The virtual dog that didn’t bark

Amid the barrage of headline-grabbing Supreme Court decisions last term, it was easy to miss the import of something the Court *didn’t* do. On March 2, 2026, it denied review of Thaler v. Perlmutter, a DC Circuit opinion holding that a virtual painting that was said to be produced solely by an artificial intelligence (AI) was not a work of authorship protected by the Copyright Act. 

At a glance, it seems easy to understand why the Court chose not to review Thaler. For decades, Copyright Office guidance has stated that only humans can be authors of copyrightable works of authorship. Courts have backed up the Office, swatting away challenges to the human authorship requirement brought on behalf of wildflowers, crested macaques, and a new age deity. Thaler could be seen as no more than a simple application of this rule: AIs are not human, so works they produce cannot be copyrighted. 

This account explains why Thaler itself, and the Court’s denial of review, occasioned little attention. A closer look at the DC Circuit’s opinion, though, reveals a more complicated picture than this narrative suggests.  

Thaler’s core argument is that, as a matter of statutory interpretation, copyright law contemplates only humans as authors. It supports this view by detailing a litany of instances in which provisions of the Act make sense only if the term “author” means “human.” Copyright duration is keyed to human lifespan; AIs do not die. The statute specifies surviving heirs to inherit an author’s copyright termination interests; AIs have no widows or children. Copyright transfers require signatures to be valid; machines can’t sign things. Protection of unpublished works operates irrespective of an author’s nationality or domicile; machines have neither. 

This too seems straightforward, until you consider that at least since the early 1900s, copyright law has included work-made-for-hire provisions. Under these provisions, the entity that hires an employee (or, in some instances, an independent contractor) who produces a copyrighted work is considered the author of that work. Yet employers are frequently non-human entities such as corporations or state governments. 

So if Coca-Cola or the State of Georgia can be “authors” within the meaning of the Copyright Act, that puts the DC Circuit’s central line of reasoning—and the entire human authorship requirement—on weaker footing. If the Act bestows authorship on nonhuman entities like juridical persons, then it can’t be the case that only humans are authors. The DC Circuit did not ignore this argument entirely. It sought to distinguish the point by observing that the Copyright Act says that hiring parties are “considered as” authors, rather than saying they “are” authors. 

But this is a minor linguistic wrinkle with which to dismiss a point that otherwise seems to torpedo the factual assertion that the Copyright Act limits authorship to humans. And even though corporate entities may only be “considered as” authors, the Act still includes variations that adapt its provisions on duration, termination, and transfer—the very ones invoked to support the human authorship requirement in Thaler—to these nonhuman legal fictions.  

The reality is that the Copyright Act is shot through with instances of nonhuman authorship. So does this scuttle the human authorship requirement? Certainly not. The Copyright Office and the courts that have affirmed its regulations on this matter were all well aware of the work-made-for-hire provisions when they did so. The point, rather, is that a compelling reading of the Copyright Act has to confront the challenge this creates for the human authorship requirement head on.  

A compelling reconciliation of the Copyright Act’s work made for hire provisions with the human authorship requirement would begin with the rather obvious point that any one statutory exception doesn’t reflexively warrant another. So the fact that the Act contemplates one kind of nonhuman author doesn’t require that it also needs to bestow this status on AIs—or monkeys, vines, or newfound deities for that matter. 

Instead, the presence of corporate authors invites exploration of the rationale behind their inclusion alongside humans as authors. One is the practical reality that if all human creators of a collaborative work like a film or a computer program shared in the copyright, administering fractionated ownership of that work would be impossible. Another is that even though the Act delegates authorship to corporate entities, the works at issue were always (pre-AI, anyway) generated by human creators.  

These points begin to explain why it makes sense to consider corporate or state hiring parties to be authors while excluding AIs (and animals, and plants, and divinities) from that category. But the Court was wise to leave the issue alone for another reason as well. The Copyright Office defended its position on constitutional as well as statutory grounds, arguing that the term “authors” in the Constitution’s IP Clause mandated human authorship.  

By affirming the requirement based only on the Copyright Act, though, the DC Circuit set up Congress to amend the statute to allow AI authorship if it sees fit. A constitutionally grounded holding would have foreclosed this possibility. Declining to review Thaler leaves the Court open to settling the constitutional question when it’s squarely presented. 

The DC Circuit’s defense of the human authorship requirement in Thaler may have been imperfect, but it got the statutory outcome right and wisely punted on the constitutional issue. And given the genuinely consequential issues bedeviling the Court, why let the perfect be the enemy of the good?