Three of the biggest studios in entertainment have been suing Midjourney since mid-2025. Disney, Universal, and Warner Bros. all filed separate cases accusing the AI image tool of copyright infringement, then consolidated them into a single fight in a California federal court. The core accusation is straightforward: type "Superman" or "Shrek" into Midjourney, and you get a highly recognizable, downloadable image of a character those studios own. The studios say that is theft.
Midjourney's defense has two parts. The first is "fair use," a legal principle that allows limited use of copyrighted material under certain conditions. AI companies across the industry have leaned on this argument, with mixed results in courts so far. The second part is sharper: an "unclean hands" argument, which says you cannot sue someone for doing what you also do.
To make that second argument work, Midjourney needs evidence. So it asked the court to compel the studios to hand over their internal AI training data, research reports, model files, and the presentations their own boards received about AI strategy. The logic is that if Disney or Universal are training AI on unlicensed copyrighted content to generate storyboards, concept art, or production assets internally, then the exact behavior they are attacking Midjourney for is also standard practice inside their own walls.
A magistrate judge sided mostly with the studios in June 2026, ruling that they only need to disclose information about AI tools they offer to the public, not internal creative systems. Midjourney is now asking the federal district judge overseeing the case to reverse that decision.
This is not a minor procedural squabble. The studios are in an awkward position. Disney has a dedicated internal AI office with over 100 people. Warner Bros. and Universal are both actively exploring or deploying AI in production workflows. Lionsgate signed a deal in 2024 to train a custom AI model on its 20,000-title library. The studios are not bystanders to AI: they are building it, buying it, and using it. Whether they do so with fully licensed data or not is precisely what Midjourney wants the court to see.
The studios' lead attorney has described Midjourney's strategy as a "fishing expedition" designed to distract from the actual copyright violations. Their position is simpler: they are not trying to ban AI, they just want their specific characters protected. What happens internally at their studios, they argue, is irrelevant to whether Midjourney infringed.
Both positions have merit. But the federal judge's ruling will carry weight beyond this case. Dozens of similar lawsuits are moving through US courts right now, involving music companies, book publishers, news organizations, and visual artists all suing AI companies over training data. If Judge John Kronstadt decides that plaintiffs in copyright cases must disclose their own AI training practices, it changes the calculation for every content owner thinking about suing an AI company. Some may decide the exposure is not worth it.
For businesses outside entertainment, the lesson is practical. Copyright and AI are on a collision course in most industries, not just Hollywood. The rules being written in this case will eventually reach sectors where proprietary data, licensed content, and AI overlap: insurance documents, architectural drawings, product photography, industrial manuals. Anyone who touches both AI tools and protected content has a stake in how this plays out.