Midjourney wants to peek inside Hollywood’s AI closet, and studios are not having it

Here’s the thing about throwing stones: it helps if you’re not living in a glass house. That’s essentially the argument Midjourney is making as it tries to force Hollywood’s biggest studios to hand over their own AI training records in an ongoing copyright battle.

The AI image generation company filed a motion on July 3, 2026, seeking expanded discovery that would give it access to internal AI business plans, research reports, datasets, and even model weights belonging to Disney, NBCUniversal, and Warner Bros. Discovery. The logic is straightforward: if these studios are also training AI models on unlicensed copyrighted content, their moral authority to sue Midjourney for doing the same thing gets considerably weaker.

The backstory: how we got here

The lawsuit saga kicked off in June 2025, when Disney and Universal/DreamWorks filed suits accusing Midjourney of using their copyrighted characters without permission to train its generative AI models. Warner Bros. Discovery piled on in September 2025. The allegations center on direct and secondary copyright infringement, the kind of claims that could reshape how the entire AI industry handles training data.

Midjourney isn’t denying the core mechanics of how its models work. Instead, it’s building a defense on two pillars: fair use and a doctrine called “unclean hands.” The unclean hands argument essentially says: you can’t sue me for something you’re doing yourself.

A magistrate judge issued a ruling on June 15, 2026, limiting discovery to consumer-facing AI applications only. In English: Midjourney could look at the studios’ public-facing AI products, but not the internal research, experimental models, or training pipelines that might reveal whether these companies are quietly doing the same thing they’re suing Midjourney for.

Midjourney appealed that ruling between June 29 and July 3, 2026, arguing that restricting discovery to consumer-facing tools misses the point entirely. The company’s legal team contends that the most relevant evidence sits in internal documents: the training datasets, the model architectures, the strategic memos about how to build AI systems using available content.

The studios push back hard

The studios have characterized Midjourney’s broadened discovery requests as a “fishing expedition.” From the studios’ perspective, what they do internally with AI has no bearing on whether Midjourney violated their copyrights.

But Midjourney’s lawyers are arguing something more nuanced. They’re saying that if training AI on unlicensed copyrighted material is an industry custom, a widespread, accepted practice across the entertainment sector, that context matters for the fair use analysis. Evidence that the plaintiffs themselves view such training as acceptable could shift those calculations.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.