#CopyrightLaw

Studio Ghibli-Style AI Images and the Legal Questions They Raise

/ Washington Journal of Law, Technology & Arts / Leave a comment
Photo by FOX ^.u11bd.^= u222b on Pexels.com

By: Esha Kher

In recent weeks, AI-generated images mimicking the iconic look of Studio Ghibli have gone viral across platforms like X and Instagram, sparking controversy. Selfies, family portraits, and memes have been transformed into soft, pastel-hued depictions that echo the dreamlike aesthetic of the legendary Japanese animation studio. Founded in 1985 by Hayao Miyazaki, Isao Takahata, and Toshio Suzuki, Studio Ghibli is renowned for its rich storytelling and distinctive, heartwarming visual style—now replicated widely through AI.

The recent trend in which users generate AI images using the latest version of OpenAI’s GPT-4o, mimicking Studio Ghibli’s aesthetic, has gained immense traction on social media, even causing server overloads. This emerging trend has sparked considerable debate, dividing public opinion into two camps: AI enthusiasts and staunch critics of AI-generated art. Supporters view the phenomenon as a tribute to Studio Ghibli’s influence and a democratization of creative tools. Critics, however, find this trend to be both a hollow and inauthentic imitation of Hayao Miyazaki’s distinct style—devoid of creative soul or artistic merit. An old clip of Miyazaki himself resurfaced during the controversy, in which he vehemently denounced AI-generated imagery as “an insult to life itself”. 

Regardless of one’s stance on the debate, the trend raises important legal questions. Do AI models rely on copyrighted material to replicate distinct visual styles? And when these outputs resemble a studio’s recognizable aesthetic, like that of Studio Ghibli, do they risk infringing on copyright or trademark protections?

How ChatGPT Generates “Stylized” Images

GPT-4o uses an autoregressive algorithm to generate images by breaking them down into visual “tokens,” which function like words in a sentence. Just as ChatGPT predicts the most likely sequence of words in a sentence, generative image models predict and assemble these visual tokens to form coherent images. Through training on large datasets of images and text, the model learns to associate certain patterns, like color palettes or brushstrokes, with specific words, encoding them as abstract “styles” within its neural network. So when a user references “Studio Ghibli,” the model doesn’t retrieve frames from actual films but instead draws on a learned mathematical representation of the studio’s aesthetic (otherwise known as “Ghibli-ness”). This ability to isolate and apply visual features across new images is known as a “style engine”. 

Copyright Law Implications 

The use of style engines has raised entirely new questions about copyright law and creative ownership. U.S. copyright law doesn’t protect certain artistic styles, the law only protects the unique ways those styles are expressed or, in other words, original works of authorship. 

However, legal experts caution that this distinction may not be sufficient. While “style” in the abstract is not copyrightable, what people casually refer to as “style” may include recognizable, discrete elements of a work of art. Therefore, the blanket statement that an artistic style isn’t protectable under copyright law may not be absolute. Courts may still find infringement if generated images include elements that are original, expressive, and substantially similar to the copyrighted works.

This legal ambiguity is at the heart of Andersen v. Stability AI, a landmark lawsuit filed in 2022 by three visual artists against AI companies Stability AI, Midjourney, and DeviantArt. The artists allege that these companies used their copyrighted artworks without consent to train AI models like Stable Diffusion, which can generate images that imitate their distinctive styles. The plaintiffs argue that such outputs constitute derivative works, and even if the results aren’t direct copies, the unauthorized use of their works in training data alone may amount to copyright infringement. Similar concerns have surfaced in other lawsuits, including Huckabee v. Meta and Millette v. Nvidia, where creators claim that their content was scrapped and repurposed by generative AI platforms, raising serious questions about how copyright applies in the context of machine learning.

Further, there is growing concern that OpenAI may have used Studio Ghibli’s films and artwork to train its generative image model without prior consent from the animation studio. This could constitute copyright infringement if the works were repurposed in a way that exceeds the scope of what is allowed under the fair use doctrine. The fair use doctrine permits the use of copyrighted material for specific purposes, such as academic research or the creation of an entirely new invention. OpenAI maintains that training its models is fair use under copyright law, but this defense is largely untested. 

