IP Frontiers: Critiquing the Copyright Office’s Rejection of AI-Generated Copyrights

By: Arjay Parhar

When the system is built around a dated statute, how does the U.S. Copyright Office (“Copyright Office”) treat unanticipated technologies? AI-generated images are notably absent from 17 U.S.C. § 102, but that has not stopped AI artists from seeking copyright protection.

In two examples, the Copyright Office rejected image copyrights for Théâtre D’opéra Spatial, an award-winning Space Opera image, and Zarya of the Dawn, a dystopian New York City graphic novel. Théâtre D’opéra Spatial’s author Jason Allen used at least 624 text prompts and input revisions in Midjourney, the AI-image generation tool, to create Théâtre D’opéra Spatial, further manipulating the image with Photoshop and enhancing the resolution using Gigapixel AI. The Copyright Office offered to grant copyright protection for Théâtre D’opéra Spatial if Allen disclaimed the image’s AI-generated elements, to which Allen refused. In its decision to revoke copyright image protection of Zarya of the Dawn, the Copyright Office concluded,

  • that the images generated by Midjourney contained within the Work are not original works of authorship protected by copyright. . . explaining that “the Office will not register works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author”. Though [the author] claims to have “guided” the structure and content of each image, the process described in the [the author’s] Letter makes clear that it was Midjourney—not [the author]—that originated the “traditional elements of authorship” in the images.

Though the Copyright Office refused to register AI-generated images here, courts may still overturn the Copyright Office’s rejections in favor of consistent policy justification for new technologies. For example, the Court used policy justifications to extend copyright protection to photographs in Burrow-Giles v. Sarony. Though the first photograph was created somewhere between 1822 and 1833, this case granted the first photograph copyright protection in 1882. At the time, cameras were technically revolutionary, but artistically chided – just like AI tools. The Court even noted the copyright criticisms of photography,

  • it is said that . . . while a photograph is the mere mechanical reproduction of the physical features or outlines of some object . . . and involves no originality of thought or any novelty in the intellectual operation connected with its visible reproduction in shape of a picture . . . the remainder of the process is merely mechanical, with no place for novelty, invention, or originality. 111 U.S. 53, 58–59 (1884).

At issue was the copyrightability of a pondering playwright’s photograph. The Court refused to prohibit protection of this image and lauded the photographer’s creative process, citing the photographer’s choice of pose, costume, background, selection of features taken by the camera, the arrangement of light, and more. Even though the relevant statute omitted photography, just as the Copyright Act of 1976 omits AI-generated images, the Court extended copyright protection to photographs as a class – ushering in a future of photography copyrights.

The Copyright Office is repeating history by failing to recognize the artistic value and technical process involved in AI-image generation tools. The decisions for Théâtre D’opéra Spatial and Zarya of the Dawn fail to understand an artistic process that can creatively use AI image generation tools. Contrary to the Copyright Office’s reasoning, an artist can progressively alter an image to reduce the variable noise of an AI image generator to narrow in on their creative vision, which contradicts the assumptions that an AI artist lacks forethought and creative control over the AI tool’s output.

The Copyright Office should be allowing artists to demonstrate intentional creative choices, such as those outlined in Burrow-Giles v. Sarony, like a subject’s pose, costume, background, selection of features taken by the fictional camera, the arrangement of light, and more. However, in the Copyright Office’s decision of Théâtre D’opéra Spatial and Zarya of the Dawn, the authors’ opportunities to demonstrate these elements were cut short. The Copyright Office rejected the 624 text prompts and input revisions in Théâtre D’opéra Spatial and the year of work in Zarya of the Dawn because “Midjourney does not interpret prompts as specific instructions to create a particular expressive result….because…Midjourney does not understand grammar, sentence structure, or words like humans. . . .[B]ecause Midjourney does not treat text prompts as direct instructions, users may need to attempt hundreds of iterations before landing upon an image they find satisfactory.” U.S. Copyright Office, Second Request for Reconsideration for Refusal to Register Théâtre D’opéra Spatial, https://www.copyright.gov/rulings-filings/review-board/docs/Theatre-Dopera-Spatial.pdf quoting U.S. Copyright Office, Cancellation Decision re: Zarya of the Dawn, https://www.copyright.gov/docs/zarya-of-the-dawn.pdf (quoting MIDJOURNEY, Prompts, https://docs. midjourney.com/docs/prompts). Contrary to the Copyright Office’s assertions that a 1:1 literary comprehension to artistic output contrasts AI image generation from more traditional computer tools such as Photoshop, the lack of a 1:1 level of control is actually a limitation that offers support for the traditional elements of authorship when authors have to make numerous choices to circumvent this limitation to be true to their creative vision. To circumvent the AI tool’s unpredictability, authors often style images using prompts such as “photorealistic”, “impressionist”, “abstract”, and “black and white noir”. Each factor in Burrow-Giles are factors an author can adapt in AI image generation: a subject’s costume, their pose, the background, the position of light and how it hits a subject. An artist revising their textual input hundreds of times over evidences artists’ using AI tools to narrow in on their creative vision – as opposed to superficially inputting a couple phrases and haphazardly choosing an immediate result. Though the Copyright Office disclaims AI tools because “users may need to attempt hundreds of iterations before landing upon an image they find satisfactory”, numerous choices made through each iteration is specifically the scenario that deserves copyright protection.

