Hollywood Titans Take on AI: Disney and Universal Sue Midjourney for Copyright Infringement

In a significant escalation of the ongoing legal battles surrounding artificial intelligence and intellectual property, two of Hollywood's most powerful studios, The Walt Disney Company and NBCUniversal (owned by Comcast), have filed a joint lawsuit against Midjourney. The San Francisco-based startup, known for its popular AI image generation service, stands accused by the entertainment giants of being a "bottomless pit of plagiarism" that produces "endless unauthorized copies" of their vast library of copyrighted works.

The lawsuit, filed in a US court, thrusts major film studios directly into the legal fray that has, until now, primarily involved artists, writers, and news organizations challenging AI companies over the use of their work for training data and the generation of potentially infringing outputs. While dozens of copyright lawsuits against AI companies are already navigating the complexities of the US legal system, including a class action lawsuit brought by visual artists against Midjourney in 2023, this action by Disney and Universal carries particular weight due to the plaintiffs' immense resources and their historical vigilance in protecting their intellectual property.

Allegations of Direct Infringement Through AI Output

At the heart of the complaint are specific examples provided by Disney and Universal, purportedly demonstrating Midjourney's capacity to generate images that closely replicate their copyrighted characters and designs. The studios argue that these outputs are not merely inspired by their work but are, in fact, infringing copies.

The complaint includes several images generated by Midjourney, presented as evidence of the alleged infringement. One notable example cited is an image depicting Yoda, the iconic Star Wars character. According to the lawsuit, this image was produced simply by inputting the prompt "Yoda with lightsaber, IMAX." The resulting image, the studios contend, is an unauthorized reproduction of their character.

Another example involves the animated film franchise The Boss Baby, owned by Universal. The studios allege that typing "The Boss Baby" as a prompt into Midjourney resulted in an image of an animated child in a tuxedo that bears a striking resemblance to the protagonist of the films. These examples are intended to illustrate the studios' claim that Midjourney's AI model directly facilitates the creation of infringing content based on their protected characters.

IP lawyer Chad Hummel commented on the significance of these examples, stating that the compilation of images in the complaint serves as compelling evidence that "the output is not sufficiently transformative." This point is crucial because, in copyright law, the doctrine of "fair use" can permit the use of copyrighted works under certain circumstances, particularly if the new work is deemed "transformative"—meaning it adds a new meaning, message, or expression to the original material.

The Fair Use Debate: Training Data vs. Output

Most AI companies facing copyright challenges have relied heavily on the fair use defense, arguing that training their models on vast datasets of copyrighted material constitutes a transformative use because it enables the creation of entirely new works, rather than simply reproducing the training data. However, the lawsuit brought by Disney and Universal takes a different approach, directly attacking the nature of Midjourney's *output*.

Matthew Sag, a professor specializing in law and artificial intelligence at Emory University, believes this distinction makes the studios' case potentially stronger and Midjourney's fair use defense more challenging. "The reason it’s different is that Disney directly attacks the output of the model," Sag explained. "It doesn’t just use a few cherry-picked examples to prove that the model was trained on its works."

Sag elaborated on the difficulty Midjourney may face in arguing that generating images closely resembling copyrighted characters is transformative. "It’s going to be very difficult for a court or a jury to accept that it is transformative to take 1,000 pictures of Darth Vader and use them to produce even more pictures of Darth Vader," he noted. This perspective highlights the core of the studios' argument: that the AI is not creating something fundamentally new, but rather generating derivative works that are too close to the originals to qualify as fair use.

Allegations of Ignored Demands and Training Data Practices

Beyond the issue of output, the lawsuit also touches upon Midjourney's practices regarding its training data and its alleged lack of responsiveness to the studios' concerns. The complaint alleges that Disney and Universal had previously requested Midjourney to "adopt technological measures" to prevent its image generators from producing infringing materials. According to the studios, Midjourney "ignored" these demands.

Furthermore, the lawsuit claims that Midjourney "cleaned" copies of Universal and Disney's work during its training process, which the studios argue "necessarily included creating more copies of the materials." This points to the controversial practice of scraping the internet for training data, a method openly discussed by Midjourney's founder.

