The US Copyright Office has reconsidered copyright protection it granted to Kristina Kashtanova for her comic book Zarya of the Dawn last fall, Portal reports. It contained images created by inputting text prompts to Midjourney, an artificial intelligence image generator.
According to this letter (PDF), sent by Robert Kasunic to her attorney, the US Copyright Office has ruled that Kashtanova is “the author of the text of the work and of the selection, coordination, and arrangement of the written and visual elements.”
However, the images themselves are “not the product of human authorship” and the registration originally granted to them has been revoked. To justify the decision, the Copyright Office cites previous instances where people failed to copyright words or songs that credited “non-human spiritual beings” or the Holy Spirit — as well as the infamous selfie-taking incident from a monkey.
The Copyright Office said it only became aware of the Midjourney images based on Kashtanova’s social media posts after the registration was granted, and then sought further information. Both Midjourney and Kashtanova are credited on the cover of the book, but according to the letter, this is the only place Midjourney appears on the 18 pages of material filed with the Copyright Office, and “The fact that the word ‘Midjourney’ appears on which appears The cover page of a work does not constitute a notification to the Office that an AI tool has created part or all of the work.”
At the end of the letter, Kasunic writes that the original certificate was issued on the basis of “incorrect and incomplete information” and will therefore be cancelled.
The artist posted about the decision on Instagram, calling it a “great day” for people who use Midjourney and similar tools. “When you put your images into a book like Zarya, the arrangement is copyrighted. The story is copyrighted as long as it’s not purely AI-produced,” she wrote, while expressing disappointment at the Copyright Office’s decision not to give her copyright to the individual images.
The Copyright Office’s decision takes into account how Midjourney generates image output by tokenizing word prompts, which it compares to training data. While other AI programs might work differently, the letter states, “The fact that Midjourney’s specific output cannot be predicted by users distinguishes Midjourney from other tools used by artists for copyright reasons.”
The Office also rejects the claim that their edits to some of the images give them copyright consideration, ruling that the changes are either “too minor and imperceptible to provide the necessary creativity for copyright protection” or that they are not based on their contributions whose could determine the transmitted information.
Kashtanova’s attorney Lindberg disagrees, saying: “There are a number of flaws in the Bureau’s arguments, some legal and others factual. However, they all seem to stem from a fundamental factual misunderstanding of the role played by chance in midjourney’s image generation.”
Among the errors he lists is the interpretation of whether or not Kashtanova contributed a “minimum” of input. Did their rapid engineering qualify as a mere suggestion, or, as he argues, did their instructions cause Midjourney to “do exactly what it’s programmed to do, and draw from an artist-chosen spot in his vast probability table, to induce the generation of to.” drive a picture”?
Lindberg claims: “AI-assisted art must be treated like photography. It’s only a matter of time.”
Kashtanova concluded her contribution by saying, “My attorneys are evaluating our options to further explain to the Copyright Office how individual images produced by Midjourney are a direct expression of my creativity and are therefore protected by copyright.”