In my previous posts on generative AI, I discussed fair use and AI training data, copyright over AI outputs, and a recent U.S. Copyright Office decision on registration for a work produced by generative AI. In the next posts in our series, I will look at claims (exemplified in a recent case against Stable Diffusion and Midjourney) that generative AI (GAI) tools should be held liable under copyright law for their ability to output works in the style of specific, named artists.
The issue in a nutshell: Artists have raised legal claims against particular users who prompt a GAI and generate an output that copies from their original expressions. However, style is not generally protected by copyright, and that’s a good thing; if one artist were given a monopoly over anime, grunge music, or other styles, that would frustrate copyright’s core purpose of supporting creativity. What’s more, GAI tools provide myriad legitimate uses, and creators of those tools generally don’t directly control the actions of their users or financially benefit from uses that may infringe on pre-existing works. As such, they generally should not be held liable when actions of their user cross over the line into infringement.
In this blog post, I address the first issue — whether copyright should protect artistic style. In my next post, I will look at whether GAI tools should be liable if users create works that may infringe on works by other artists.
What’s the issue?
Before looking at the legal issues involved, it’s important to start with the concerns animating artists’ claims here. One part of these concerns seems to be related to the more general worry about AI disrupting the market for human work. Several months ago, an MIT Technology Review article wrote that Greg Rutkowski’s name was used at least 93,000 times to produce images using Stable Diffusion. Greg Rutkowski is an artist from Poland who is known for producing epic fantasy scenes, having created illustrations for things like Dungeons and Dragons and Magic the Gathering. If Stable Diffusion can create Rutkowski-like works by being instructed to “produce an epic fantasy scene with a fire breathing dragon fighting a noble warrior in the style of Greg Rutkowski,” why would anyone hire Rutkowski for his art any more? This AI generated content could be produced much more quickly and at a lower cost than commissioning Rutkowski for a custom piece. And, of course, this could apply to any artist and any style.
Even if the ability to generate similar works with AI doesn’t affect artists’ income or the economic value of the original artists’ works, artists still may feel that this devalues their work and their efforts in a more fundamental way. Artists spend time and effort establishing, developing, and refining their styles. If, all of a sudden, anyone could produce works in that style simply by adding some keywords into an AI text prompt box, it may appear to devalue the human creative effort and the integrity of their works.
Does copyright protect artistic style?
Concerns about devaluing the works created by human artists are important considerations — but are there problems that copyright law can and should fix? That is, can and should artists be able to claim copyright protection over the general style and feeling of their works, as opposed to the specific design elements expressed in their specific works?
One fundamental principle of copyright law is that copyright does not protect ideas, but instead protects the specific expressions of ideas that artists create through their art. As the Supreme Court wrote in Google v. Oracle: “copyright protection cannot be extended to ‘any idea, procedure, process, system, method of operation, concept, principle, or discovery ….’ [17 U.S.C.] § 102(b). These limitations … have often led courts to say, in shorthand form, that, unlike patents, which protect novel and useful ideas, copyrights protect ‘expression’ but not the ‘ideas’ that lie behind it.”
Looking at this in the context of style, style alone is not usually considered the subject matter of copyright. Consider, for example, music genres. Copyright does not extend to something like genre. So, copyright doesn’t protect the style of music “grunge,” and in fact, many bands who work in the same genre have similar sounds. Genres are, in fact, defined by stylistic similarities. So, for instance, early 1990s grunge bands had sounds that were, in many ways, alike: sludgy, heavy electric guitars, intense vocals, elements of both punk rock and heavy metal music. At the same time, however, copyright does protect specific expressions of that style: Nirvana’s Smells Like Teen Spirit, Pearl Jam’s Even Flow, Silverchair’s Tomorrow. Similarly, copyright doesn’t protect animation style. The American-made animated television show Avatar: The Last Airbender shares a similar animation and storytelling style with many Japanese anime shows. Indeed, people often think Avatar: The Last Airbender is, itself, a Japanese show. However, it is not a copy of any other show, even if it appears similar to anime.
Copyright doesn’t protect things like style and genre, because doing so would limit the ways that others can create other works, thereby chilling their ability to express themselves through their works. If the law allows artists to extend the scope of their copyright monopolies beyond their works to include elements outside their specific expression, it would compromise the ability of future artists to be inspired by and build from the art that they experience and love. A band like Silverchair, which sounds (at times) a bit like a mix of Nirvana and Pearl Jam, might have run into copyright problems if those bands had a copyright over their grunge sounds. Or a show like Avatar might never have been able to exist if the anime style itself was the subject of copyright.