In a 2022 interview with Forbes, Midjourney CEO David Holz described their training process: "It’s just a big scrape of the internet. We use the open data sets that are published and train across those." He acknowledged the difficulty in identifying copyright owners for the vast number of images used, stating, "There isn’t really a way to get a hundred million images and know where they’re coming from. It would be cool if images had metadata embedded in them about the copyright owner or something. But that's not a thing; there's not a registry.” This perspective contrasts sharply with the studios' assertion that Midjourney should have taken steps to prevent infringement, implying a responsibility to manage or filter their training data more carefully.

The Emerging Market for AI Training Data Licensing

The legal challenges surrounding AI training data have spurred the development of a new market focused on licensing creative works specifically for AI models. As lawsuits have highlighted the potential legal risks of using unlicensed data scraped from the web, startups and organizations are emerging to act as intermediaries between content creators and AI companies.

Companies like ProRata and organizations such as the Dataset Providers Alliance are examples of this burgeoning ecosystem, facilitating ethical licensing agreements for training data. This trend suggests a potential path forward for AI companies to acquire data legally, although it requires significant effort and cost.

Even companies involved in ongoing AI litigation are engaging in licensing deals. The New York Times, which is currently in a lawsuit with Microsoft and OpenAI over copyright issues related to training data and output, recently inked a deal with Amazon to license its journalistic works for AI training. Similarly, WIRED's parent company, Condé Nast, signed a licensing agreement with OpenAI in August 2024. These developments underscore Hummel's point that "What’s very clear now is that there is a vibrant market for licensing," suggesting that AI companies have alternatives to simply scraping the internet.

An Existential Threat to the Film Industry?

In previous AI copyright cases, judges have often emphasized the need for plaintiffs to demonstrate tangible financial harm caused by the AI companies. The lawsuit filed by Disney and Universal frames the issue not just as individual instances of infringement but as a fundamental threat to the economic viability of the motion picture industry.

The complaint asserts that "Midjourney’s bootlegging business model and defiance of US copyright law are not only an attack on Disney, Universal, and the hard-working creative community that brings the magic of movies to life, but are also a broader threat to the American motion picture industry which has created millions of jobs and contributed more than $260 billion to the nation’s economy." This language positions the lawsuit as a defense of the entire industry's economic foundation, arguing that unchecked AI-generated copies undermine the value of their creative output and the livelihoods of those who create it.

The studios are essentially arguing that if AI can generate high-quality, unauthorized reproductions of their characters and stories on demand, it could devalue their intellectual property, reduce the incentive for creative investment, and ultimately harm the industry's ability to generate revenue and employ artists, writers, actors, and countless others.

AI's Rapid Integration and Differing Views in Hollywood

Despite the legal challenges and concerns raised by unions like SAG-AFTRA, which fought for guardrails on AI use during contract negotiations, generative AI technology is already rapidly changing aspects of the film industry. Companies continue to develop increasingly sophisticated tools, such as Google's Veo 3 video generator, capable of producing realistic video content quickly.

Within Hollywood, there are differing perspectives on the adoption of AI. While some view it with caution or alarm due to potential job displacement and copyright issues, others see its potential as a creative tool. Notably, acclaimed director James Cameron, known for blockbusters like Titanic and Avatar, has joined the board of AI image generator Stability AI and advocates for its use to potentially reduce filmmaking costs and enhance creative possibilities. This division highlights the complex reality of AI's integration into the industry, where legal battles are unfolding alongside technological adoption and creative experimentation.

The Potential Impact of a Disney/Universal Victory

The outcome of the lawsuit between Disney/Universal and Midjourney has the potential to set significant precedents for the generative AI industry. Disney, in particular, has a long-standing reputation for being fiercely protective of its intellectual property, famously extending copyright terms and pursuing legal action against unauthorized uses of its characters, including the iconic Mickey Mouse, parts of which recently entered the public domain (though later versions remain protected).

