Midjourney is trying to widen discovery in its copyright fight with major Hollywood studios, arguing that Disney, Universal, and Warner Bros. should have to disclose more about how they use generative AI themselves. The move, reported by TechCrunch from a recent court filing, could matter beyond this case because it goes to a central question in AI litigation: whether entertainment companies can claim market harm from AI systems while also building similar tools internally.

The immediate dispute is procedural, not a final ruling on copyright liability. But discovery fights often shape the factual record that later drives settlement pressure, trial strategy, and public narratives. In this case, Midjourney says the studios should not be allowed to limit document production to AI use that resulted in “consumer-facing” images and videos. The company wants broader disclosure, including internal creative uses that it believes could support its fair use defense and challenge the studios’ claims about competitive harm.

According to TechCrunch, Disney and Universal sued Midjourney last year over alleged copyright infringement, citing the image generator’s ability to produce pictures resembling well-known characters such as Bart Simpson and Darth Vader. Warner Bros. filed a similar suit months later. Midjourney has argued that training on copyrighted images is protected by fair use.

The discovery fight is about more than prompts and outputs

The current flashpoint comes after a judge reportedly ruled that the studios would have to turn over some information about their generative AI activity, but only where that activity led to consumer-facing outputs. Midjourney is now seeking to remove that limitation.

As described by TechCrunch, Midjourney argues that the narrower scope lets the studios selectively produce only the material that helps their case. In the company’s view, broader internal records could show that the same rights holders suing over AI-generated images may also be experimenting with AI for activities like storyboarding, concept development, or ideation.

That matters because copyright cases involving AI are not only about whether a model was trained on protected works. They are also about alleged market substitution and commercial harm. If a plaintiff argues that an AI system damages licensing markets or displaces human-created work, evidence that the plaintiff uses similar systems internally could complicate that theory, even if it does not resolve the legal question on its own.

Midjourney is also reportedly asking for all prompts the studios used in Midjourney, plus the outputs those prompts generated, not just the prompts that produced the allegedly infringing results highlighted in the complaints. That request suggests the company wants a fuller picture of how the studios tested Midjourney, what they were able to generate, and whether their own use patterns undercut parts of their claims.

Why Midjourney wants internal AI workflows on the record

Midjourney’s filing, as quoted by TechCrunch, frames the withheld documents as potentially revealing whether the studios are doing “exactly what they are suing Midjourney for doing.” That is advocacy language from a litigant, not an established fact. Still, it points to a broader defense strategy now common in AI copyright disputes: try to show that incumbents object publicly to model training practices that they may tolerate, explore, or replicate privately.

The company’s example, again cited by TechCrunch, is internal model development for storyboarding or content ideation. If studios are building image-generation workflows for those purposes, Midjourney appears to believe that could help it argue that such practices are emerging industry custom rather than clear-cut infringement. Whether a court ultimately accepts that reasoning is another matter. Internal adoption of generative AI would not automatically prove that any studio trained on unlicensed copyrighted works, nor would it automatically validate Midjourney’s own training methods.

But from a litigation standpoint, the ask is strategic. Discovery into internal AI use could surface documents about experimentation, procurement, policy, vendor relationships, or model training assumptions. Even if those records do not transform the legal standard, they could shape how a judge or jury views the practical realities of AI adoption in Hollywood.

For AI companies watching the case, this is a reminder that copyright litigation is increasingly turning into a contest over workflow evidence. Questions about training data remain central, but courts are also being asked to weigh what kinds of internal enterprise use are considered normal, risky, competitive, or harmful.

The studios say the case is about copying, not stopping AI

The studios have pushed back on Midjourney’s broader discovery demands. TechCrunch reports that lead attorney David Singer previously described the company’s request as a “fishing expedition.” Singer also said the studios are not trying to stop AI technology or shut down Midjourney’s business, but instead want the company to stop copying films and TV shows and distributing or creating works containing unauthorized versions of famous characters.

That distinction is important. Rights holders across media have generally tried to avoid looking anti-technology in public court filings and press statements. Their preferred position has been narrower: generative AI may continue, but not in ways that reproduce protected characters, styles, or copyrighted works without permission.

