Midjourney Requests Hollywood Studios Disclose Their Internal Use of AI in Litigation Battle Over Copyright
Table of Contents
You might want to know
1. Could evidence of studios using AI internally affect Midjourney's fair use defenses?
2. What scope of documentation is Midjourney seeking, and why does it matter for the case?
Main Topic
The dispute between AI image-generator Midjourney and three major Hollywood studios centers on whether the studios must disclose details about their own use of generative artificial intelligence during discovery. Midjourney was sued last year by Disney and Universal for alleged copyright infringement, with plaintiffs asserting that Midjourney’s models can produce images replicating studio-owned characters such as Bart Simpson and Darth Vader. Warner Bros. later filed a similar suit. Midjourney contends that training on images of copyrighted characters can be covered by fair use, and it has sought evidence that could support that legal position.
At issue in the current phase of litigation is what documentation the studios must provide to Midjourney. A judge already ordered the studios to produce information about their use of generative AI, but limited that requirement to instances where such AI resulted in "consumer-facing" images or videos. Midjourney argues that this limitation is too narrow. In its most recent court filing, the company urged the court to compel broader disclosure, asserting that the restricted scope allows the studios to selectively produce materials that bolster their market-harm claims while withholding documents that would bolster Midjourney’s defenses.
Midjourney’s filing explains why broader discovery is relevant: if the studios internally use image-generation tools for non-public tasks such as storyboarding, concepting, or ideation, that conduct could show that downloading and training models on copyrighted content is an industry practice—even practiced by the studios themselves. Such evidence, Midjourney argues, would be probative of customary behavior within the industry and might weaken assertions that Midjourney’s actions are uniquely harmful or outside ordinary commercial practices.
Concretely, Midjourney seeks not only production of documents describing studios’ generative-AI use but also the prompts studios used when they interacted with Midjourney and the outputs that resulted. The startup contends that restricting production to prompts tied to allegedly infringing images fails to capture the full scope of relevant materials and could conceal patterns or practices useful to its defenses.
The studios’ lead counsel, David Singer, has characterized Midjourney’s requests as overly broad and akin to a "fishing expedition." He has emphasized that the studios' objective is not to halt AI development or to destroy Midjourney’s business, but rather to stop what they view as unauthorized copying and distribution of their copyrighted films, shows, and famous characters. Singer asserts the studios merely seek injunctive and remedial relief targeted at preventing the reproduction and public dissemination of their protected characters and works without permission.
Legally, the dispute highlights the tension between discovery rules designed to ensure both parties can fairly present their cases and the protections against unbounded, burdensome requests. Midjourney frames the sought materials as central to its fair use defense and to demonstrating industry norms. The studios frame the requests as potentially invasive and not sufficiently tied to the harms they claim. A broader production could reveal internal workflows, training datasets, prompts, and non-public outputs, all of which might bear on liability, market harm, and the reasonableness of each party's behavior.
Beyond the immediate evidentiary fight, the case poses broader questions about how courts should treat use of copyrighted material in training and developing generative models, and whether parity in practice across an industry should influence adjudication. If studios themselves rely on unlicensed copyrighted content internally, that fact could influence a court’s assessment of industry standards and the availability or scope of remedies. Conversely, studios may argue that internal, non-public use differs materially from a public-facing service that distributes derivative works widely.
In sum, Midjourney is pressing for expanded discovery to uncover whether the studios' own AI practices align with or contradict their allegations. The outcome will shape not only the parties’ litigation positions but also how courts evaluate evidence about internal AI practices in determining fair use and appropriate relief.
Key Insights Table
| Aspect | Description |
|---|---|
| Key Fact 1 | Midjourney seeks broader discovery to determine whether studios use AI internally in ways similar to Midjourney’s practices. |
| Key Fact 2 | A prior ruling required disclosure only for "consumer-facing" outputs; Midjourney argues that limitation conceals relevant non-public practices. |
Afterwards...
Looking forward, this litigation highlights the need for clearer norms and possibly regulatory or judicial guidance on how copyrighted material may be used to train generative models. Policymakers, industry participants, and courts will benefit from exploring robust auditability standards for training datasets, transparent documentation of model development practices, and standardized approaches to prompt and output retention policies. Such efforts could reduce disputes about discoverability while promoting responsible AI development and fairer resolution of intellectual property issues.
Additionally, technical advances in provenance, watermarking, and dataset lineage tracking could help demonstrate whether specific outputs derive from particular copyrighted works. Investing in these technologies—and in cross-industry agreements on acceptable internal uses—would offer practical mechanisms to balance innovation with rights protection.
Finally, the case underscores that litigation outcomes will shape industry behavior. Greater transparency about AI training and usage, whether voluntarily adopted or compelled by courts, may become a central element of trust between creators, platforms, and AI developers. Exploring interoperable standards for dataset documentation and consent mechanisms for copyrighted content could meaningfully reduce conflict and provide clearer expectations for all parties.