Trademark Law Implications

Beyond copyright law, Studio Ghibli could assert that OpenAI’s generation of “Ghibli-style” images infringes upon its trademark rights under the Lanham Act. While an animation style—such as Studio Ghibli’s distinctive visual aesthetic—is not independently protected by trademark law and does not trigger the traditional “likelihood of confusion” test used in trademark infringement claims, other aspects of trademark law may still apply. 

When there is no registered trademark involved, Section 43(a) of the Lanham Act (15 U.S.C. § 1125(a)) provides broader protection by prohibiting false endorsement, sponsorship, or affiliation. Under this provision, a claim can arise when: (1) a defendant uses elements closely associated with a person or brand, such as names, visual likenesses, or identifying characteristics, and (2) this use creates a false impression in the minds of consumers that there is a connection, endorsement, or affiliation with the original brand. 

In this context, while Studio Ghibli may not have a registered trademark on its animation “style” per se, OpenAI’s promotion of “Ghiblification” experiments—along with OpenAI employees sharing Ghibli-style portraits of themselves on social media—could potentially give rise to a false endorsement claim under § 43(a). This is especially true if such references imply to the public that Studio Ghibli has authorized or collaborated with OpenAI in developing these tools. Even allowing prompts such as “in the style of Studio Ghibli” could lead consumers to mistakenly believe that Studio Ghibli has endorsed or is affiliated with the image generation process. While this may not amount to traditional trademark infringement, it opens the door to a false association claim under the broader protections of the Lanham Act. 

Conclusion

The rise of AI-generated art in the style of Studio Ghibli underscores the growing legal uncertainty surrounding copyright and trademark protections.​ As of now, Studio Ghibli has not initiated any legal action against OpenAI regarding the AI-generated images mimicking its distinctive animation style. Nonetheless, there might be grounds to pursue action under U.S. law. While U.S. copyright law does not formally protect artistic styles, the line becomes blurry when AI outputs closely resemble original works in expression and substance. At the same time, Studio Ghibli may have a claim under trademark law, particularly if the AI-generated images create consumer confusion or falsely present an endorsement by the studio. By capitalizing on Ghibli’s recognizable aesthetic, OpenAI risks infringing on the studio’s brand and artistic reputation. These unresolved legal questions underscore the need for updated legal frameworks that account for how AI systems produce and distribute creative content without prior consent from creators.

Reel Rights: Copyright’s Collision With Documentaries

/ Washington Journal of Law, Technology & Arts / Leave a comment
Photo by Ryutaro Tsukata on Pexels.com

By: Alexander Tranquill

The Rise of Tiger King

Remember the start of the Covid-19 pandemic? Like myself, many Americans turned to TV to find comfort, and, in those first few surreal weeks, many found themselves watching one particularly enthralling, peculiar, and utterly outlandish story: Tiger King. Tiger King is a Netflix documentary series released in March 2020, which details the increasing tensions between rival big cat eccentrics, eventually culminating in Joe Exotic’s arrest in a murder-for-hire plot of rival Carole Baskin. While Tiger King initially generated massive media attention, it has more recently been the subject of intense copyright litigation.

If you are unfamiliar with this story, Joe Exotic was an internet personality long before Tiger King. With a substantial presence on YouTube, the Netflix documentary heavily relied on video footage originally created by Exotic and his employees. The suit now at issue, Whyte Monkee v. Netflix, centers on Netflix’s use of a video that shows Exotic giving an eulogy at his late husband’s funeral. The video was originally shot by Timothy Sepi, an employee and videographer at Exotic’s Gerald Wayne Zoo. However, Sepi now claims he never gave Netflix permission to use his footage, thus forming the basis for his copyright infringement claim.

During litigation, the district court originally found that Netflix’s use of the footage fell under the fair use exception to copyright infringement. This decision was later reversed by the Tenth Circuit, but, after a great deal of consternation and a flurry of amicus briefs, the Tenth Circuit later vacated its ruling and granted a petition for rehearing. Though the final decision is still pending, this case is significant because the decision has  major implications for documentary filmmakers, while also raising important questions about the rights of content creators in our age of smartphones and social media––where personal footage is often reused by others.

Copyright Protections and Fair Use

To understand the legal questions raised in Whyte Monkee, we must explore the interaction between copyright protections and the doctrine of fair use. Overall, copyright is a type of intellectual property that protects original works of authorship (i.e. paintings, photos, writings, movies) against use by others. However, authorship is a fairly low bar, requiring only a minimal level of creativity. If any creativity can be shown, copyright protections immediately attach when the work is fixed (published) in a tangible medium. As a result, recordings, much like Sepi’s home-video, are often considered copyrightable.

