Disney and Universal Sue Midjourney: Navigating AI Copyright Challenges and Fair Use Implications

On June 11, Disney and Universal filed a complaint in the Central District of California against Midjourney, a popular text-to-image AI tool.

The plaintiffs allege that Midjourney trained its AI models on data that includes the studios’ copyrighted works, including a variety of fictional characters. The plaintiffs provide examples of instances where Midjourney’s output closely resembled or directly copied the plaintiffs’ copyrighted works. Although Midjourney has not filed an answer, Disney and Universal assert in their complaint that “whether an infringing image or video is made with AI or another technology does not make it any less infringing.”

[1]

The new case against Midjourney is just the last of several dozen copyright cases pending against AI developers. Although no court has issued a decision on the merits against a developer of generative AI (GenAI), federal judges in several cases are beginning to grapple with the fact-intensive nature of the fair use doctrine and the impact of an emerging licensing market for training data.

Background

Fair use is an affirmative defense against copyright infringement, codified under Section 107 of the Copyright Act. Courts balance four factors:

1. The purpose and character of the use, including whether such use is of a commercial nature or is for nonprofit educational purposes.
2. The nature of the copyrighted work.
3. The amount and substantiality of the portion used in relation to the copyrighted work as a whole.
4. The effect of the use upon the potential market for or value of the copyrighted work.

Two Federal Judges Hint at GenAI Fair Use Analysis

Even before the Midjourney action was filed, two judges in the Northern District of California provided early indications of how they may analyze the fair use factors in cases involving GenAI technologies. Both cases were filed by authors who allege that AI developers infringed their copyrights by using the authors’ copyrighted books as training data without authorization or license.

In a recent summary judgement hearing, Judge Vince Chhabria suggested that the fourth fair use factor, which requires judges to consider the impact of the unauthorized use on the market for the original work, may be most influential in his analysis. But while he seemed convinced that AI tools will affect the literary market as a whole, he noted that there seemed to be a lack of evidence that the plaintiffs’ specific works would be affected.

In another recent hearing, Judge William Alsup suggested that an AI developer violated the Copyright Act when it downloaded pirated versions of the plaintiffs’ copyrighted works, but that the developer’s subsequent uses of those pirated copies to train its AI models may qualify as fair use. Much of Judge Alsup’s concerns seemed to center on the fact that the developer obtained the plaintiffs’ works through known piracy websites without paying any fees. Judge Alsup cautioned that he is “leaning” towards this finding but could change his mind when he ultimately rules on the Anthropic’s motion for summary judgment. As a remedy, he suggested that the court may require the developer to pay for the initial copies it acquired.

Questions About a Potential Licensing Market

AI developers in both cases have argued that there is no cognizable licensing market for AI training data. Nevertheless, several owners of copyrighted works have already entered into partnerships with AI developers to license protected works, which suggests at least a “potential” licensing market. For example, media companies that own large amounts of copyrighted works, such as the Associated Press and The New York Times, have reached deals with some AI developers to license their copyrighted works. The US Copyright Office has also encouraged AI developers to pursue licensing options for training data when available.

Key Takeaways

Owners of copyrighted works and AI developers should closely monitor these and other AI copyright cases, which could provide some of the earliest indications of whether use of copyrighted content to train AI models is fair use, and if so, under what circumstances.

The AI, Metaverse, and Blockchain team at ArentFox Schiff is closely monitoring developments in the various copyright infringement lawsuits against AI developers, and will provide critical updates and analysis on these issues. In the meantime, please reach out to a member of our team with any questions about how these pending decisions may impact your business.