Can AI Steal an Artist’s Style? Legal Risks and IP Challenges in the Age of Generative Art

Explore legal risks of AI-generated art, copyright limits on artistic style, and how to protect creative work in the evolving IP and AI legal landscape.l

Dean Taylor

5/31/20256 min read

As AI-generated art explodes across social media, artists and IP lawyers face a pressing question: Can style itself be stolen—and if so, is it legally protected? This post explores a high-profile controversy surrounding OpenAI’s image generation tool and what it means for copyright, training data, and the legal rights of creators.

OpenAI recently released an image generation tool that has demonstrated capabilities beyond that of current competing models like Stable Diffusion. A controversial feature (which existed in other companies’ models previously) allows users to transform photos into the “style of” well known artists. A viral trend started when this new image generation model was released involved users converting images into a style known as Studio Ghibli.

Studio Ghibli is a world-famous Japanese animation studio established in 1985 by directors Hayao Miyazaki and Isao Takahata alongside producer Toshio Suzuki. The name “Ghibli,” inspired by the Arabic word for a hot desert wind, reflects the founders' mission to bring a fresh, innovative spirit to the animation industry. Since its inception, the studio has earned global recognition for its breathtaking hand-drawn animation, imaginative storytelling, and deep emotional resonance. Iconic films like My Neighbor Totoro (1988), Princess Mononoke (1997), and Spirited Away (2001),which won the Academy Award for Best Animated Feature in 2003, showcase Ghibli’s signature blend of fantasy, environmental themes and richly detailed worlds.

Since that time Studio Ghibli has become an animation and cultural cornerstone celebrated for its commitment to traditional animation techniques and its ability to craft stories appealing to all ages. Their films often explore profound topics, such as humanity’s relationship with nature, while delivering unforgettable characters and visuals. This dedication to artistry and meaning has influenced countless creators worldwide, cementing Studio Ghibli’s legacy as a trailblazer in animation history.

But all that time, effort, goodwill and audience enthusiasm - what is it worth. Is it worth protecting? Can it be protected?

Studio Ghibli's Position

Studio Ghibli hasn't issued an official statement, but Hayao Miyazaki, the co-founder, has criticized AI-generated art in general calling it “an insult to life itself.” Miyazaki's comments in a 2016 documentary also underline his displeasure with AI art. He expressed disgust at AI animation, stating, “I am utterly disgusted. If you really want to make creepy stuff you can go ahead and do it. I would never wish to incorporate this technology into my work at all,”

Unexpected Detail

An unexpected aspect is the viral social media trend of converting images to this style. Despite being embroiled in multiple infringement lawsuits, even OpenAI CEO Sam Altman joined the trend by changing his X profile picture to a Studio Ghibli-style image.

Legal Considerations

Copyright law protects specific expressions, but does not explicitly protect general styles. This complicates the question of whether studios developing recognized styles have any recourse. While generating images in Studio Ghibli's style might be permissible, using their works for AI training without permission could be a violation. I have written about here more than once regarding the ongoing cases litigating just that argument. Fair use is a potential defense, but courts are still deciding its applicability to AI models in text, audio, image and video generation.

Copyright and AI-Generated Art

Copyright law, particularly in the U.S., protects original works of authorship, including literary, dramatic, musical, and artistic works, but as mentioned above, not ideas or styles. This distinction is crucial for the Studio Ghibli case and others like it that have already been filed. Specific characters, scenes, and sequences from Studio Ghibli films are protected, but the general anime style, characterized by clean lines, vivid expressions, and stylized forms, is not explicitly covered by copyright. As it is currently interpreted, only the creation of actual characters or scenes that are contained in existing Studio Ghibli works would be protected.

The U.S. Copyright Office has previously ruled that works generated solely by AI without human input are not copyrightable. I wrote about the Thaler case previously (US appeals court rejects copyrights for AI-generated art lacking 'human' creator). However, the question here is not the copyrightability of the generated images but whether their creation by the OpenAI model itself violates Studio Ghibli's copyright.

