Midjourney has filed a motion to compel Disney, Universal, and Warner Bros. to disclose the full scope of their internal generative AI usage, a move that could force the studios suing it for copyright infringement to reveal whether — and how extensively — they rely on the same category of technology they claim caused them harm.

The motion, docket entry 112 in case 2:25-cv-05275 (Disney Enterprises Inc. v. Midjourney Inc., U.S. District Court, Central District of California), was filed on March 28, 2026, according to court records archived via CourtListener. Disney and Universal had jointly brought the underlying copyright infringement suit against Midjourney in June 2025; Warner Bros. filed its own separate action in September 2025. The two cases have since been consolidated for discovery purposes.

The Discovery Dispute

At the center of Midjourney's motion is a prior judicial order that narrowed what the studios were required to produce. Magistrate Judge Joel Richlin's June 2026 discovery ruling limited the studios' obligations to generative AI usage that resulted in consumer-facing images and video. TechCrunch reported on July 4, 2026 that Midjourney's motion explicitly seeks to remove that "consumer-facing" restriction and compel disclosure of internal, non-public AI usage as well — covering workflows that never reached an audience but may have touched copyrighted characters or source material in training or production pipelines.

Midjourney's core defense in the underlying litigation is fair use: the company contends that training image-generation models on copyrighted characters does not constitute infringement. Expanding the discovery scope is legally coherent within that argument. If the studios are themselves training models on proprietary or licensed IP, or using generative tools in internal production workflows involving the same characters, that conduct may bear on the equities of a fair use analysis — particularly the fourth factor, which weighs market harm and substitution. It would also speak to any argument about the normative expectations of the industry.

David Singer, lead counsel for the three studios, characterized Midjourney's document requests as a "fishing expedition," per Variety's earlier reporting. That framing is the standard response to broad third-party or adversarial discovery, but the "consumer-facing" limitation already imposed by Judge Richlin suggests the court had some sympathy for the studios' position — at least on scope. Midjourney's motion now asks him to reconsider.

What Is Actually at Stake

The discovery fight matters beyond procedural housekeeping. If granted in full, the order would compel three of the most powerful media conglomerates to produce internal documentation on AI adoption at a level of granularity they have not publicly disclosed. Studios routinely use machine-learning tools in visual effects, localization, marketing, and post-production — but the specific systems, vendors, and character assets involved in those workflows are not public information. Broad compelled disclosure would be unprecedented in scope.

Worth flagging: the strategic logic here is not merely defensive. A defendant that can put before a jury evidence that the plaintiffs are commercially deploying substantially similar technology — even internally — has a richer narrative for the fair use argument than one arguing the abstract principle alone. Whether that evidence is ultimately admissible for that purpose is a question for the district judge, but getting it into the record through discovery is the prerequisite.

The studios' position is also not without its own tension. Their core claim is that Midjourney's training on copyrighted characters caused economic harm to rights holders. But if internal AI workflows at those same studios process the same characters without per-use licensing or compensation to the underlying rights owners — writers, artists, actors whose likenesses appear in source material — the question of who owes what to whom becomes considerably murkier. Courts have not yet resolved what "fair use" means for AI training at scale, and discovery that surfaces studio practice on both sides of the v. could inform that analysis.

For now, the motion sits with Judge Richlin. His June order drew a line at consumer-facing output; Midjourney is asking him to erase it. The ruling, whenever it comes, will set a meaningful precedent for how broadly courts are willing to let AI copyright defendants probe the internal technology practices of the corporations suing them.