At the same time, however, the line between style and expression can be fuzzy. Artists often work to develop idiosyncratic styles that differentiate their works from other artists. I think of the example of artists who illustrate superhero comic books. These artists often tell stories with the same characters, but present those characters in their own styles, such that you can easily tell one artist’s work from another. Scotti Young’s version of Spiderman is very different from Todd McFarlane’s or Jack Kirby’s. Each shares some similarities with the others — the character itself and the red and blue super suit — but are also very different. Indeed, they are so different, and have such strong styles, that you can recognize other works by the same artists, even without being told it’s them. Jack Kirby is one of the most influential comic artists of all time; if other artists create works that evoke his work, does that mean they are copying him? Moreover, while a character like Spiderman may be protected by copyright, it is not clear how the style in which he is portrayed is part of that. Where does the style end and the expression begin?
Additionally, some case law recognizes a degree of copyright protection over style. Professor Ed Lee, director of the program on intellectual property law at Chicago-Kent College of law, points to two cases on style and copyright: Malden Mills, Incorporated. v. Regency Mills and Steinberg v. Columbia Pictures Industries. In Malden Mills, the 2nd Circuit Court of Appeals held that two works were “substantially similar” because they shared a “style of representation” among other similarities. In Steinberg, the Southern District of New York wrote that “style is one ingredient of ‘expression.’”
In a similar vein, the 9th Circuit Court of Appeals in Williams v. Gaye upheld a jury verdict, finding that Robin Thicke and Pharell Williams copied Marvin Gaye’s Got to Give it Up with their song Blurred Lines. Notably, Thicke mentioned in several interviews that he was directly inspired by Marvin Gaye’s song. In fact, the dissenting judge in the case wrote that: “The majority allows the Gayes to accomplish what no one has before: copyright a musical style. ‘Blurred Lines’ and ‘Got to Give It Up’ are not objectively similar. They differ in melody, harmony, and rhythm. Yet by refusing to compare the two works, the majority establishes a dangerous precedent that strikes a devastating blow to future musicians and composers everywhere.” Notably, the majority in that case denied these claims made by the dissent, calling them “hyperbole” and writing specifically that the decision “does not grant license to copyright a musical style.”
I believe that the best way to think about style’s position in copyright law is that design elements like style, design, and look are all parts of artistic expression, but cannot alone belong to a single artist. When two works are stylistically similar, that is part of the bigger question of whether the later work is “substantially similar” to the former — but it is only part. Under copyright law, a secondary work does not need to be a verbatim or identical copy to constitute copyright infringement. Instead, the secondary work only needs to be “substantially similar” to infringe on the first. While substantial similarity is “an elusive concept, not subject to precise definition”, we know that two works have to be both extrinsically similar (based on their objective features) and intrinsically similar (based on what an ordinary person would subjectively think) for one to infringe on the other. Stylistic similarity can be part of these questions — are there objectively similar elements of the two works, more than standard elements that belong to a particular genre, and would an ordinary person think they are similar? — but it cannot be determinative of “substantial similarity” by itself.
Style and artistic expression are certainly linked, but they are not the same thing. In some instances, mimicking an artist’s style may constitute copyright infringement, but when this is true will necessarily require a case-by-case analysis. Simply working in a style that is similar to another artist does not automatically mean the work is a copy. Moreover, there may be legitimate reasons to produce works in another artist’s style. The law should allow artists to be inspired by each other, even if that means having similar styles. Doing otherwise would chill expression, ultimately harming human creativity and the purposes of copyright law itself.
Like the rest of the world, CC has been watching generative AI and trying to understand the many complex issues raised by these amazing new tools. We are especially focused on the intersection of copyright law and generative AI. How can CC’s strategy for better sharing support the development of this technology while also respecting the work of human creators? How can we ensure AI operates in a better internet for everyone? We are exploring these issues in a series of blog posts by the CC team and invited guests that look at concerns related to AI inputs (training data), AI outputs (works created by AI tools), and the ways that people use AI. Read our overview on generative AI or see all our posts on AI.
Note: We use “artificial intelligence” and “AI” as shorthand terms for what we know is a complex field of technologies and practices, currently involving machine learning and large language models (LLMs). Using the abbreviation “AI” is handy, but not ideal, because we recognize that AI is not really “artificial” (in that AI is created and used by humans), nor “intelligent” (at least in the way we think of human intelligence).