Arguments for Copyright Violation

1. Training Data Concerns: If OpenAI trained its model on Studio Ghibli's copyrighted works without permission, this could constitute copyright infringement. Several ongoing cases have this issue as yet unresolved. Evan Brown, an intellectual property lawyer at Neal & McDevitt, noted in an interview that it's plausible OpenAI achieved the Ghibli style by training on millions of frames from the studio's films. The common defense to all of these AI related cases has been fair use. Such use might not qualify as fair use, especially if it's for commercial purposes, given ongoing lawsuits like the New York Times' case against OpenAI for using articles without attribution.

2. Derivative Works: If the generated images are too similar to specific Studio Ghibli works, they could be considered derivative works, requiring permission from the copyright holder. This is particularly relevant if the AI outputs replicate protected elements like character designs or iconic scenes, potentially violating Studio Ghibli's exclusive rights.

3. Ethical and Legal Precedents: The rapid advancement of AI has led to multiple lawsuits challenging the use of copyrighted data for training, such as the class action against Microsoft, GitHub, and OpenAI over GitHub Copilot.

Arguments Against Copyright Violation

1. Style Not Protected: Christa Laser, an IP law professor at Cleveland State University, argues that while individual works are protected, artistic style is not. Matthew Sag, an Emory University law professor, reinforced this, stating, “If you just evoke the vibe of somebody else's creative work, it generally doesn't violate their copyright,” emphasizing that anime styles are a cornerstone of AI image generation. Id.

2. Fair Use for Training: Training AI models on publicly available data, including copyrighted works, could be defended under fair use, particularly if the use is transformative. The RAND Corporation's paper on AI impacts on copyright law discusses this, noting that courts are still deciding whether such training falls under fair use protections.

3. User-Generated Content: The tool's output is user-initiated, with prompts like "turn this photo into Studio Ghibli style," suggesting a collaborative process that might dilute direct liability. This aligns with arguments that AI-generated art, when prompted by humans, could be seen as a new creative expression, not a direct copy.

Relevant Legal Precedents and Current Cases

Ongoing Lawsuits

Multiple lawsuits, such as those by the New York Times and visual artists against Stability AI, challenge the use of copyrighted material for AI training, with defendants arguing fair use as discussed above.

The Copyright Office's ongoing report on AI and copyright, issued in parts, addresses digital replicas and copyrightability, with Part 3 forthcoming on training data implications

Rob Rosenberg, former general counsel at Showtime, suggested Studio Ghibli might claim violations under the Lanham Act for false advertising or trademark infringement, though he noted it's a "complex question." Conversely, experts like Sag argue Studio Ghibli has few legal moves, given style isn't protected.

Fake Legal Letters Accompanying Fake Art

Some social media users have responded to expressed ethical and legal concerns by creating fake legal letters, adding to the controversy.

Conclusion and Practical Advice for You

1. Investigate Training Data: Normal discovery in these cases or any involving AI systems should include questions designed to determine if the AI model was trained on the client's copyrighted works, as this is a key area of contention. Review contracts and data sources used to train the model or any related models.

2. Assess Similarity: Evaluate examples of generated images your client brings to you for derivative work claims, focusing on specific protected elements rather than general style. Expert testimony on visual similarity may be crucial.

3. Stay Updated on Legal Developments: Monitor ongoing lawsuits, such as those against OpenAI and Stability AI, and the U.S. Copyright Office's forthcoming reports for guidance. Google alerts is a free tool you can use to track articles and events occurring in relevant cases.

4. Consider Fair Use Defenses: Prepare arguments on fair use, emphasizing transformative use and public benefit, but be aware courts are still deciding this for AI training.

5. Explore Alternative Claims: If copyright claims are weak, consider trademark or Lanham Act claims for false advertising, especially if the tool misrepresents affiliation with your client’s works as may be the situation with Studio Ghibli.

I wish I was good enough at drawing or painting to make that type of art. Just wasn’t something I was gifted with in this life. I can sympathize with skilled artists spending years developing a unique style and then watching, starting in 2023, a variety of AI models able to produce an endless supply of images in just their style. In contrast, as I have written previously, if the courts impost copyright restrictions on these models, their innovation will not entirely stop, but it will become prohibitively expensive.

This is the challenge of the legal and political system going forward. Protecting copyright and other IP rights but not slowing innovation as countries perceived as our adversaries are likely not worried as much about IP rights and can innovate more easily.