On March 18, 2025, the United States Court of Appeals for the District of Columbia Circuit affirmed the denial of Stephen Thaler’s application to register A Recent Entrance to Paradise, a purportedly machine-authored work, holding that the Copyright Act of 1976 “requires all eligible work to be authored in the first instance by a human being.” This is a welcome result both as a matter of doctrine and principle, with the D.C. Circuit attempting to preserve an appearance of copyright’s legitimacy in the post-human era. At the same time, the requirement’s practical effect will be thin – it “does not impede the protection of works made with artificial intelligence,” as Judge Millet wrote. It comes down to nominally taking as author “the person who created, operated, or used artificial intelligence—and not the machine itself.”
Thaler v. Perlmutter is an easy case making good law – or at least that’s the way D.C. Circuit wrote the opinion, engaging “traditional tools of statutory interpretation” to show that “author” refers “only to human being.” The court did not address the question of authorship’s constitutional bounds nor whether Thaler could be considered an author, an argument waived by virtue of Thaler’s original submission before the United States Copyright Office (USCO), which the appellant rather regrettably and unsuccessfully attempted to introduce during judicial review. Whether one can author outputs of prompted AI will be considered in another test case, Allen v. Perlmutter, and is “neither here nor there” in Thaler.
To recount the facts and history, Thaler’s registration application designated the “Creativity Machine” as the work’s sole author and himself as the owner. Throughout his dealings with the USCO, Thaler confirmed the work lacked “traditional human authorship” and was “autonomously generated by an AI.” This led to the refusal to register for want of human authorship, a decision upheld by the USCO’s Review Board, and later the District Court. The USCO has long maintained that both copyrightability and registrability require human authorship, with the current edition of the Compendium stating that the Office will refuse to register works if a human being did not “create the work,” including regarding works “produced by a machine.” As an aside, this policy has been reflected in the 2023 Registration Guidance, the 2025 Report on copyrightability of AI-generated works, and in several registration decisions. Human authorship was held to be a copyright subsistence prerequisite by the District Court, where Judge Howell called it copyright’s “bedrock requirement.”
The D.C. Circuit grounded its affirmation of the statutory requirement of human authorship in textual analysis and, rather strikingly, an inference that the Congress adopted the USCO’s “longstanding interpretation” of authorship when passing the 1976 Act. (Note that the Office formally adopted the requirement three years prior). As a matter of textual analysis, the court found that many of the Act’s provisions “only make sense” if an author is a human being, and so the best reading includes the requirement that “authors must be human, not machines.” Namely:
- Since copyright vests initially with the author, an entity which cannot own property cannot be an author.
- Duration of copyright is limited to author’s lifespan or a “period that approximates how long a human might live.” Yet, “machines do not have ‘lives’ nor is the length of their operability generally measured in the same terms as a human life.”
- While the Act’s inheritance provision speaks of author’s “widow,” “widower,” “children,” and “grandchildren,” the court observes that machines “have no surviving spouse or heirs.”
- Copyright transfers need to be in writing and signed, while machines “lack signatures” and the “legal capacity to provide an authenticating signature.”
- Although unpublished works are protected regardless of the author’s nationality or domicile, machines possess neither.
- Joint works require authorial intent to merge contributions. While “[a]uthors have intentions,” machines “lack minds and do not intend anything.”
- The court shows that whenever the Act discusses machines, the “context indicates that machines are tools, not authors.” Pointing to the definition of computer programs, which treats the words “machine,” “device,” and “process” synonymously, the court cites numerous provisions “consistently” taking machines as “mechanisms that assist authors, rather than as authors themselves.”
These provisions are taken to “collectively” prove, together with the 1976 Act’s structure and design, humanity is a “necessary condition” for authorship. The statute’s natural meaning, we learn, is reinforced by the last sixty years of USCO’s reports and regulations. The court proclaims that the “Copyright Act makes no sense if an ‘author’ is not a human being.” However, while it is true that to substitute “machine,” “monkey,” “forest,” “celestial being,” for “author” makes little sense, the analysis limiting authors to natural persons is not obviously convincing. (The court’s glossing over the references to, or exclusions of, work for hire from e.g., duration and inheritance provisions does not quite remove doubt, either).