Given the studios' resources and determination, this case is likely to be a protracted and closely watched legal battle. A ruling in favor of Disney and Universal could force AI image generators like Midjourney to implement stricter filtering mechanisms to prevent the creation of images resembling copyrighted characters, potentially requiring significant changes to their models or prompting systems. It could also strengthen the argument that generating derivative works closely based on existing IP, even through an AI, constitutes infringement and is not protected by fair use.

Conversely, if Midjourney successfully defends itself, perhaps by arguing that its output is indeed transformative or that the studios' claims of direct infringement are unfounded or difficult to prove on a large scale, it could embolden other AI companies and influence the direction of future AI development and deployment in creative fields.

Regardless of the outcome, this lawsuit underscores a critical tension at the intersection of rapid technological advancement and established legal frameworks designed to protect creative works. It highlights the urgent need for clarity on how existing copyright law applies to generative AI and whether new legal or licensing models are necessary to balance innovation with the rights of creators and copyright holders. The battle between Hollywood's giants and a leading AI startup is poised to shape the future landscape of both industries.

The Core Legal Arguments Explored

To fully appreciate the stakes of this lawsuit, it's important to delve deeper into the legal concepts at play, particularly copyright infringement and the defense of fair use. US copyright law grants creators exclusive rights to reproduce, distribute, perform, display, and create derivative works based on their original creations. Disney and Universal argue that Midjourney's AI is directly violating these rights.

Their primary claim rests on the generation of images that are substantially similar to their copyrighted characters. The examples of Yoda and The Boss Baby are intended to serve as proof points. For a court to find infringement, the studios must demonstrate two things: ownership of valid copyrights (which is undisputed for their major characters) and that the defendant (Midjourney) copied protected elements of their work. In the context of AI, "copying" can refer both to the ingestion of copyrighted data during training and the generation of output that is substantially similar to the original.

Midjourney's likely defense will center on fair use. The fair use doctrine is a limitation on copyright, allowing limited use of copyrighted material without permission for purposes such as criticism, comment, news reporting, teaching, scholarship, or research. Courts typically analyze four factors to determine if a use is fair:

1. **The purpose and character of the use:** Is it commercial or non-profit educational? Is it transformative? Disney and Universal argue Midjourney's use is commercial and the output is not transformative. Midjourney might argue its AI is a tool for user creativity, and the training process itself is transformative.
2. **The nature of the copyrighted work:** Is it factual or creative? Creative works like characters receive strong protection.
3. **The amount and substantiality of the portion used:** How much of the original work was used? In AI training, this involves ingesting potentially millions of images. In output, it's about the similarity of the generated image to the original character.
4. **The effect of the use upon the potential market for or value of the copyrighted work:** Does the new use harm the market for the original or derivative works? Disney and Universal argue that AI-generated copies directly compete with and devalue their official merchandise, films, and licensed products, posing an "existential threat."

The studios' focus on the *output* being non-transformative is a key strategic move. While courts are still grappling with whether training AI models on copyrighted data constitutes fair use, arguing that the *result* of the AI process is directly infringing derivative work might be a more straightforward path to proving harm and lack of transformation, especially with compelling visual examples.

The Training Data Conundrum

While the lawsuit emphasizes infringing output, the issue of training data remains a significant backdrop. Midjourney's admission that it used a "big scrape of the internet" highlights a common practice among early generative AI models. This scraping often includes vast amounts of copyrighted material, from images and text to music and code, collected without explicit permission or licensing.

The legal argument against this practice is that the act of copying material to train an AI model constitutes unauthorized reproduction, a violation of copyright. AI companies counter that this copying is internal, temporary, and transformative, enabling the creation of a new capability (the AI model) rather than distributing the original works. They argue it's akin to a human artist studying thousands of paintings to learn technique and style.