The fair use exception to copyright infringement allows a party to use a work without the permission of the creator if the copying is done for a limited or “transformative” purpose. While there are no hard and fast rules, courts will consider four factors in determining fair use: (1) the purpose and character of the use, (2) the nature of the copyrighted work, (3) the amount and substantiality of the portion used, and (4) the effect of the use on the market for the copyrighted work. Recently, courts have taken particular interest in the first factor, considering the significance of any changes made to the original while also assessing the purpose of the work.  

The Supreme Court’s Warhol Decision

Recently, the Supreme Court issued a detailed analysis of this first factor in Warhol v. Goldsmith. Here, Warhol was sued for a series of silkscreen prints he created of Prince, which he based on a copyrighted image captured by Goldsmith. Considering the purpose and character of Warhol’s silkscreens, the Court clarified that it is no longer sufficient for a work to simply add “new expression” to the original; the key question is whether the work serves “a purpose distinct from the original.” While purpose is not necessarily limited, derivative works should comment on, criticize, or provide otherwise unavailable information to the original. Therefore, although Warhol added new artistic expression to the original, his work did not constitute fair use because its purpose and character aligned with Goldsmith’s––both works were licensed to media companies, merely being used “to illustrate a magazine about Prince with a portrait of Prince.” Thus, because Warhol’s work simply used Goldsmith’s image as a template for the same commercial purpose, it failed to seriously comment on, criticize, or add information to the copyrighted image. 

Applying the Warhol Precedent

Following in the footsteps of Warhol, the Tenth Circuit overturned the district court’s decision in Whyte Monkee, relying on the purpose and character of the use. While the district court found Netflix’s use transformative as it incorporated the funeral clip into a broader narrative, the Tenth Circuit, citing Warhol, concluded that fair use requires the derivative work to serve a distinct purpose. Specifically, the court required the derivative work to critically comment “on the substance or style of the original composition.” With this backdrop, the court found that Netflix only used the funeral footage to show Exotic’s purported megalomania and showmanship. Accordingly, Netflix failed to seriously comment on the style of the video clip itself, instead using the video to “target[] a character in the composition.” Therefore, because Netflix used the funeral footage to detail Exotic’s life and not to comment on the style of the footage, the Tenth Circuit found that the first factor weighed against fair use. 

So, what is all the uproar about? The amicus briefs suggest that this decision will have a chilling effect on the documentary industry, confining filmmakers to commenting on the composition of footage itself (i.e. lighting, angles, editing). In the en banc rehearing, the court re-examined Netflix’s intent behind including the funeral clip in the documentary, focusing on Netflix’s use of the video to detail Exotic’s callous attitude. Thus, the court’s review likely reflects an effort to broaden the meaning of transformative purpose.

In its initial decision, the Tenth Circuit severely narrowed Warhol’s definition of distinctive purpose, requiring a derivative work to critically comment “on the substance or style of the original composition.” While this was an important factor in Warhol, it is not the only relevant factor in examining the purpose and character of a new work. First, the Court in Warhol explicitly looks to purposes outside critical commentary to determine fair use. For example, the Court found that Warhol and Goldsmith’s works shared the same commercial purpose—both were used to “illustrate a magazine about Prince with a portrait of Prince.” Furthermore, the Court in Warhol notes that the “degree of difference” between the works is relevant in the fair use analysis, being weighed together with purpose to determine whether the derivative work is transformative. Ultimately, the Tenth Circuit seems poised to consider other factors in its analysis of the purpose and character of the use. Such a decision would better support filmmakers by providing them greater access to material as they attempt to capture many of the compelling narratives in our world today. As a result, Netflix should continue to assert that its use of the video serves to illustrate Exotic’s personality—distinct from Sepi’s purpose of simply commemorating the funeral. Further, Netflix should revive its district court arguments, claiming the documentary is substantially different from Sepi’s video because it continually interrupts this video with comments from the deceased’s mother and ties the video into the broader story arc to highlight Exotic’s character. Ultimately, these arguments mirror the Supreme Court’s focus in Warhol, offering the Tenth Circuit a precedential foundation to recognize a broader interpretation of transformative use.