AI is Not Creative Per the USCO and the Courts – And That’s a Good Thing

Recently, Wen Xie argued on IPWatchdog that the U.S. Copyright Office (USCO) and the U.S. Patent and Trademark Office (USPTO) have reached different conclusions regarding “the creative and conceiving capabilities of machines,” which leads to intellectual property (IP) law being self-contradictory. According to Xie, the USCO presumes that artificial intelligence (AI) is creative, while the USPTO does not reach a similar conclusion regarding AI inventorship.

I disagree. Firstly, the USCO did not find AI-generated works to be “creative.” It did the exact opposite; just like the courts, the USCO is embracing traditional copyright causation and its human-centeredness, declining protection of AI-generated works. Secondly, patents and copyright are doctrinally and conceptually different. Thus, even if as a matter of patent law AI was “inventive,” or as a matter of copyright it were “creative,” they would have no bearing on each other. Thirdly, I argue that there is no good reason for the law to change. AI is not deemed creative – and it should not be.

Recent Copyright and AI Jurisprudence

Contrarily to Xie’s argument, the USCO’s recent guidance, the Copyright Review Board’s decisions in Théâtre D’opéra Spatial, A Recent Entry to Paradise, Zarya of the Dawn, and the district court judgment in Thaler v. Perlmutter, all found that human creativity is the sine qua non of copyrightability, in all cases refusing to register works lacking human creative involvement or control. AI-generated works are unprotectable not because it was AI and not the human who supplied originality – they are unprotectable because the purported human author did not make an original contribution to the work in question. Put simply, the necessary causal link between the human and the work was broken.

In A Recent Entry, the Board found that the courts “have uniformly limited copyright protection to creations of human authors,” and relied on Urantia Found. v. Kristen Maaherra, to find that human creativity was necessary. Thus, the reason why AI-generated content that is more than de minimis should be explicitly excluded from registration applications is not because AI is capable of originality or creativity. It is because this output is not original to the purported human author, who neither created the relevant elements of the work independently, nor supplied the necessary creativity.

This does not mean AI is deemed legally “creative” or capable of conceiving things. We are speaking of the insufficiency of the human’s contribution, not the AI’s transcendence. All kinds of things, animals, and events can break a causal chain, undermining one’s claim to obtain property, such as copyright. But, in principle, a human can take an unoriginal, source material and transform it in a creative way, to make one’s own. This was the reasoning underlying the Zarya of the Dawn decision.

As I have already argued, all of copyright’s conceptual building blocks—the idea-expression dichotomy, originality, authorship, and the concept of a protectable work—are premised on the notion of human creativity. Autonomous creations of AI systems, consequently, fall outside of copyright’s purview, being akin to ideas, facts, or subject-matter predicated by technical considerations, rather than authorial creativity.

This reasoning is confirmed by each of these decisions, including the Théâtre D’opéra Spatial. The Board reassured that “human authorship is a bedrock of copyright” and thus, if someone does not author a particular subject-matter or does not supply original, creative expression, they cannot own it.

Patents and Copyrights are Different: The USPTO and USCO Need not Reach the Same Conclusion 

Xie argued that the two Offices should reach similar conclusions regarding the protectability of AI outputs, especially in regard to how “the creative and conceiving capabilities of machines” are assessed in each domain. While this may be so as a matter of policy – a point returned to below – doctrinally, patents and copyrights speak different languages. One is of ideas, novelty, and inventions – the other of creative expressions, originality, and works. Thus, in principle, both the causal inquiry and the conclusions regarding protectability can differ.

While authors and courts sometimes have spoken on this matter recklessly, the Supreme Court laid out the difference first in Baker v. Selden, and re-emphasized it a century later in Mazer v. Stein. “Unlike a patent, a copyright gives no exclusive right to the art disclosed; protection is given only to the expression of the idea—not the idea itself.” In this way, each type of intellectual property right exists in different types of creations, which arise in a different ways, and have different requirements for protection. “[C]opyright protects originality rather than novelty or invention,” which is the domain of patents, said the Court in Mazer,Finally, the Court found that protection in one domain neither entails nor precludes protection in the other.

