Midjourney asks court to expose studios' own AI usage

Key Takeaways

• Midjourney filed to expand discovery, demanding studios reveal all internal AI development, not just consumer-facing outputs
• The startup argues studios may be training their own AI models on unlicensed copyrighted content behind closed doors
• A judge previously limited disclosure to consumer-facing AI; Midjourney calls this cherry-picking

Midjourney wants to flip the script on the Hollywood studios suing it for copyright infringement. In a new court filing, the AI image-generation startup is asking a judge to force Disney, Universal, and Warner Bros. to disclose how they use generative AI internally. The argument: if studios are training their own AI on unlicensed copyrighted material, that would undermine their claims against Midjourney.

The case centers on whether Midjourney's training on copyrighted images, which allows users to generate characters like Darth Vader and Bart Simpson, constitutes fair use. Disney and Universal filed suit in 2025, and Warner Bros. followed months later. Midjourney has maintained that its training methods fall within legal bounds.

Why Midjourney wants full AI disclosure

A judge previously ruled that studios must produce documents about their generative AI usage, but only when that usage led to "consumer-facing" videos and images. Midjourney now argues this limitation lets studios hide the most relevant evidence.

“The documents they are withholding are precisely those that would reveal whether, behind closed doors, they are doing exactly what they are suing Midjourney for doing.”

— Midjourney court filing

The startup claims that if studios are building AI models for internal use, such as storyboarding or content ideation, and those models were trained on unlicensed content, that fact would establish an "industry custom." This matters because industry custom can inform how courts interpret fair use. If major studios engage in the same practice they accuse Midjourney of, it becomes harder to argue that practice is inherently harmful.

What Midjourney is demanding

Beyond internal AI development records, Midjourney wants access to every prompt the studios ran through its platform and the resulting outputs. Currently, the studios only need to produce prompts that generated allegedly infringing images. Midjourney argues the full picture matters.

The studios' lead attorney, David Singer, has called this a "fishing expedition." He also clarified the studios' position: they do not seek to shut down Midjourney or halt AI development entirely. Their complaint is narrower. They want Midjourney to stop copying their movies and TV shows, and stop distributing, publicly displaying, or creating derivative works featuring their characters without authorization.

The stakes for both sides

This case is one of dozens testing how copyright law applies to generative AI. The outcome could set precedent for how AI companies train models and how traditional content owners protect their intellectual property.

Midjourney, valued at roughly $5.5 billion with an estimated $200 million in annual revenue, operates with fewer than 100 employees. It faces studios with combined market caps in the hundreds of billions. But the discovery dispute could shift the narrative. If internal documents reveal that Disney or Universal trained proprietary AI tools on third-party copyrighted content, the studios would face uncomfortable questions about their own practices.

For tech decision-makers watching this case, the implications extend beyond entertainment. Any company building or using AI tools trained on external data should pay attention to how courts interpret fair use and what documentation might become discoverable in litigation.

What happens next

The judge will rule on Midjourney's motion to expand discovery. If granted, the studios would need to produce internal AI development documents regardless of whether the resulting outputs reached consumers. That could reveal partnerships with AI vendors, in-house model training data, and internal policies around licensing.

Both sides have reasons to settle before such documents become public. But neither has signaled willingness to back down. The case will likely proceed through 2026, with discovery rulings shaping the battlefield.

Frequently Asked Questions

What are Hollywood studios suing Midjourney for?

Disney, Universal, and Warner Bros. allege that Midjourney infringed their copyrights by training its image-generation models on their characters, including Darth Vader and Bart Simpson, allowing users to generate unauthorized images of those characters.

What is Midjourney's fair use defense?

Midjourney argues that training AI models on publicly available images, including copyrighted ones, is permitted under fair use doctrine. The startup contends this is similar to how human artists learn by studying existing works.

Why does Midjourney want studios to reveal their AI usage?

Midjourney claims that if studios are training their own AI models on unlicensed copyrighted content, that would establish an industry custom that supports the fair use defense and undermine the studios' claims of market harm.

What did the judge previously rule about discovery?

The judge ruled that studios must disclose generative AI usage, but only when it resulted in consumer-facing videos and images. Midjourney is now challenging that limitation as too narrow.

How could this case affect other AI companies?

The ruling on fair use could set precedent for how AI companies can legally train models on external content. It may also establish norms around what documentation becomes discoverable in AI-related copyright litigation.

Source: TechCrunch / Anthony Ha