In the past few weeks, I’ve discussed at some length my skepticism about AI-generated content and what legal protections might be available to tech-savvy users who rely on AI to create text, images, and even music.
The debate has raged for as long as AI has been on the public’s radar, with opinions split about whether AI heralds the “next big thing” or whether it’s all “sound and fury, signifying nothing.”
In a huge blow to AI enthusiasts and artists who lean on AI for their creations, the US Copyright Office issued a decision on February 21st, 2023 (click the link to view the entire PDF file), unequivocally and categorically rejecting the validity of copyright for AI created works.
What prompted the Copyright Office ruling?
In 2022, comic creator Kristina Kashtanova (whose preferred pronouns are they/them and not she/her, which will become important shortly) filed for copyright protection before the U.S. Copyright Office for their digital graphic novel, Zarya of the Dawn. While Kashtanova is noted as the primary author, and the writing is all their own, the imagery for Zarya was created by an AI program known as Midjourney, which Kashtanova utilized by creating the input strings and commands that generated the final imagery pivotal to the graphic novel.
This created a substantial problem, as AI-created content has been found by both the U.S. Copyright Office and the courts to be ineligible for copyright protection, owing to the lack of a recognizably human creator or human creativity and intent.
Just because someone enters a string of text into a search engine doesn’t grant them control over the results they receive (remember Whitehouse.gov vs Whitehouse.com?), and just because someone tells an AI program what they want to see in the final result doesn’t mean they have ongoing iterative control over the process of creating the content in question.
As an aside, “ongoing iterative control” was not used in the US Copyright Office’s decision, but is a descriptor I settled on to explain the difference between human-created art and AI-generated “art.” A human painting or drawing a picture will choose the size of canvas, the subject matter, the color palette, the correct brushes, pencils, pens, etc., and can apply and edit all of the above until the final result aligns with the artist’s intent. The same applies to writing, where a human author comes up with an idea, uses their vocabulary and available tools (keyboard, pen, pencil, clay tablets and chisels, etc.), and edits the output until it satisfactorily reflects what the author wants to say or convey. Likewise, a musician writing a song must choose tempo, rhythm, harmony, melody, and (if applicable) lyrics, put them all together, and create the final piece. All these are examples of ongoing iterative control.
AI cannot do this.
It can only mimic or ape ongoing iterative control because it can only follow directions, and then only within the limits of its library of information.
As I have said before, it is no more “creative” than a myna bird mimicking human speech. A myna bird can tell you it wants a cracker, or quote Shakespeare, but there’s no evidence that it understands what a cracker even is, let alone issue a reasoned argument for why the troubled Dane in Hamlet was more of a cautionary tale than a model of honorable conduct.
Midjourney’s own documentation seems to bear out this interpretation, because of the way in which its AI interprets information. In the “Grammar” section of Midjourney’s “Prompts” subcategory, the following paragraph proved pivotal in the Copyright Office’s determination:
The Midjourney Bot does not understand grammar, sentence structure, or words like humans. Word choice also matters. More specific synonyms work better in many circumstances. Instead of big, try gigantic, enormous, or immense. Remove words when possible. Fewer words mean each word has a more powerful influence. Use commas, brackets, and hyphens to help organize your thoughts, but know the Midjourney Bot will not reliably interpret them. The Midjourney Bot does not consider capitalization.
Kashtanova claimed that because they were the one to enter the keywords and strings into Midjourney, their input reduced, or elevated, Midjourney’s role in the creation of the graphic novel to the same input as a pencil, paintbrush, piano, or hammer and chisel in the hands of an artist—in short, because they were the one who entered the keywords into the program, the resulting images were a result of their own creativity.
However, the Copyright Office disagreed, saying, in part, “…[T]he fact that Midjourney’s specific output cannot be predicted by users [emphasis added] makes Midjourney different for copyright purposes than other tools used by artists” and that the use of AI did not constitute “original work of authorship fixed in any tangible medium of expression…”, which is a key and longstanding tenet of copyright law.
