AI as Author: Thaler v. Perlmutter Now Before the DC Circuit

by Dennis Crouch

The leading case on copyrightability of AI created works is now pending before the Court of Appeals for the District of Columbia. The case, Thaler v. Perlmutter, No. 23-5233 (D.C. Cir. 2024), centers on Dr. Stephen Thaler’s attempts to register a copyright for an artistic image autonomously generated by his AI system that he has named the “Creativity Machine.” The U.S. Copyright Office refused registration on the basis that the work lacked the required human authorship. Thaler filed suit challenging this determination.  The parties have now filed their briefs, along with one law professor amicus brief in support of Thaler.

Stephen Thaler developed an AI system he calls the Creativity Machine. Using this system, he autonomously generated a 2-D artwork titled “A Recent Entrance to Paradise.” In November 2018, Thaler filed an application with the Copyright Office seeking to register a copyright in this AI-generated work.

The Copyright Office refused registration in August 2019, stating that it “cannot register this work because it lacks the human authorship necessary to support a copyright claim.” After requests for reconsideration were denied, Thaler filed suit against the Copyright Office in June 2022, arguing that “requiring human authorship for registration of copyright in a work is contrary to law.”

In August 2023, the district court granted summary judgment in favor of the Copyright Office. The court held that under the Copyright Act, “United States copyright law protects only works of human creation” and found no basis to conclude that copyright extends to works generated autonomously by AI without human involvement. Thaler appealed to the D.C. Circuit.  Briefs have been filed, but the court has not yet set a date for oral arguments.

• Thaler Opening Brief
• Thaler Amicus Brief
• Thaler Copyright Office Brief
• Thaler Reply Brief
• Thaler Appendix to Appeal

Arguments on Appeal:

In the parallel patent case, the appellate court found that the Patent Act expressly requires a human inventor based upon the definition of an inventor as an “individual.” But the copyright statute is different – it does not appear to expressly require a human author. And, in fact, under the work-made-for-hire doctrine, corporations and other non-human entities are regularly regarded as the legal author of created work (although there is an underlying human creative force).  At the same time, the copyright law also does not expressly state that a copyright can persist even with no human originality “in the loop.”  This largely leaves the courts to decide whether or not to presumptively require a human.

Thaler’s main argument is that the Copyright Act does not expressly require human authorship for a work to be copyrightable. The statute refers only to “original works of authorship,” 17 U.S.C. § 102(a), without mandating that authors be human. As I did above, Thaler points out that the Act already accommodates non-human authors in the form of corporate entities under the work-made-for-hire doctrine. He contends there is no principled reason to treat AI systems differently under the statute. Thaler argues this interpretation is most consistent with the constitutional goal of the Copyright Clause to “promote the progress of science and useful arts.” U.S. Const. art. I, § 8, cl. 8. Incentivizing AI-generated works, in his view, serves that purpose.

The Copyright Office, in response, marshals the statutory text as evidence that Congress primarily contemplated human authors, or at least a human originator. It cites the numerous references in the Act to an author’s “life,” “death,” “children,” and “widow/widower” as making sense only in the context of human authorship.  The Office also relies heavily on Burrow-Giles Lithographic Co. v. Sarony, which rooted authorship in “intellectual invention” and “mental conception.” 111 U.S. 53, 58, 60 (1884).  Of course, our AI tools easily pass the requisite test, so long as they do not include a human litmus test.  While Burrow-Giles held a photograph taken via camera (a machine) to be copyrightable, the crux of the decision focused on the photographer’s choices that were creative enough to warrant copyright. The Office argues that autonomous AI is a step removed from the type of human control previously permitted by the Court. The Office also cites its own longstanding practices of requiring human authorship for registration.

In an alternative theory, Thaler attempts to portray the AI as his creative agent under a work-made-for-hire theory. But, in my view, the Copyright Office persuasively counters that the work-made-for-hire doctrine, properly understood, would require a contractual relationship–and up to now, a human cannot “hire” or “commission” a machine.

The supporting amicus brief amplifies some of Thaler’s themes. It frames the issue as whether copyright law will remain stuck in the past or flexibly adapt to new technologies – as it has in the past with photography, sound recordings, and other once-novel media. The amici emphasize the economic and policy benefits of granting protection to AI works to spur investment and maintain American leadership in creative innovation. They also highlight the global context, noting that many other jurisdictions are recognizing copyrights in computer-generated works even absent direct human authorship. Amici further argue that the Copyright Office’s human authorship rule lacks any statutory basis and that, functionally, AIs are not so different from corporations when it comes to non-human creation.

Others have argued that authorship should be seen as a required causal act leading to copyright, and in American law such an act requires an actor with personhood status. From that perspective, the idea that a program, acting independently, could qualify as an author challenges the fundamental nature of personhood.  But, this causation element is not developed in either the code or the case law. Shyamkrishna Balganesh, Causing Copyright, 117 Colum. L. Rev. 1, 78 (2017) (arguing for an independent causation requirement in copyright law).

The issue of AI copyright is now squarely before the courts, although it has been brewing for many decades.  A good starting point in the analysis might begin with the 1979 CONTU final report that was commissioned by congress who had established the National Commission on New Technological Uses of Copyrighted Works (CONTU) in 1974. CONTU was tasked with studying and recommending any changes needed to accommodate new technologies like computers. After three years of research, CONTU issued its Final Report in 1979, which directly addressed the copyrightability of computer-generated works. The report unanimously concluded that “works created by the use of computers should be afforded copyright protection if they are original works of authorship within the Act of 1976.” However, the report focused only on situations where some human authorship is involved, with the computer merely an “assisting instrument” used by the human author, akin to a camera or typewriter.  In his 1993 article, Harvard Law professor (and CONTU member) Arthur R. Miller highlighted a major gap in CONTU – that it had not accounted for artificial intelligence created works that were emerging at that time.  Miller argued that “if the day arrives when a computer really is the sole author of an original artistic, musical, or literary work (whether a novel or a computer program), copyright law will be embracive and malleable enough to assimilate that development into the world of protected works.” Arthur R. Miller, Copyright Protection for Computer Programs, Databases, and Computer-Generated Works: Is Anything New Since CONTU?, 106 HARV. L. REV. 977 (1993).

Although not an AI case, one of the more interesting decisions on the subject, Urantia Foundation v. Maaherra involves a book purportedly authored by celestial beings. 114 F.3d 955 (9th Cir. 1997).  The Ninth Circuit ultimately ruled that while the copyright laws do not explicitly demand “human” authorship, they do require a demonstrable element of human creativity.  Still, the book itself was copyrightable as a compilation by the humans who organized and transcribed the celestial messages. This decision provides a crucial precedent, suggesting that for computer-generated works, similar principles could apply—where the role of a human, possibly the programmer or the user interacting with the software, would be essential in establishing copyright claims, as they introduce the requisite creativity and organization into the final product.