Vulnerability of Artistic Style in the Age of Generative AI

Maggie Lee

Animation studios have distinct, recognizable styles, such as Hayao Miyazaki’s Studio Ghibli’s combination of simple and clear lines with muted color tones,[1] or Disney’s vibrant and expressive CGI art style in films like Tangled and Frozen. Even if one has not seen a particular Ghibli film, nor was given any indicia of the movie’s origin, individuals familiar with the Miyazaki’s distinctive style would instantly discern its origins upon encountering a character from said film. Ghibli characters are characterized by their simple faces: expressive yet unadorned eyes, and noses and mouths delineated with few subtle lines.[2]

With the development of generative AI, fans of these studios have been able to easily recreate their own images or create new images in the style of these animation studios.[3] When such an AI-generated image is shared publicly, the public generally associates the respective styles with these internationally recognized studios, irrespective of whether the image originates with the studio. However, there are also cases where an artist’s style gains immense popularity without clear attribution. For example, Greg Rutkowski, a digital artist known for his dreamy fantasy landscapes created in classical painting styles, has created illustrations for popular games such as Dungeons & Dragons and Neverwinter.[4] Rutkowski’s distinctive style became one of the most commonly used prompts to crate dreamy, fanciful images in the AI generator Stable Diffusion.[5] While the initial user recognizes Rutkowski’s style when using his name as a prompt to generate an image, social media users encountering Rutkowski-style images may not recognize his work or even know who Rutkowski is.

Although the prompt “Greg Rutkwoski” was later removed in the newer version of Stable Diffusion,[6] artists like Rutkowski are still worried that “[s]omeone can create something in a matter of five seconds using my name, or any other artist’s name, as a prompt, as a guideline for style.”[7] This leads to significant repercussions for small artists, as their demand for commissions diminishes when AI can effortlessly produce art mimicking their style either at no cost or through a subscription-based service. In fact, since the emergence of generative AI, Rutkowski said he has received far fewer requests from authors who need covers for their fantasy novels.[8]

Is there a way for artists to protect their distinctive styles from being copied by AI, or others more generally? Copyright law suggests otherwise, distinguishing between an idea, which is freely available to all, and expression, which is protectable. An artist’s style is considered to be an idea; thus, imitation of a style is not infringing.[9] As a district court once put it, “Picasso may be entitled to a copyright on his portrait of three women painted in his Cubist motif. Any artist, however, may paint a picture of any subject in the Cubist motif, including a portrait of three women, and not violate Picasso’s copyright so long as the second artist does not substantially copy Picasso’s specific expression of his idea.”

Although an artist could make a copyright infringement claim based on copying that occurs (if any) during the AI system’s training data, the resulting AI work in the style of an artist would not be infringing unless the work copied specific, protectable expressive elements from the artist’s underlying work. For example, in a (non-AI) case, the movie poster for Moscow on the Hudson clearly copied Saul Steinberg’s hallmark “sketchy, whimsical style.”[10] The court took the “striking stylistic relationship” between the two works into account in finding infringement “since style is one ingredient of ‘expression.’”[11] However, had there only been “striking stylistic relationship” without copying of other protectable expressive elements, the court would likely have not found infringement.

While an artist cannot receive copyright protection for this artistic style, could he at least gain attribution rights? In the U.S., an artist is entitled to be identified by the name of his work and prevent his name from being attributed to a work he did not create,[12] but only in relation to a specific work of visual art. Unfortunately, an artist does not have attribution rights to his style.

Currently, copyright law does not offer protection for an artist’s distinctive style, considering it an idea rather than an expression eligible for copyright. Artists also lack attribution rights for their styles, unable to prevent their style from being misattributed or copied. Although large studios could negotiate with or threaten generative AI over their rights,[13] small artists are at a significant disadvantage when it comes to bargaining power. Regrettably, the law in its current state does not provide protection for the distinctive styles of artists, leaving them vulnerable to exploitation in the age of AI.

