The legal battle between AI image generator Midjourney and three of Hollywood’s largest studios has taken a sharp procedural turn, with the startup now demanding that Disney, Universal, and Warner Bros. disclose how they themselves use generative artificial intelligence. At the heart of the dispute is a discovery motion that could force the studios to hand over internal documentation about their own AI training practices, potentially revealing whether they engage in the same kind of unlicensed content usage they are suing Midjourney over. The motion, if granted, would fundamentally reshape the trajectory of one of the most closely watched copyright cases in the emerging field of generative AI, and it raises questions about industry-wide double standards that extend far beyond the specific allegations in the lawsuit.
The Core Dispute: Fair Use vs. Infringement
Disney and Universal filed suit against Midjourney last year, alleging that the startup’s image-generation models could reproduce copyrighted characters such as Bart Simpson and Darth Vader. Warner Bros. followed with its own lawsuit a few months later, citing similar concerns over characters like Superman and Batman. The central legal question in each case is whether training AI models on copyrighted images constitutes infringement or is protected under fair use doctrine. Midjourney has consistently argued that its training methodology falls within fair use, a position that has drawn both support and criticism from legal scholars and industry observers.
The fair use defense in AI training cases remains unsettled law, and the outcomes of these studio lawsuits could set precedents that affect how generative AI companies operate in the United States for years to come. Midjourney’s strategy of turning the tables on the plaintiffs — requesting that the studios reveal their own AI usage — represents an attempt to demonstrate that the industry standard is far murkier than the studios’ legal complaints suggest.
What Midjourney Is Asking the Court to Compel
The immediate procedural fight centers on the scope of documentation that the studios must produce during discovery. A judge previously ruled that the studios are required to provide information about their generative AI usage, but only when that usage led to consumer-facing videos and images. Midjourney, in its latest filing, is asking the court to remove that limitation, arguing that it unfairly allows the studios to cherry-pick documents that support their claims of market harm while withholding material that would support Midjourney’s defenses.
The startup contends that the documents the studios are withholding are precisely those that would reveal whether, behind closed doors, they are doing exactly what they are suing Midjourney for doing. Specifically, Midjourney argues that if the studios are developing image-generating AI models for internal use in storyboarding or ideating content for film or television, that evidence would equally demonstrate that it is an industry custom — even among the studios themselves — to download and train AI on unlicensed copyrighted content.
Beyond internal training practices, Midjourney is also requesting that the studios reveal all the prompts they used within Midjourney’s platform, as well as the resulting outputs, not merely the prompts that produced the allegedly infringing images. This broader request is designed to uncover patterns of use that could undercut the studios’ claims of harm or demonstrate that the studios themselves have benefited from the very technology they now seek to restrict.
The Studios’ Counterargument: A Fishing Expedition
David Singer, the lead attorney representing the studios, has characterized Midjourney’s requests as a fishing expedition — a broad, speculative search for evidence that the startup hopes will bolster its defense but lacks a specific basis in the facts of the case. Singer has also stated publicly that the studios do not seek to stop AI technology or even to shut down Midjourney’s business. Rather, he has said, they simply want Midjourney to stop copying their movies and TV shows and to stop distributing, publicly displaying, publicly performing, and creating derivative works that include copies of their famous characters without authorization.
This framing positions the studios as defenders of intellectual property rights who are not opposed to AI innovation per se, but who object specifically to the unauthorized use of their proprietary content for commercial AI training. The question of whether the studios themselves engage in similar practices, however, could complicate that narrative considerably. If discovery reveals that the studios have trained internal AI models on unlicensed third-party content, or that they have used Midjourney itself for commercial purposes while simultaneously suing the company, their legal and moral standing would be significantly weakened.
What Is the Core Legal Question in the Midjourney Studio Lawsuits?
The core legal question is whether training an AI model on copyrighted images without a license constitutes copyright infringement or falls under fair use protections. Midjourney argues that its use of publicly available images for training is transformative and therefore permissible under fair use, while the studios argue that the reproduction of their characters in AI-generated outputs directly harms their intellectual property rights and market value. The discovery dispute adds a secondary question: whether the studios’ own AI practices are relevant to determining industry standards and fair use.
Industry Implications: Double Standards and Strategic Leverage
This case highlights a broader tension in the entertainment and technology industries. Hollywood studios have been among the most vocal critics of generative AI, yet many of them are simultaneously exploring or deploying AI tools for internal production workflows, visual effects, and even content generation. The line between using AI as a production tool and using it to replicate copyrighted material is not always clear, and internal practices may not always align with public legal positions.
If Midjourney succeeds in compelling broader discovery, the resulting documents could provide an unprecedented window into how major studios actually use AI behind the scenes. Such disclosures could influence not only the outcome of this specific case but also the broader regulatory and legislative conversation around AI training data and fair use. Legislators and policymakers, who are already grappling with how to regulate generative AI, would gain access to real-world evidence of industry practices rather than relying solely on theoretical arguments from either side.
For Midjourney, the discovery motion is also a strategic lever to shift the narrative. By painting the studios as hypocrites who condemn practices they themselves engage in, the startup hopes to weaken the moral force of the plaintiffs’ arguments and potentially sway public opinion and judicial perception. Whether this strategy succeeds depends on what the discovery process actually uncovers — and on whether the judge agrees to expand the scope of documentation that must be produced.
Technical Dimensions: Training Data and Fair Use Doctrine
Understanding the technical underpinnings of this dispute is essential. Generative AI models like Midjourney are trained on vast datasets comprising billions of images scraped from the internet. The training process involves analyzing patterns in those images and learning to generate new images that mimic the styles, subjects, and compositions found in the training data. When a user prompts Midjourney to generate an image of a character like Bart Simpson, the model draws on its training data, which likely includes numerous images of that character from publicly available sources.
Fair use doctrine, as codified in U.S. copyright law, considers four factors: the purpose and character of the use, the nature of the copyrighted work, the amount and substantiality of the portion used, and the effect of the use on the potential market for the original work. Midjourney’s argument hinges on the first factor — that training is a transformative, non-expressive use — and on the fourth factor, where the company seeks to show that its models do not harm the market for the original works. The studios’ internal AI practices could be relevant to both factors, particularly if the studios themselves use similar training methods for purposes that are arguably transformative or that do not harm the market in the ways they claim.
The Road Ahead: What to Watch in This Case
The judge’s ruling on the discovery motion will set the stage for the next phase of this litigation. If the court grants Midjourney’s request to compel broader documentation, the studios will face a choice: comply and potentially expose sensitive internal practices, or resist and risk sanctions or adverse inferences. If the court denies the motion, Midjourney will have to defend itself without the benefit of what it views as critical exculpatory evidence.
Beyond the immediate procedural skirmish, this case is likely to be one of the first major tests of how copyright law applies to generative AI training in the entertainment sector. The outcome could influence not only Midjourney’s business but also the practices of other AI companies and the strategies of content owners seeking to protect their intellectual property in an era of increasingly capable generative models.
The broader question — whether the industry that is most vocal about the dangers of AI is itself quietly adopting the same technologies — will linger regardless of the immediate ruling. The documents Midjourney is seeking, if they ever see the light of day, could provide an answer that no amount of legal argument alone can deliver.