The Copyright Office also wrongly assumes that AI artists are not the masterminds behind their creation. The Copyright Office differentiates AI artists from other artists by contrasting the relationship between the artist and the AI image. While the Copyright Office assumes that a photographer makes several decisions that ultimately result in the taken photograph and the photographer uses reality as an inspirational springboard to create their artwork, the Copyright Office believes that “[b]ecause of the significant distance between what a user may direct Midjourney to create and the visual material Midjourney actually produces, Midjourney users lack sufficient control over generated images to be treated as the ‘master mind’ behind them”. The Copyright Office believes that an AI-generated image is ultimately randomly generated (at least those created by Midjourney); therefore, no matter how precise an author uses these tools, no causal link exists between an author’s text input and the image. Thus, there can be no human originality or creativity. However, this limited view of using an AI-image generation tool only makes sense when the tool is superficially used. When the tool uses hundreds or even thousands of revisions and inputs, the author is exerting more exact control over the image by progressively altering the image to a specific style with envisioned details.

Further, the Copyright Office blanketly attributes human originality to a fictionalized exacting control in artforms, but it should instead embrace a more realistic conception of control in the creative process. Photographs, paintings, poetry, literature, inventions, all these artforms may begin with a mental conception, but that initial conception often morphs with every stroke of the brush, every poured emotion, every failed experiment. A new image, emotion, message, or product often readapts and converges into a unique presentation different from the initial conception. Attributing a fictional exacting control to only AI-generated images would unfairly subject AI artists to a standard of their own. As the author of Zarya of the Dawn explains, “[i]t 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. . . . [I]ndividual images produced by Midjourney are [a] direct expression of my creativity. . . .” Likewise, the Théâtre D’opéra Spatial’s 624 text prompts and input revisions should evidence the author’s creative vision and artistic control, each prompt a stroke towards the artist’s vision. Instead, the Copyright Office incorrectly attributed the creative process as evidencing the “sweat of the brow” doctrine, which refuses copyright protection lacking a creative conception even if an artist works hard on a work. However, that evidentiary classification is mistaken because that evidence is specifically important for establishing the creative vision, as several iterations and text prompts are specific requirements unique to creatively and effectively using AI tools.

Though human originality and creativity will always exist in several planes, the tools for human creativity have been steadily eroding many financial and technical barriers barring an idea’s manifestation. As the famous actor, musician, and producer, Donald Glover, explains in the context of music and implying with the existence of AI tools, creativity,

  • used to be about access. . .[I]f you had access, you would be able to make stuff and then if you happened to be good it was different. Now everyone has access and it’s way easier to make all these things. It’s way easier to make music. We’re always racing towards very good ideas. That’s all that matters at the end of the day.” Donald Glover (Childish Gambino) Breaks Down His Most Iconic Characters | GQ, YouTube (Apr. 4, 2023).

As Donald Glover alludes to, AI tools are not substituting human originality, they are just eroding the barriers necessary to bring forth an idea. Just as digital audio workstations greatly reduced music production requirements from an ensemble of instruments and a recording studio to one person with a computer, so too are AI tools simply eroding many technical skills necessary to bring forth an idea. Just because an author is not using traditional more labor-intensive methods of bringing forth an idea does not mean that the idea itself is lacking human originality. AI tools are bringing people closer than ever to their human ideas. The Copyright Office should and will hopefully follow suit in recognizing AI tools as barrier-eroding tools, and not the originating authors themselves.

In the meantime, authors using AI tools should carefully document every part of their creative process to weave a story that demonstrates a creativity and original authorship that deserves copyright protection. One day, one of these artists may create an image with an extensive and progressively involved process that elevates the status of other similarly created AI-generated images.