 

[1] Steve Chang, The Art Style of Studio Ghibli (How To), Sapling Corp (Nov. 30, 2023), https://www.saplingcorp.com/journals/30/studio-ghibli-art-style [https://perma.cc/T54U-TBRV] [https://web.archive.org/web/20240321185104/https://www.saplingcorp.com/journals/30/studio-ghibli-art-style].

[2] Id.

[3] For example, the program AnimeGenius offers users the ability to “[t]urn [their] imagination or real photos into detail[ed] Studio Ghibli AI art,” and “[e]asily create Ghibli style arts, characters, landscapes or any image,” by using text prompts or inputting the user’s actual images. Studio Ghibli AI Generator, AnimeGenius, https://animegenius.live3d.io/features/studio-ghibli-ai-generator [https://perma.cc/R759-4SKL] [https://web.archive.org/web/20240321185500/https://animegenius.live3d.io/features/studio-ghibli-ai-generator] (last visited Mar. 21, 2024); Similarly, the program “getimg.ai” offers users the ability to “[c]reate Disney characters, images and posters . . . [by] choos[ing] between two Disney Pixar-based AI models to create Disney characters from text or apply[ing] Disney filters to any image or photo.” Disney Pixar AI Generator Online, Getimg.ai, https://getimg.ai/use-cases/disney-pixar-ai-generator [https://perma.cc/5EUA-JAND] [https://web.archive.org/web/20240321185921/https://getimg.ai/use-cases/disney-pixar-ai-generator] (last visited Mar. 21, 2024).

[4] Ian Dean, “It’s Terrifying”—Greg Rutkowski is the Most Prompted Artist on Stable Diffusion, Creative Bloq (May 31, 2023), https://www.creativebloq.com/features/greg-rutkowski-ai-art-prompts [https://perma.cc/5VVJ-8LT4] [https://web.archive.org/web/20240321190620/https://www.creativebloq.com/features/greg-rutkowski-ai-art-prompts].

[5] Rutkowski’s name was used as a prompt around 93,000 times, compared to Picasso, whose name appeared as a prompt around 2,000 times. Melissa Heikkilä, This Artist is Dominating AI-Generated Art. And He’s Not Happy About It, MIT Tech. Rev. (Sept. 16, 2022), https://www.technologyreview.com/2022/09/16/1059598/this-artist-is-dominating-ai-generated-art-and-hes-not-happy-about-it/ [https://perma.cc/6C98-PFCD] [https://web.archive.org/web/20240321190318/https://www.technologyreview.com/2022/09/16/1059598/this-artist-is-dominating-ai-generated-art-and-hes-not-happy-about-it/].

[6] Kasmir Hill, This Tool Could Protect Artists from A.I.-Generated Art That Steals Their Style, N.Y. Times (Feb. 13, 2023), https://www.nytimes.com/2023/02/13/technology/ai-art-generator-lensa-stable-diffusion.html [https://perma.cc/D5K7-NBV3] [https://web.archive.org/web/20240321191038/https://www.nytimes.com/2023/02/13/technology/ai-art-generator-lensa-stable-diffusion.html].

[7] Dean, supra note 4.

[8] Hill, supra note 6.

[9] Dave Grossman Designs, Inc. v. Bortin, 347 F. Supp. 1150, 1156–57 (N.D. Ill. 1972).

[10] Steinberg v. Columbia Pictures Indus., Inc., 663 F. Supp. 706, 712 (S.D.N.Y. 1987).

[11] Id.

[12] 17 U.S.C. § 106A.

[13] After a social media trend where owners used Microsoft’s AI-generator to create images of their pets in the style of Disney, Microsoft temporarily blocked the term “Disney” from being entered as a prompt after Disney raised its concerns. Cristina Criddle, Microsoft Tweaks AI Image Generator Over Disney Dogs Poster Trend, Fin. Times (Nov. 17, 2023), https://www.ft.com/content/64958031-6d7b-4c2c-aeb3-d4df0b04ae32 [https://perma.cc/M7V4-PMCQ] [https://web.archive.org/web/20240321192506/https://www.ft.com/content/64958031-6d7b-4c2c-aeb3-d4df0b04ae32].