- Discovery Fight: Midjourney is asking U.S. District Judge John Kronstadt to broaden discovery into studio artificial intelligence use.
- Evidence Scope: The request covers business plans, training datasets, model weights, board presentations, prompts and outputs.
- Studio Response: The studios say the lawsuit targets alleged copying and unauthorized character outputs, not AI technology itself.
- Legal Caveat: The discovery dispute does not decide liability, but it could shape what enters the defense record.
Midjourney has filed a motion asking U.S. District Judge John Kronstadt to widen discovery into studio AI use, pushing Warner Bros. Discovery, Disney and Universal to disclose more about their own AI practices in an ongoing copyright case over AI-generated characters. Its request turns the ongoing studios’ infringement suit into a procedural fight over how much behind-the-scenes AI planning a plaintiff must reveal when it sues an AI developer.
A June 15 discovery order limited production to consumer-facing AI applications and rejected broader material as irrelevant to Midjourney’s infringement liability. Discovery, the pretrial exchange of evidence, is now the battleground because Kronstadt’s review could decide whether internal AI planning, datasets and board-level materials enter the case.
Studios have already agreed to provide consumer-facing AI application material, but Midjourney wants the line moved inward. Consumer tools may reveal only public-facing experiments, while internal plans and board presentations could show how studios evaluate AI for production work.
Studio counsel David Singer argued that plaintiffs do not seek to stop AI technology or shut down Midjourney’s business. They want Midjourney to stop copying their movies and TV shows and stop creating derivative works that include copies of famous characters without authorization.
Studios Face Pressure Over Their Own AI Use
Midjourney’s request covers AI business plans and research reports, training datasets, model weights and board presentations tied to how the studios use AI tools for film and television work. Model weights are the parameters that help define how an AI model behaves, so the request reaches beyond public chatbots or consumer-facing products and into technical systems the studios may treat as internal.
Midjourney also wants prompts and resulting outputs from the studios’ use of its platform, rather than only prompts that produced images the plaintiffs claim are infringing. If Kronstadt grants that request, the studios’ own Midjourney interactions would become part of the same evidence fight as their internal AI planning.
A broader order would reduce the studios’ control over which market-harm materials enter the record while withholding defense-relevant documents becomes harder. In practical terms, Midjourney wants material that could weaken the studios’ market-harm theory or support defenses tied to their own AI conduct.
Midjourney also argues that training AI on publicly available images can be fair use. Midjourney attorney Bobby Ghajar contends that evidence of the studios’ own AI training practices could support fair-use and unclean-hands defenses. Unclean hands is a defense arguing that a plaintiff’s related conduct can limit the relief it can obtain.
What the Court Fight Does Not Decide
The fair-use lane separates training on copyrighted works as a claimed transformative use from outputs that closely resemble protected characters. Studio plaintiffs frame the lawsuit around alleged copying and unauthorized depictions of famous characters. Kronstadt’s review will test how far defendants can probe a plaintiff’s AI development before liability is resolved.
Because liability remains unresolved, arguments over whether studios can challenge AI while using it center on evidence scope. Discovery does not decide infringement or whether any studio engaged in wrongful conduct. It can still shape the case by determining whether internal AI development, safeguards, market-harm documents and requested relief remain inside the evidence record.
Before the case reaches liability questions, Kronstadt must weigh targeted discovery tied to market harm and safeguards against broader requests about internal AI development and use. His review turns less on whether AI training is lawful than on which categories of studio material are relevant enough for Midjourney to inspect.
AI developers beyond Midjourney have reason to watch that boundary. If courts let defendants examine a plaintiff’s internal AI practices, copyright suits over generative systems may involve more reciprocal discovery. Studios could also preserve a boundary around internal experimentation unrelated to specific accused outputs.
The Lawsuit Background and Next Step
The current motion sits inside a broader copyright lawsuit over images resembling studio-owned characters. Character-output allegations are the reason the studios are in court, but the immediate issue is what information changes hands before the court weighs liability.
Disney and Universal filed an infringement suit against the AI company. in 2025. Warner Bros. filed a related suit involving Superman, Batman and other characters.
An earlier artists’ AI-training lawsuit from 2023 shows how the fair-use argument remains central to image-generator cases since then.
Kronstadt’s next ruling will decide whether the discovery limit keeps many internal AI materials outside the case or adds them to the defense record Midjourney wants to build.