Indeed, what the Court made clear in Feist v. Rural, is that authorial works need to be original; that is, both created independently and “creative.” Other cases, such as Bleistein v. Donaldson, spoke of original expressions as “personal reaction upon nature,” where the author contributes “something recognizably his own,” per Alfred Bell. All of this is missing from the language of patent law – and, at the same time, makes protectability of works “autonomously” generated by AI at odds with copyright doctrine and theory, as I have argued at length and in short form.

Is AI Creative?

Fundamentally, there are two questions here. One is whether AI is “creative” in law, the other is whether it is creative in fact. As already explained, neither the courts nor the USCO have found AI to be creative as a matter of copyright law. As a reminder, lawyers operate with law’s meaning of terms, in a shared fantasy of sorts, an ontological space different from the natural one. That is, they care little what the philosophers and neuroscientists say about creativity, but what the law takes it to be. The other, related question is whether AI is creative in fact. This is rather complicated because we do not really know what it means to be creative. The point is, however, that they are not creative in the same way we are, or at least there is no good reason to think so. As Chomsky said:

“‘Think’ is an informal term of human language. It means what it means in its normal usage and its normal usage is applied to human activity. So, the question whether machines can think is like asking whether submarines can swim. What metaphor do we decide to pick? That’s a meaningless question… “

This picks up on what Alan Turing wrote 70 years ago, a line of argument which has recently become fashionable again. We cannot meaningfully use anthropocentric concepts, grounded in human reality, and apply them to machines. AI does not have such grounding, it is a “syntax without semantics.” It is neither embodied, social, nor subjective the way we are. Is Naruto creative? Just like with AI, the output appears creative, and whatever our philosophical assessment, for now, the law does not deem it as such.

Should Copyright Deem AI Creative?  

Currently, only human authorial contributions are copyrightable. Human creators are the subjects of copyright law, while works, that is expressions of human intellectual conception, are the object. After all, the constitutional justification for copyright law is human welfare, and it is difficult to imagine how the legal standard could be altered to include non-human, apparently creative contributions.

The recent Federal Trade Commission Roundtable on Creative Economy and Generative AI (followed by the Comment submitted to the USCO) opened with Chair Khan expressing concern than AI can “significantly disempower creators and artists” and concluded with Commissioner Bedoya saying that it is “wildly unrealistic and frankly dangerous to think that [AI] will break every bound of human creativity.” The post-humanist copyright moment is in retreat.

Indeed, setting aside all of the doctrinal differences between copyrights and patents, if there is a good example of the “historic kinship” of which the Supreme Court spoke in Sony Corp. of Am. v. Universal City Studios, it is precisely that the Intellectual Property Clause is supposed to benefit social welfare – with the People, their progress, learning, creations and inventions, being an ultimate referent. It is not surprising that, just as copyright authorities reaffirmed the place of human authorship, the Supreme Court in Thaler v. Vidal reaffirmed that an inventor must be human. In both cases, the answer comes from simple statutory interpretation, colored by the wider constitutional framework, and a basic causal inquiry. It is the person who contributes an original expression or an invention or discovery which deserves protection in the form of an IP right, and that person must be human.

Copyright law is an especially good example where the claims of AI creativity or calls for the recognition of broad AI personality may effectively detract from the common good: whether of the public domain, or of rights of human authors, both of which the IP clause is supposed to protect. Not long ago, in addition to IP provocateurs, authors claimed AI should be granted constitutional rights. Today, it is increasingly seen that there is a zero-sum element, that humanistic ethics should guide legal and political development, and that our culture is ultimately at stake. The least we can do is to preserve the shape of copyright’s concepts and its causal inquiry.

Editor’s note: Wen Xie sent the following statement in response to this article:

“I would like to thank Mr. Blaszczyk for his insightful and well-written article.  Since the publication of my article, I have appreciated many comments from copyright experts, including Mr. Blaszczyk and Professor Devlin Hartline at the Hudson Institute, expressing their views on the matter. IPWatchdog provides a valuable forum for concerned experts and practitioners within the legal community to discuss our insights from different perspectives and fields during this important time of critical emergent technologies and in light of the President’s directive that our major IP offices collaborate to advise the Administration on the treatment of AI in the innovative ecosphere. Please read the works of Professor Hartline for more on copyrights in the modern digital age.”