A.I. in the Second and Ninth Circuit
There are two circuits that form the powerhouses of creativity within the United States: the Ninth Circuit, which includes California (pumping movies out of Hollywood and software like video games out of Silicon Valley and LA), and the Second Circuit, which includes New York – the place you end up as an actor after Hollywood says no, but a recurring role on one of the 20 Law & Order shows says yes.
Though decisions regarding artificial intelligence (AI) can come from anywhere to move their way up the appellate system & finally come to their ultimate conclusion before the Supreme Court, eyes are trained on these two circuits in particular, especially after the recent New York Times v. OpenAI case in the Second Circuit.
With that in mind, there are a few decisions from these circuits that are particularly worthy of consideration, as well as Supreme Court precedent that they may dovetail with eventually (just keep in mind, this is by no means a comprehensive list of the draft picks or context for future Supreme Court cases on this issue).
Second Circuit Cases
Hachette v. Internet Archive
In Hachette, the Defendant, Internet Archive, published books whose copyrights were owned by Hatchette Book Group, Penguin Random House, HarperCollins, and Wiley, online. Their defense was that posting books online via their Controlled Digital Lending practice was fair use. The court disagreed, finding that the online copies of the books were harming the market for these books and operating as a substitute for the products of the copyright owner.
While this does not necessarily apply directly to AI, it does put in some modest guardrails that would provide a strong argument for finding copyright infringement if an AI reproduces a significant portion of a book or other copyrighted work. Will the infringement fall on the programmers of the AI producing the infringing prompt, or on the individual putting in a prompt that may have demonstrated some degree of negligence or intentional misconduct in eliciting an infringing response? That’s a good question…
NYT v. OpenAI
OpenAI is essentially the organization that kicked off the debate around copyright and artificial intelligence after it released the kraken/let the dogs out/jailbroke the genie in making ChatGPT available to consumers. In perhaps the largest development in AI and copyright since it became the issue that was living the motto of “move fast and break things,” a judge in the Second Circuit ruled in favor of OpenAI, dismissing the case because the news organizations involved were not able to show that OpenAI’s use of their copyrighted content did not cause them any actual injury: “I am not convinced that the mere removal of identifying information from a copyrighted work – absent dissemination – has any historical or common-law analogue.”
The judge further reasoned that it was unlikely that ChatGPT would produce a copyrighted article, or even a large enough part of it to argue for copyright infringement, as ChatGPT draws from millions of pieces of data – it’s essentially a “bed of nails” logic where no single point of potential copyright infringement passes the threshold to qualify that differentiates itself from Hatchette in not clearly producing an infringing amount of any one source.
Ninth Circuit
Naruto v. Slater
Naruto v. Slater was a case brought by PETA on behalf of a monkey that took a (legit cute) selfie after a wildlife photographer left his camera unguarded. Long story short – the Ninth Circuit found that the monkey DID have constitutional standing to bring a lawsuit, but did NOT have standing under the Copyright Act of 1976 as this law exclusively applies to humans. This quote encapsulates the heart of the court’s reasoning:
“The court in Cetacean did not rely on the fact that the statutes at issue in that case referred to ‘persons’ or ‘individuals.’ Id. Instead, the court crafted a simple rule of statutory interpretation: if an act of Congress plainly states that animals have statutory standing, then animals have statutory standing. If the statute does not so plainly state, then animals do not have statutory standing. The Copyright Act does not expressly authorize animals to file copyright infringement suits under the statute. Therefore, based on this court’s precedent in Cetacean, Naruto lacks statutory standing to sue under the Copyright Act... Based on this court’s decision in Cetacean and the text of the Copyright Act as a whole, the district court did not err in concluding that Naruto – and, more broadly, animals other than humans – lack statutory standing to sue under the Copyright Act.”