However, critics, including the plaintiffs in various lawsuits, argue that the scale and nature of AI training are fundamentally different from human learning and require licensing. The emergence of a licensing market, as evidenced by deals like NYT-Amazon and Condé Nast-OpenAI, suggests that the industry is moving towards recognizing the need to compensate creators for the use of their work in training data. Disney and Universal's lawsuit, while focusing on output, implicitly challenges the legitimacy of a training process that relies on unlicensed scraping, especially if that process directly leads to infringing outputs.

Economic Implications and the Future of Creativity

The lawsuit's framing of AI infringement as an "existential threat" to the film industry underscores the deep economic anxieties within creative sectors. The concern is that if AI can generate content that mimics the style, characters, or even specific works of human creators without compensation or licensing, it could disrupt traditional business models, devalue creative skills, and lead to job losses.

For studios like Disney and Universal, their characters and franchises are their most valuable assets. The ability of an AI to generate unauthorized images of Mickey Mouse, Spider-Man (licensed by Sony but part of the Disney/Marvel universe), or characters from Universal's extensive library could dilute the value of their brands, impact merchandise sales, and potentially even affect box office revenue if AI-generated content becomes a substitute for official productions or licensed fan art.

The lawsuit highlights the tension between technological innovation and the protection of creative labor. Proponents of generative AI argue that it is a powerful tool that can enhance human creativity, democratize content creation, and open up new artistic possibilities. Opponents, including many artists and unions, fear that it is a tool for mass infringement that threatens their livelihoods and the value of human artistry.

The outcome of this case could influence how AI companies develop and deploy their models, potentially pushing them towards licensed datasets or implementing more robust filtering to avoid generating copyrighted material. It could also impact future negotiations between studios, unions, and AI companies regarding the ethical and legal use of AI in production workflows.

Broader Landscape of AI Copyright Litigation

The Disney/Universal lawsuit is not occurring in a vacuum. It is part of a growing wave of litigation challenging various aspects of generative AI. These cases involve different types of creative works and different legal theories:

• **Visual Artists vs. AI Image Generators:** Several lawsuits have been filed by groups of artists against companies like Midjourney, Stability AI, and DeviantArt, alleging that their works were included in training datasets without permission and that the AI outputs are infringing.
• **Authors vs. AI Text Generators:** Prominent authors have sued companies like OpenAI and Meta, claiming their books were used to train large language models without consent, and that these models can generate text that infringes their copyrights.
• **News Organizations vs. AI Companies:** The New York Times' lawsuit against OpenAI and Microsoft is a high-profile example, alleging infringement through training data and the generation of output that competes with or misrepresents NYT content.
• **Code Copyright:** Lawsuits have also been filed regarding the use of copyrighted code from platforms like GitHub to train AI coding assistants.

Each of these cases explores different facets of the AI copyright problem, from the legality of training data scraping to the transformative nature of AI output and the potential for market harm. The Disney/Universal case adds the significant weight of major entertainment studios and focuses specifically on the output of character-based imagery, a core asset of the film industry.

The legal landscape is still very much in flux, with courts across the US grappling with novel questions raised by generative AI. There is no established consensus on how existing copyright law applies, and it is possible that new legislation may eventually be needed to provide clarity.

Conclusion: A Defining Moment for AI and IP

The lawsuit filed by Disney and Universal against Midjourney represents a pivotal moment in the intersection of artificial intelligence and intellectual property law. By directly challenging a leading AI image generator on the grounds that its output constitutes unauthorized copies of their iconic characters, the studios are pushing the boundaries of the legal debate beyond just the issue of training data.

The case will likely hinge on the interpretation of fair use and whether generating images that closely resemble copyrighted characters, even with user prompts, is considered transformative or infringing. The studios' argument that this practice poses an "existential threat" to the film industry highlights the significant economic stakes involved.

As this lawsuit progresses alongside other ongoing AI copyright cases, the outcomes will undoubtedly help shape the future of generative AI development, deployment, and regulation. They will influence how AI companies source and use data, the types of outputs they are permitted to generate, and the potential for compensation or licensing models for creators whose work contributes to AI capabilities. The battle between Hollywood's most powerful storytellers and the creators of powerful AI tools is set to define the legal and economic boundaries of artificial creativity for years to come.