On January 29, 2025, the U.S. Copyright Office issued its long-awaited report on authorship and AI. This Report followed its earlier AI Guidance on Works Containing Material Generated by Artificial Intelligence published on March 16, 2023.
Has the U.S. Copyright Office’s position on AI-generated works changed since March 2023?
On the surface, the Report might strike people as the same position. But, as soon as it issued, I had a different reaction, which was later borne out by the Office’s own registrations of works originally generated with AI. In some significant ways, the Office’s view has evolved. But, in other significant ways, it remains the same. As AI generators continue to advance with new functionalities offering creators even more options, I suspect the Office’s receptivity to registering partly AI-generated works will increase, as evinced by the Office’s own repeated qualifier that its view is based on the “current technology.”
A. Ways in which the U.S. Copyright Office’s stance has evolved
1. The U.S. Copyright Office has de-emphasized the “traditional elements of authorship”
One noticeable change in the 2025 Report is its de-emphasis of the so-called “traditional elements of authorship.” The Report only mentions the term once–and only in a quotation from Register of Copyrights Kaminstein, but not as a part of its own analysis. By contrast, the 2023 AI guidance invoked the term 4 times–and did so as a key test or doctrine in the Office’s analysis of human authorship. In its rejection of Jason Allen’s registration of an AI-generated image, the Office likewise invoked the “traditional elements of authorship” 3 times.
Of course, the Office can invoke this term in the future. But it is striking that in its 2025 Report to Congress, it only used the term once–and only in a quotation, not in an application. If the Office intended to make this term a key part of its analysis, going forward, one would expect the Report to have explained and applied the term more than once.
In my scholarship and comment to the Copyright Office (see Related Stories), I criticized the use of “traditional elements of authorship” as wholly unsupported by the Progress Clause, Supreme Court precedent, or the text of the Copyright Act. No federal court has ever used that term to deny a person a copyright.
2. The U.S. Copyright Office has identified several ways in which human authors can qualify their AI-generated works for copyright
The other significant development is that 2025 Report delineates at least 3 different ways in which human authors can qualify their works for copyright, with the underlying AI-generated elements excluded. After receiving approximately 5,000 comments on the issue of authorship, the Report does provide greater explanation on the ways in which human creators can create works using AI and qualify the works for copyright under the Copyright Office’s view.
i. Selecting, coordinating, arranging, or editing of AI generated elements (post-initial generation)
Perhaps the most important page of the entire 2025 Report is page 26. It recognizes how human creators can qualify as the author of works, initially generated through AI, based on the human’s further selection, coordination, or arrangement of elements.
The Copyright Office even recognizes that a human’s use of inpainting (on Midjourney called “vary region and remix prompting” and inpainting on Stable Diffusion) may be the basis for such authorship, provided the selection, coordination, or arrangement satisfies originality. After the 2025 Report was published, a number of copyright registrations of initially AI-generated works were allowed based on selection, coordination, or arrangement of elements (see below Related Stories). We can expect more works will so qualify in future registrations.
If the authorship is based on selection, coordination, or arrangement, the AI generated elements must still be excluded. As the Office Report explained, “For example, a film that includes AI-generated special effects or background artwork is copyrightable, even if the AI effects and artwork separately are not.“
The Copyright Office’s position moves closer to the one I advanced in scholarship and my comment to the Office (see Related Stories). In these writings, I stressed the importance of recognizing authorship based on selecting and arranging AI-generated elements, provided originality is satisfied. The Office still maintains that “prompts alone” cannot render an output that satisfies human authorship, even if a person engages in iterative prompting of many images. But the Office’s Report now recognizes that inpainting and other post-initial generation functions of a generator used to modify an initial AI output may satisfy it. This recognition makes less salient the Office’s own distinction of “prompts alone”—which may, in fact, be exactly what the Office intended. Although the 2023 AI guidance mentioned selection and arrangement as a potential basis for authorship, it said virtually nothing about how a human creator can do so using AI. By contrast, the 2025 Report–which was drafted after the Office studied information submitted in thousands of comments–goes into much greater detail with specific examples of uses of AI, such as inpainting.
The Office’s recognition of authorship based on selection and arrangement is apparent in its registration of a visual work aptly titled A Single Piece of American Cheese. Invoke, the corporate creator of the work, recorded the creative process on video, demonstrating the many elements that a human creator, Kent Keirsey, edited through inpainting after the initial AI-generation of the work. This example also suggests a video recording might provide good evidence of the level of human selection and arrangement to bolster one’s claim of authorship.
ii. Starting first with human expressive input, such as a human drawn sketch
The Report also recognizes that a human author may create a work of expression that serves as the input for an AI generator. If the human expression is retained in the ultimate work, the human author can claim copyright with respect to their elements, but not the AI-generated elements. The Report explained: “Their own creative expression will be protected by copyright, with a scope analogous
to that in a derivative work. Just as derivative work protection is limited to the material added
by the later author, copyright in this type of AI-generated output would cover the perceptible
human expression. It may also cover the selection, coordination, and arrangement of the
human-authored and AI-generated material, even though it would not extend to the AI-generated elements standing alone.”
For example, the Report discusses the hand-drawn work by Kristina Kashtanova:
iii. Assistive uses of AI tools
Then, the Report distinguished use of AI “as a a tool to assist in the creation of works” versus “using AI as a stand-in for human creativity.”
“While assistive uses that enhance human expression do not limit copyright protection, uses where an AI system makes expressive choices require further analysis. This distinction depends on how the system is being used, not on its inherent characteristics.”
“Commenters also identified situations where creators have begun to experiment with
using AI as a brainstorming tool. The Recording Academy, for instance, stated that “[m]any
Academy members already use generative AI as a tool to assist them in creating new music,”
including through song ideation. Another stakeholder noted that AI can be used to structure
or create a preliminary outline for literary works. In these cases, the user appears to be
prompting a generative AI system and referencing, but not incorporating, the output in the
development of her own work of authorship. Using AI in this way should not affect the
copyrightability of the resulting human-authored work.”
The Report’s recognition of the use of AI as an assistive tool might not eliminate the Motion Picture Association’s concern that AI can be used in a variety of ways in film production. Indeed, the Report’s exclusion of AI-generated special effects from copyrightability is likely to be a continuing source of concern: “a film that includes AI-generated special effects or background artwork is copyrightable, even if the AI effects and artwork separately are not.” If the special effects footage is not copyrightable, anyone is free to copy it.
B. Ways in which the U.S. Copyright Office’s stance is same as 2023
The main way in the Copyright Office’s position is the same as it was in 2023 is its drawing a bright-line against “prompts alone” generating any output that can qualify as a work of human authorship at least based on the “current generally available technology.”
In the Office’s view, the problem is that AI generators under current technology do not enable humans “to exert so much control over how their expression is reflected in an output that the system’s contribution would become rote or mechanical.”
But, as mentioned above, inpainting does involve prompts, albeit used in a way that enables people to target the specific locations and elements within an AI-generated image to change. So, as long as AI generators include inpainting and other editing options, as Midjourney does, the generators already provide people with ways in which they can qualify for copyright based on their selection, coordination, and arrangement of elements, in post-initial generation editing.
* * *
The Report is well worth a read. I’ve summarized only the parts of it related to the Copyright Office’s position on how human creators can qualify for copyright in creating works using AI. The parts highlighted below:
Related Stories