What this means, in a nutshell, is that while Kashtanova’s writing and the story of Zarya are protected by copyright but the accompanying graphics are not because they fail the most basic possible test for copyright to be applicable: namely, that of obvious and indisputable human authorship.
What the Artist Says
Kashtanova and their attorney, Van Lindberg, disagree with the US Copyright Office’s stance on the authorship of the images so essential to Zarya’s very existence as a graphic novel. While Lindberg says the ruling constituted “a success” on one level because the US Copyright Office had previously threatened to reject Zarya’s copyright in toto, he also asserts that the US Copyright Office had erred badly in failing to weigh “the input of the human” more heavily than “the output of the tool.” In his words, “A ‘modicum’ of [human] input is all it takes.” To bolster his argument, he points to Jackson Pollock’s infamous splatter painting and animals “photobombing” otherwise seriously intended photos in a humorous way as examples of works of art that were created with human input—but without any clear or obvious intent as to what the final outcome would be, despite the fact that the resulting works were eligible for copyright. From this perspective, AI is no different.
However, this flies in the face of both settled law and precedent, notoriously that of Dr. Stephen Thaler’s DABUS AI system and its trials and tribulations overcoming US Copyright Office dictums, as well as that of many of the Copyright Office’s opposite numbers internationally. Professors Jane C. Ginsburg and Luke Ali Budiarjo wrote an analysis of the legal arguments, pro and con, around which machine-authored content debates revolve. Specifically, they identified two key factors in authorship, as quoted by IP Watchdog in the article about Dr. Thaler linked above: “(1) a mental step, or the conception of a work; and (2) a physical step, or the execution of the work.” As AI is currently incapable of Step 1, Step 2 by definition is a nonstarter—both are reliant upon the human being “in the driver’s seat” at all times during the creative process, else there cannot be said to BE a creative process as we understand it.
Kashnatova took to Instagram on Wednesday, February 23rd, to air their joy at the upholding of the copyright for Zarya and their intentions for protecting the art they claim they played a decisive role in creating, saying:
I received the Copyright Office’s decision today about Zarya of the Dawn. The great news is that they affirmed my copyright, so Zarya of the Dawn will stay officially registered.
This is a great day for everyone that is creating using Midjourney and other tools. When you put your images into a book like Zarya, the arrangement is copyrightable.The story is copyrightable as well as long as it’s not purely AI produced. That covers a lot of uses for the people in the AI art community.
I was disappointed in one aspect of the decision. The Copyright Office didn’t agree to recognize my copyright of the individual images. I think that they didn’t understand some of the technology so it led to a wrong decision.
It is fundamental to understand that the output of a Generative AI model depends directly on the creative input of the artist and is not random.
My lawyers are looking at our options to further explain to the Copyright Office how individual images produced by Midjourney are direct expression of my creativity and therefore copyrightable.”
The Bottom Line
While Kashnatova has the right to appeal the decision, and to take it to the Copyright Office’s Review Board if appeal proves unsuccessful, it seems highly unlikely that they are likely to find the satisfaction they hope for in that milieu.
In short, in this situation, it seems safe to say that, once again, AI and the law have butted heads—and AI has come up short, as I expect it will continue to do for quite some time to come.
ABOUT JOHN RIZVI, ESQ.
John Rizvi is a Registered and Board Certified Patent Attorney, Adjunct Professor of Intellectual Property Law, best-selling author, and featured speaker on topics of interest to inventors and entrepreneurs (including TEDx).
His books include “Escaping the Gray” and “Think and Grow Rich for Inventors” and have won critical acclaim including an endorsement from Kevin Harrington, one of the original sharks on the hit TV show – Shark Tank, responsible for the successful launch of over 500 products resulting in more than $5 billion in sales worldwide. You can learn more about Professor Rizvi and his patent law practice at www.ThePatentProfessor.com
Follow John Rizvi on Social Media