http://ipkitten.blogspot.com/2025/02/a-single-piece-of-us-copyright-are-ai.html
Creation of the work
The step-by-step creation process of the work is explained in the “Copyright Whitepaper” prepared by Kent Keirsey (CEO of Invoke AI and the applicant for the registration) to demonstrate the USCO that the work “exhibits sufficient creative authorship to warrant registration”.
First, Mr Keirsey entered the following prompt into InvokeAI: “fractured glass, faces in the facets, surreal pattern of glazed brushstrokes, spaghetti noodle hair” and “blurry, out of focus, sketch. photo”. From these instructions the AI tool provided him with three image options.
Then, the applicant chose one of them to work further on his digital canvas.
Finally, Mr Keirsey, “inpainted” the AI-generated image, which is a process that allows users to make changes on the AI-generated image:
Inpainting in generative AI is the process of selectively modifying or regenerating parts of an image while maintaining consistency with the surrounding elements. Invoke’s canvas is uniquely designed to give artists a high degree of control, allowing them to precisely define where and how regions are regenerated.
In the end, Mr Keirsey obtained the image below.
USCO so far…
While addressing the copyrightability of AI outputs, the USCO has consistently stated that human authorship is a “bedrock requirement” and a principle deeply rooted in the US copyright law, and therefore, refused the registration of outputs that cannot demonstrate an adequate level of original human authorship. This has also been voiced in the US District Court for the District of Columbia in its ruling in Thaler v. Perlmutter, which accordingly, held that A Recent Entrance to Paradise, an output generated wholly by an AI tool, could not qualify for copyright protection.
The Office has not only refused the registration of wholly AI-generated outputs, but also AI outputs that it deemed to lack sufficient human contribution, for example, Théâtre D’opéra Spatial and the “artwork generated by artificial intelligence” in Zarya of the Dawn [IPKat here and here]. Perhaps the closer the Office ever got to protecting an AI image was when it registered Rose Enigma in 2023. However, even then, both in its Public Record System and its recently published Copyrightability Report, the USCO explicitly excludes any non-human expression and states that: “Registration [is] limited to unaltered human pictorial authorship that is clearly perceptible in the deposit and separable from the non-human expression that is excluded from the claim.” It appears that the USCO still protects the wholly human-created sketch, rather than the AI output that stems from the said original drawing and accompanying prompts.
| The hand-drawn sketch provided to the AI tool as an expressive input by Kristina Kashtanova, the author |
| Rose Enigma: The image generated by AI based on the author’s drawing |
Against this background, granting registration for A Single Piece of American Cheese is certainly a step forward in responding to AI art, however not without flaws.
Mr Keirsey claimed that he “selected, coordinated, and arranged numerous AI generated image fragments into a single, unified image” and therefore, A Single Piece of American Cheese “can and should be thought of as a kind of collage of separate elements that together form something new”. He further supported these by submitting a video that demonstrates each step of the creation process. The Office agreed.
| The initial AI-generated output and the final work Théâtre D’opéra Spatial, for which registration was sought by Jason Allen but refused by the USCO |
| The initial AI-generated output and the final work, A Single Piece of American Cheese, for which the USCO granted copyright protection |
It might, however, be the case that Mr Keirsey simply made a stronger and more persuasive argument, providing more compelling evidence to the Office than Jason Allen, creator of
Théâtre D’opéra Spatial. Mr Allen had also provided Midjourney with a significant amount of prompts, creative inputs, and made creative selection, coordination, and arrangements in relation to the AI output, which resulted in the creation of a “new visual content”. As the nature of their contributions to the final output was essentially the same, it is possible that Mr Allen’s failure to secure registration was due to a lack of strong supporting evidence.Concluding remarks: AI-generated artworks or compilations?
The USCO provided protection to A Single Piece of American Cheese – excluding individual AI-generated elements – on the basis that the user of Invoke AI creatively modified and changed the intermediary product. This language appears to draw from the copyrightability of compilations [see Feist].
It is also curious that the Office issued this decision to register Mr Keirsey’s work just one day after publishing its report on the copyrightability of AI outputs. This timing seems to demonstrate the correct application and interpretation of its recommendations, which inter alia suggest that prompts alone are not always sufficient to demonstrate originality. Therefore, the “creative selection, coordination, or arrangement” of AI outputs, any “creative modifications” made to initial or intermediary outputs, as well as “expressive inputs” should also be considered in determining originality. It might have been a more useful approach if the USCO adopted a broader interpretation of ‘prompts’ as all the actions it refers to, in essence, instruct AI to act in a certain way or to perform a certain task, be it in the form of textual instructions, human-drawn sketches, or modifications on the initial AI output. A broader interpretation of ‘prompting’ could have rendered the Office’s decisions regarding the copyrightability of works in which AI tools are used more predictable. For now, this inconsistency and ambiguity – or even arbitrariness, as seen in the opposing decisions on Théâtre D’opéra Spatial and A Single Piece of American Cheese – will likely persist.
Finally, if the USCO did actually demonstrate the correct application of its recommendations with its decision on A Single Piece of American Cheese, then it seems that the Office provided a 52-page report merely to say that AI outputs are to be protected as compilations, where applicable. The value of generative AI technology, however, lies not only in its ability to produce compilations, but also in its capacity to generate artistic, musical, or literary works, based on the original instructions users provide at various stages of the process.
Content reproduced from The IPKat as permitted under the Creative Commons Licence (UK).