For Disney, Universal, and Warner Bros., the narrower framing also helps maintain a line between internal experimentation and external commercialization. A studio might argue that even if it uses generative AI in controlled internal contexts, that does not excuse an outside model provider from training on copyrighted material or distributing outputs that resemble protected characters at scale.

This is likely where the fight will continue. Midjourney appears to want evidence that blurs the studios’ distinction between internal use and external infringement. The studios, by contrast, appear to be trying to confine the case to alleged copying and output distribution.

Evidence, claims, and what is still unknown

The reporting in this story comes from TechCrunch’s account of court filings and prior statements by the parties. The underlying event is not a product launch or public benchmark release but a procedural step in an ongoing lawsuit. That means several important points remain unverified or unresolved.

First, Midjourney’s assertion that broader discovery would reveal comparable studio behavior is a claim by a defendant seeking litigation advantage. The filing, as summarized by TechCrunch, suggests the possibility that studios may be training or using generative systems internally, but the available reporting does not establish the scope, methods, data sources, or legality of any such work.

Second, the judge’s prior ruling, as described by TechCrunch, appears to require some disclosure tied to consumer-facing outputs. Without the full court order in the source material, the precise boundaries of that obligation are not fully clear from the reporting alone.

Third, the underlying copyright claims against Midjourney remain contested. The studios allege infringement tied to character-based outputs, while Midjourney says its training practices fall under fair use. No final ruling on that core issue is reported here.

In short, the strongest claims in the current dispute are litigation claims, not established facts. Builders and buyers should read them as part of an adversarial process rather than as confirmed evidence about how Disney, Universal, Warner Bros., or Midjourney actually trained or deployed internal systems.

What this means for enterprise AI and creative tooling

For teams building creative AI products, the case underscores that discovery risk now extends beyond datasets. If a company sells image generation into media, advertising, design, or preproduction workflows, internal customer usage can become legally relevant when plaintiffs argue market harm or substitution.

That has practical implications for enterprise AI vendors. Product teams may need clearer logging, retention, and governance around prompts, outputs, and model provenance. They may also need more explicit customer contracts covering training data, acceptable use, indemnity, and audit access. The request for complete prompt histories in Midjourney is a concrete signal: prompt records are not just product telemetry; in some contexts, they may become litigation evidence.

For studios and other large enterprises, the dispute is a warning that internal experimentation may not stay private when a company also brings claims against outside vendors. The more organizations adopt generative AI for ideation, storyboarding, or concept work, the harder it may be to keep a clean separation between “we oppose this market” and “we use similar tools in-house.”

The broader enterprise AI market should also note the competitive angle. If major media companies are building or testing internal generative systems, that supports the view that incumbents want both legal protection and strategic optionality. They may challenge external platforms like Midjourney while still investing in proprietary creative pipelines.

What to watch next

The next concrete signal is whether the court grants Midjourney’s request to broaden discovery beyond consumer-facing outputs. A ruling in its favor could expose more internal AI documentation from Disney, Universal, and Warner Bros. and potentially influence how other AI copyright defendants frame similar requests.

Another key point to watch is whether any disclosed materials, if produced, speak only to tool usage or go further into model development and training practices. Those are very different categories. Internal use of generative software does not necessarily mean a studio trained models on unlicensed data.

It will also be worth watching whether this dispute changes settlement posture. Discovery battles can raise litigation costs and reputational risks on both sides. If either party believes broader disclosure could be damaging, that can affect leverage.

Finally, this case may become a reference point for other disputes involving AI image generation, famous characters, and fair use. Even without a final merits ruling, discovery decisions can influence how future plaintiffs and defendants structure their arguments.

Creati.ai perspective

The most important part of this story is not that Midjourney and Hollywood disagree on copyright; that was already clear. What stands out is the fight over internal AI usage as evidence. Courts are being asked to evaluate not just what models output, but how entire industries are adopting them behind the scenes.

For builders, that means compliance cannot stop at model training narratives. Governance around prompts, internal pilots, customer workflows, and provenance records is becoming part of the competitive stack. For enterprises, especially in media, the lesson is equally sharp: if you plan to challenge third-party AI vendors in court while adopting generative AI internally, assume that your own usage patterns may become part of the case.