News
Midjourney Demands Disney, Universal and Warner Bros Reveal Their Own AI Practices
Midjourney has filed a motion with a federal court to lift restrictions on its access to film studios' internal documents about their own AI use. The company is defending itself against a copyright infringement lawsuit by arguing that the plaintiffs themselves train models on unlicensed data.
Midjourney filed a motion this week with Judge John Kronstadt of the federal court in California to overturn a ruling that limited the company's access to internal documents from Disney, Universal and Warner Bros Discovery concerning their own use of artificial intelligence. It's the latest chapter in the high-profile dispute over whether image generators infringe copyrights on characters such as Batman, Superman and Darth Vader.
The dispute dates back to June 2025, when Disney and Universal filed suit alleging that Midjourney's image generator produces unauthorized versions of characters such as Bart Simpson and Darth Vader. In September, Warner Bros Discovery joined the lawsuit with its own set of protected characters, including Batman and Superman. The companies are seeking damages and an injunction against further generation of images based on their intellectual property.
Midjourney's Defense Strategy
Midjourney is relying on a fair use defense and argues that the studios themselves engage in the exact practices they accuse the company of. In June, the magistrate judge limited the scope of disclosed information solely to AI applications aimed directly at consumers, which Midjourney's lawyers say excluded the most important evidence from the proceedings: the studios' internal tools for generating storyboards and film concepts.
If the plaintiffs are doing exactly what they're trying to punish us for, that evidence goes to the heart of our fair use and unclean hands defense - Bobby Ghajar, attorney for Midjourney
What Discovery Would Reveal
In the new motion, Midjourney's lawyers are demanding the studios' AI business plans, research reports, training datasets, model weights and presentations prepared for their boards. They argue that if the studios train their own image-generating models on unlicensed third-party material for internal purposes, that proves scraping and training AI on protected content is standard practice within the film industry itself, including among the very entities now suing Midjourney.
This is a fishing expedition designed to distract from Midjourney's own infringement - David Singer, lead attorney for the studios
Implications for the AI Industry
The case is being closely watched across the generative AI sector, since the outcome could shape how courts evaluate fair use defenses for models trained on protected material. Similar disputes are unfolding in parallel in the music industry, where rulings are expected in summer 2026 in cases against the platforms Suno and Udio, while in Germany a Munich court is set to rule on July 31 in the GEMA v. Suno case.
For Polish creators, publishers and companies that use image generators, the outcome of this proceeding carries indirect but real significance. If American precedent establishes that training models on unlicensed data is common practice even among major media conglomerates, that could weaken publishers' standing in similar disputes playing out in Europe, including in the context of the EU's AI Act and its rules on training-data transparency.
Judge Kronstadt has not yet ruled on the scope of document disclosure. If he grants Midjourney's motion, the studios will have to hand over documents they have so far consistently sought to keep confidential.
Sources: Variety (variety.com), Winbuzzer (winbuzzer.com), Artnews (artnews.com), The AI Insider (theaiinsider.tech)