While artificial intelligence is not exactly an “animal other than a human,” considering it analogous to such, or crafting a new presumption in the lack of clarity of congressional intent for the purposes of copyright law, may not be too far-fetched for a Supreme Court that has not shied away from taking some big swings on controversial issues Congress can or will not address in the legislative process, i.e. Roe v. Wade. Though there are frankly several matters that have sorely disappointed me in the institution that leads the courts I practice law in, I will at least admit that I was impressed by the way the current Supremes pressed the arguments on both sides and the government attorney for AWF v. Goldsmith.
For a more detailed breakdown of Naruto v. Slater, read our article here.
Supreme Court
The two cases which, in my humble opinion, are the best to look at currently for guidance on the copyrightability of AI-generated content includes the copyrightability of products of an automated process in Burrow Giles Lithography Co. v. Sarony, and the guardrails set into place for fair use recently in The Andy Warhol Foundation for the Visual Arts v. Goldsmith.
Andy Warhol Foundation for the Visual Arts (AWF) v. Goldsmith
In this case, the court was faced with the issue of whether a change in meaning or message, which can be entirely subjective on the part of the recipient of said meaning or message, can qualify as “transformative.” This is of particular importance as it is often the heart (and most entertaining part) of a fair use analysis - while no single factor is dispositive, these cases often turn on this issue.
There were several additional issues, such as the commercial purpose of the infringing work and the market in no way punishing book sales when a more popular media form adapts a book’s storyline, but the big takeaway is that the changing of meaning or message of a work infringing on the copyright of another does not necessarily qualify for a finding that the infringing work is transformative. It is difficult to figure out exactly what effect this will have on AI litigation, but at the very least changes that are seen as relatively minor that do little other than alter the meaning or message of someone else’s original work (or at least it’s essential elements) will likely not qualify as transformative (and thus heavily-weighted in a fair use argument) without additional factors in its favor.
For a more detailed breakdown, catch our article on the certiorari briefs and oral arguments.
Burrow-Giles v. Sarony
In Burrow-Giles Lithographic Co. v. Sarony, once upon a time when photography was still a relatively new art form, the court decided on whether a photographer had the copyright on their photo despite the fact that the photo went through an automated process once the shutter was opened to produce the photo. This process was completely bereft of any human creativity that parallels putting a prompt into an AI to obtain a generated text or image.
Copyright law at the time did not explicitly require “human authorship” BUT, as per the Trade-mark Cases (1879), required that “works” be the product of human creativity as “the fruits of intellectual labor… founded in the creative powers of the mind.” Looking to photography, the court had to decide whether the process between opening the shutter and producing the negative cut the human creativity off from the final product sufficiently to say that the photographer did not qualify for the copyright.
In rendering its decision, the court reasoned that the photographer had created a “…useful, new, harmonious, characteristic, and graceful picture, and that said plaintiff made the same at his place of business in said City of New York, and within the United States, entirely from his own original mental conception, to which he gave visible form by posing the said Oscar Wilde in front of the camera, selecting and arranging the costume, draperies, and other various accessories in said photograph, arranging the subject so as to present graceful outlines, arranging and disposing the light and shade, suggesting and evoking the desired expression, and from such disposition, arrangement, or representation, made entirely by the plaintiff… the terms 'author,' 'inventor,' and 'designer,' as used in the art of photography and in the complaint, mean the person who so produced the photograph."
In considering the case, it can be reduced to somewhat of a logical formula:
Posing/lighting/choosing shutter speed = Writing a promp = A
Clicking shutter = Pressing ‘Enter’ after entering prompt = B
Negative = Generated Text or Image; X
A + B = X;
X HAS a copyright in photography, but NOT in artificial intelligence.
Read our breakdown of Burrow-Giles v. Sarony here.
If the Supreme Court were faced with the question of AI outputs and copyright and based their decision on this precedent, it is easy to see how AI could be considered analogous to a tool like a camera: you start with human creativity, put it through an automated process, and the person who used the automated process, as long as they had sufficient creative input and a degree of control over the output, would likely qualify for a copyright. However, the automated process in AI may be differentiated from photography, or legal protections for creators whose work is used in the models of generative AI may be formed at different points, such as the information that is used to train the models.