Midjourney is asking a federal court to force Disney, Universal, and Warner Bros. to hand over records of their own internal AI usage, arguing the studios suing it for copyright infringement are likely training image models on the same unlicensed material. The startup filed the discovery motion this week in the consolidated copyright case brought by the three studios, whose combined market capitalization exceeds $400 billion.
The strategy is straightforward: if Midjourney can show the studios are quietly downloading and training on copyrighted images for storyboarding or content ideation, its fair-use defense hardens into an industry-custom argument. That is a materially stronger legal posture than defending the practice as a startup outlier.
The dispute traces back to last year, when Disney and Universal sued Midjourney over outputs that reproduced characters including Bart Simpson and Darth Vader. Warner Bros. joined a few months later with its own filing. Midjourney has maintained that training on copyrighted images qualifies as fair use — the same defense OpenAI and Anthropic have raised in their respective copyright suits.
Key facts
- 01Midjourney is seeking to compel Disney, Universal, and Warner Bros. to disclose their internal generative AI usage as part of discovery.
- 02Disney and Universal sued Midjourney last year over characters including Bart Simpson and Darth Vader; Warner Bros. followed months later.
- 03A judge previously limited studio disclosures to AI usage that produced consumer-facing content — Midjourney wants that limit overturned.
- 04Midjourney argues training on copyrighted images qualifies as fair use, an industry-wide practice.
- 05Studios' lead attorney David Singer called Midjourney's demand a 'fishing expedition.'
A judge previously ruled the studios must produce information about their generative AI usage, but only where that usage produced consumer-facing videos or images. Midjourney wants that boundary erased. In its filing, the startup said the limit lets the studios cherry-pick documents supporting their market-harm claims while withholding evidence that would support Midjourney's defense.
The startup went further, claiming the withheld documents are precisely those that would show the studios doing behind closed doors what they are suing Midjourney for doing in public. Midjourney also wants disclosure of every prompt studio employees have entered into its own product, along with the resulting outputs — not just the prompts that produced allegedly infringing images. That reciprocal-discovery framing is unusual and, if granted, would force the studios into a defensive posture on their own AI workflows.
The scope matters because internal-use AI in Hollywood is not hypothetical. Studios have publicly discussed AI for previsualization, storyboarding, script analysis, and VFX iteration. What has not been disclosed is what training data those internal tools rely on, or whether any of it includes competitors' copyrighted material scraped without licensing.
David Singer, lead counsel for the studios, has called Midjourney's demand a fishing expedition. He has framed the studios' position as narrow: they are not trying to stop AI or shut Midjourney down — only to stop the reproduction of their specific characters without authorization.
That framing is meant to insulate the studios' own AI ambitions from the outcome of the case. If the court accepts it, the studios can win a narrow injunction against character replication while continuing to build and license generative tools of their own. If Midjourney's broader discovery request lands, that separation collapses. Whatever the studios' internal AI teams have been doing enters the public record.
The case sits inside a growing docket of copyright litigation that will determine what training data AI companies can legally use. The New York Times case against OpenAI, the Anthropic authors settlement earlier this year, and the various music-industry filings against Suno and Udio all revolve around variations of the same question. Midjourney's twist is procedural — using discovery to argue that the plaintiffs are engaged in the same conduct they are suing over.
The counterweight is that discovery motions of this scope rarely succeed at their maximum ask. Courts tend to narrow rather than expand the scope of what one party can extract from another, and the studios' argument that Midjourney is fishing has procedural weight. Even a partial win — forcing disclosure of studio AI usage beyond consumer-facing output — would still hand Midjourney a materially better factual record than it has today.
For the broader AI market, the significance is that Midjourney is turning a defensive lawsuit into an offensive fact-finding operation against the plaintiffs' own AI practices. If the tactic works, expect every AI company facing a copyright suit from a media or publishing incumbent to run the same play. The plaintiffs in these cases are almost never AI-free themselves, and that asymmetry — between what they claim in court and what their AI teams are doing internally — is now a lever every defendant will try to pull.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