For over a century, the U.S. legislation has explicitly recognized non-human authors (17 U.S.C. § 201(b); 17 U.S.C. § 26 (1909)). In so-called works made for hire, the employer is by default “considered” both the owner of “all of the rights comprised in the copyright” and “the author for purposes of this title.” In those cases, the human actual creator is never considered the author; this is why courts sometime contrast the author-in-law with the author-in-fact by calling the latter “‘author’ in the colloquial sense.” Unlike the laws of Canada or most European jurisdictions, both the 1909 Act and the 1976 Act treat the issue as one of “authorship and not of transfer of rights; the employer is presumed to be the author initially and not by virtue of a post-creation transfer.” This deemed authorship, a legal fiction, was historically important in institutionalization of corporate liberalism – so much so that most valuable copyrights belong to corporations, commentators call copyright an “engine of inequality,” and argue that the traditional concept of authorship had been abandoned long before AI.
By all accounts, the D.C. Circuit rightly rejected Thaler’s arguments that the machine could be considered his employee (which, perhaps surprisingly, was not devoid of scholarly support) and that § 201(b) allows to copyright creations of non-humans. However, compared with the District Court, the Circuit court offered rather labored exegesis, claiming that the word “considered” does “critical work,” showing that the statute merely “transfer[s] instantaneously” the “copyright and authorship protections,” and that Congress was “careful to avoid using the word ‘author’ by itself to cover non-human entities.” There is something truly scholastic in pondering the realities involved in legal fictions. That said, the Circuit court’s analysis does not disrupt the doctrine in any practical sense; it leads to a result which allows to prevent explicitly equating a human with a machine; and allows to proclaim a humanistic victory of the law, even though corporations are the law’s people when it matters for them. Finally, as Judge Millet emphasizes, the human authorship requirement “does not impede” the protection of AI-generated works – its role is rather symbolic.
As if attempting to convince utilitarian readers to anthropocentrism, the court wholeheartedly embraces the incentive theory, repeatedly speaks of the public benefit, and even quotes the Google v. Oracle dismissal of the “special reward to the author” – rather curiously, both given several SCOTUS dicta embracing the concept, and the contention that “[a]uthors are at the center of the Copyright Act,” which appears a page later. Perhaps quoting the case infamous for sidestepping the question of copyrightability is not accidental. In its analysis, the court does not speak of originality, downplays registration applications’ rejections, noting that some disagree with the USCO’s decisions, and reassures that the human authorship requirement is a mere formality, which may come to be reevaluated by Congress in the future – for example, once AI becomes advanced enough to respond to economic incentives, like in Star Trek – but not by the courts in the meantime.
In conclusion, Thaler v. Perlmutter affirms the central place of the human being in copyright’s doctrinal architecture. The D.C. Circuit proclaims that machines are mindless tools which do not need incentives nor possess subjectivity, downplays the importance of corporations in modern copyright law, and emphasizes the public benefit. The court makes equally clear that neither this decision nor the human authorship requirement will pose practical obstacles to the growing importance of AI-driven cultural production. In this way, the judgement continues the legacy of Bleistein v. Donaldson Lithographing Co. and Feist v. Rural, known for exalting authorial creativity while in fact lowering standards, focusing on economic growth, and in Bleistein’s case, hiding the work for hire behind personality language. I expect the upcoming Allen case to take this up even further – both affirming the human authorship requirement as a matter of formality and depriving it of any substantive weight.
The Author is preparing a law review article on this topic titled “Posthuman Copyright: Copyright, AI, and Legitimacy.”
________________________
To make sure you do not miss out on regular updates from the Kluwer Copyright Blog, please subscribe here.
