As first reported by CNet, the AI-company Invoke, successfully registered a single AI-generated image for a work aptly titled “A Single Piece of American Cheese.” (Invoke is the author as an employer under the work-made-for-hire doctrine.) But Invoke did so only based on the “selection, coordination, and arrangement” of elements otherwise generated by an AI generator.
This work is the first AI-generated image the Copyright Office has registered.
This approach to AI-generated works is what I’ve been advancing for nearly two years in my writings: WaPo Op-Ed, Comment to USCO, law review essay, and law review article. They should be recognized as works of authorship as long as the human author makes a minimal level of creative selection, coordination, or arrangement of the underlying elements, which needn’t be copyrightable in themselves.
I may still disagree with the Copyright Office’s apparent position that prompting itself cannot effectuate any selection, coordination, and arrangement (the Copyright Office’s latest Report seems to recognize such decisions effectuated by other means, including ones offered by generators, as long as not in the form of a mere prompt. I will follow up in a subsequent post.).
Kent Keirsey, CEO of Invoke, said to CNet they used “inpainting” to generate 35 elements within this image. I argued in my law review article and Comment to the USCO that inpainting should count as a way human authors can select, arrange, and coordinate elements with AI generators (and gave examples of doing so in both writings). The Copyright Office did not use the word “inpainting” in its Report, but it did refer to the terms that some AI generators use to describe the same functionality:
USCO REPORT: “Many popular AI platforms offer tools that encourage users to select, edit, and adapt AIgenerated content in an iterative fashion. Midjourney, for instance, offers what it calls ‘Vary Region and Remix Prompting,’ which allow users to select and regenerate regions of an image with a modified prompt….Other generative AI systems also offer tools that similarly allow users to exert control over the selection, arrangement, and content of the final output.
“Unlike prompts alone, these tools can enable the user to control the selection and
placement of individual creative elements. Whether such modifications rise to the minimum standard of originality required under Feist will depend on a case-by-case determination.In those cases where they do, the output should be copyrightable.”
This is a welcome step in the right direction for the U.S. Copyright Office.
But the Office’s dichotomy between “prompts alone” versus inpainting (aka “vary region” on Midjourney or “generative fill” on Photoshop) is a false one. Inpainting involves prompt engineering as well, albeit, yes, the creator can target specific elements that have been already generated to revise, alter, or completely change to something else by using prompts. Granted, inpainting does offer more direct control over the elements within an image–that’s in part why creators love the functionality and ability to edit every single element within an image. But, for the purposes of a selection, coordination, or arrangement of elements, there shouldn’t be an artificial distinction between prompts versus prompts via inpainting. The true test should be whether the putative author has made a minimally creative selection, coordination, or arrangement through whatever means.
The Office doesn’t get the last word, let’s not forget. The courts have yet to rule on any AI-prompt engineered work. Jason Allen has a lawsuit challenging the Office’s prior approach. Stay tuned.
Keirsey’s hands-on role in the creation process was key to his application for copyright. Invoke first applied for a copyright for “A Single Piece of American Cheese” in August 2024 and was denied because the U.S. Copyright Office said it “lack[ed] the human authorship necessary to support a copyright claim,” according to correspondence between Invoke’s legal team and the office reviewed by CNET. Invoke followed up with more evidence, including a timelapse video of the image’s creation and an explanation of how Keirsey was involved in the creation process. The Copyright Office gave its stamp of approval on Jan. 30.
More on the creative process Keirsey used:
Related Stories
One response to “AI-generated image received copyright registration based on “selection, coordination, and arrangement.” Yes, in the United States. How?”
There’s a level of absurdity in how the Copyright Office approached this.
It’s fully in line with their recent report, and their existing stance, that copyright registration can only be granted to the human elements.
But when you break it down in this particular case – and limiting it to a single element for argument’s sake – what the copyright office is saying is this:
The placement of the cheese on top of the head is human artistry, and deserving of copyright registration.
However, the cheese itself is not carved into rock, not painted onto canvas, not sketched onto paper, not digitally inked in pixels, not a masked out photograph of a half-melted slice of cheese – no, it is A.I.-generated, and therefore the cheese itself is not deserving of copyright registration.
Quite literally *almost any other process* by which the depiction of a slice of cheese would have come into being would be eligible, *except* A.I.-generation no matter the nature of that A.I.-generation or the level of control a human has over that generation.
( The major other exception would be if they took an existing depiction of a half-melted slice of cheese – in which case the registration of copyright would presumably rest with *its* copyright holder, if ever sought. )
There will come a time where this is untenable, and the Copyright Office’s recently released discussion report on the economic aspects of generative A.I. gives keen insight into why that is the case.