As part of an ongoing legal dispute with three Hollywood studios, generative AI startup Midjourney is asking the studios to disclose how they use artificial intelligence technology in their operations. Last year, Disney and Universal Pictures sued Midjourney for alleged copyright infringement, accusing its image generation model of being able to generate images of copyrighted characters including Bart Simpson and Darth Vader. Warner Bros. subsequently filed a similar lawsuit.
Midjourney contends that using images of copyrighted characters to train its AI model falls within the scope of fair use. At issue is the scope of documents studios are required to submit during the discovery phase. Previously, judges have ruled that the studios must provide information related to their use of generative AI, but this obligation is limited to the final output of "consumer-oriented" videos and images.
In a new court filing, Midjourney seeks to overturn the restriction, arguing that the definition is "unfair" and allows studios to "selectively submit documents they believe support their claims of market harm, while depriving Midjourney of documents that would assist in its defense." The company further claims that the documents currently being withheld from the studio "exactly reveal whether they were doing the same conduct behind the scenes as alleged in the lawsuit against Midjourney."
For example, Midjourney pointed out that if these studios are developing image-generating AI models for in-house lens design or creative conception of film and television content, then this evidence will also show that downloading and using unauthorized copyrighted content to train AI is an industry practice even within the studio. In other words, Midjourney is trying to weaken its case against Midjourney by showing the court that major Hollywood studios may rely on similar technology and data sources in their internal processes.
In the document, Midjourney also argued that studios should disclose all text prompts (prompts) they use on the Midjourney platform and the corresponding generated results, not just those that are alleged to have produced allegedly infringing images. By expanding the scope of the disclosure, Midjourney hopes to take a more comprehensive look at how studios use generative AI tools in actual creative processes to support their "industry practice" and "fair use" defense logic.
David Singer, the lead attorney representing the studio, previously accused Midjourney of requesting evidence as a "casting net" operation aimed at indiscriminately obtaining a large amount of internal information. He also stated that the studio "does not seek to prevent the development of AI technology, nor does it even intend to close Midjourney's business," but "simply hopes that Midjourney will stop copying the content of their movies and TV series, stop disseminating, publicly displaying, publicly broadcasting and creating derivative works containing replicas of their famous characters without authorization."
The current legal dispute surrounding the boundaries of generative AI training data and copyright is still advancing. The battle between Midjourney and Hollywood studios over the scope of discovery is becoming an important part of clarifying the norms and responsibilities of AI use in the industry.