflect the reality of human–machine creativity.

Footnotes

1 It is interesting to note that, in his appeal to the District Court of the Copyright Office's denial of his application, Dr. Thaler introduced new facts, namely, that he "provided instructions and directed his AI to create the Work," that "the AI is entirely controlled by [him]," and that "the AI only operates at [his] direction." The District Court refused to consider these newly-asserted facts, because it was bound only to review the decision of the Copyright Office based on the record that was actually before the Office at the time of the Office's final decision. It is unclear whether the new facts would have mattered. Based on the Office's current position regarding AI outputs, the application likely would still have be denied. However, this information suggests that there may have been sufficient human authorship present in the applied-for work.

2 Dr. Thaler has asserted that work-made-for-hire is in fact an appropriate framework, with the Creativity Machine as the putative employee/creator and himself as the employer/commissioner. This position has received recent favorable attention; on October 31, 2025, a group of scholars filed an amicus brief in support of treatment of AI-generated works as works-made-for-hire, among other points. Ivan Moreno, Academics Back IP Rights For Generated Art At High Court, IP Law 360 (Nov. 3, 2025) https://www.law360.com/ip/articles/2406913.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.