Midjourney Pushes Hollywood to Disclose Its AI Playbook in Copyright Fight

Midjourney’s legal battle with three of Hollywood’s biggest studios has entered a new phase: the image-generation startup is now asking a federal court to force Disney, Universal and Warner Bros. Discovery to disclose how they use artificial intelligence behind the scenes.

The request adds a sharp twist to a dispute that already sits at the center of one of the entertainment industry’s most consequential questions: whether companies that complain about AI copying their characters are, at the same time, using similar tools internally to speed up development. Midjourney says that possibility matters because it could help show that AI-assisted image creation has become a normal business practice in film and television, not an aberration invented by a rogue startup.

The studios, for their part, have accused Midjourney of building and operating image models that can produce copyrighted characters and artwork without permission. Midjourney has responded that training on copyrighted material is protected by fair use, and that the studios are trying to narrow discovery in a way that shields information useful to the startup’s defense.

The fight is no longer just about whether Midjourney’s outputs resemble Bart Simpson or Darth Vader. It is increasingly about what kinds of internal AI use the studios themselves can be compelled to reveal, how far discovery should reach, and whether Hollywood can challenge generative AI from one side of the courtroom while potentially deploying similar tools from the other.

The latest court fight is about discovery, not just infringement

At the center of the current dispute is the document-production phase of the case, known as discovery. That stage is where both sides exchange records, communications and technical material that may support or weaken their claims. Midjourney wants the court to expand what the studios must turn over.

According to the startup’s latest filing, a previous judicial ruling already required the studios to disclose information about their own generative AI use, but only when that use involved consumer-facing images or videos. Midjourney is now challenging that limitation and arguing that it creates an unfair evidentiary imbalance.

The company says the restriction allows the studios to select only the material that helps their argument that Midjourney harms their markets. In Midjourney’s view, that leaves out records that could support its own defense: that AI-generated content has become embedded across the entertainment industry, including within the studios suing it.

In practical terms, Midjourney wants access to internal materials that could show whether the studios have adopted AI tools for storyboarding, concept development, visualization or other early-stage creative work. If those tools were trained on copyrighted material without licensing, the startup argues, that could undercut the studios’ claim that Midjourney’s approach is uniquely harmful or outside industry norms.

Why Hollywood’s internal AI use matters to Midjourney’s defense

Midjourney’s argument is not just rhetorical. It goes directly to a familiar issue in copyright litigation: industry custom.

By seeking internal records, Midjourney appears to be trying to show that major entertainment companies themselves rely on unlicensed or semi-licensed training practices when they experiment with generative tools. If a studio uses image-generation systems for internal ideation, concept art or preproduction, the startup argues, that could help demonstrate that the conduct is part of a broader business practice rather than a violation unique to Midjourney.

The company has suggested that if the studios are building or testing tools for internal storyboarding and content ideation, that would show they are engaged in the same general category of activity they now condemn in a competitor. Midjourney’s logic is that such evidence could matter both for fair use analysis and for assessing the studios’ claims of market harm.

That is especially relevant in a case involving creative workflows, where the line between inspiration, reference, transformation and direct copying is often legally contested. Hollywood studios have spent years experimenting with machine learning for editing, dubbing, visual effects, localization, audience analytics and preproduction. Midjourney wants a fuller record of that experimentation, not just the parts that support the studios’ complaint.

The startup’s theory of the case

Midjourney’s broader theory is that AI-generated art should be treated as part of a changing media workflow, not as an isolated infringement machine. If studio teams are already using generative systems to brainstorm scenes, test concepts or create rough visuals, then the startup argues that the market for such tools is not hypothetical.

That matters because copyright plaintiffs often try to show that a defendant’s use substitutes for or harms a protected market. Midjourney wants to argue the opposite: that the studios may be using the same kinds of tools for similar purposes, which would suggest a market in transition rather than one being destroyed by a single company’s products.

The startup also appears to be laying the groundwork for a potentially important public-interest narrative. If Hollywood uses generative AI internally while suing a startup for making generative AI available to the public, Midjourney can frame the case as an attempt by powerful incumbents to regulate competitors while keeping productive uses for themselves.

Studios accused Midjourney of making unauthorized copies of famous characters

The lawsuit itself began with a familiar complaint from copyright holders facing generative AI systems: that the technology can produce highly recognizable characters and imagery without permission.

Disney and Universal sued Midjourney last year, alleging that the startup’s image models were capable of generating characters associated with their franchises, including figures such as Bart Simpson and Darth Vader. Warner Bros. Discovery later brought its own case.

From the studios’ perspective, the issue is not whether artificial intelligence is useful. It is whether a commercial product can be trained and deployed in ways that reproduce or approximate copyrighted characters, style elements and other protected creative assets without a license.

They argue that the startup is effectively building a business on the back of works and characters that took decades and billions of dollars to develop. The studios have also signaled that they see the case as part of a broader effort to protect creative industries from unlicensed AI exploitation.

Midjourney rejects that framing. It contends that model training on copyrighted images qualifies as fair use under U.S. law and that generative systems necessarily learn from large-scale collections of visual data, much like human artists and designers do when they absorb prior art and popular imagery.

What the studios say they want

According to the studios’ lawyer, the goal is not to destroy Midjourney or ban AI outright. Instead, the companies say they want the startup to stop using and distributing copies of their movies and television properties, and to stop publicly displaying or performing works that include their characters without authorization.

Representing the studios, attorney David Singer has characterized Midjourney’s requests for internal records as overly broad, arguing that the company is effectively probing for material beyond what is necessary to litigate the case.

Singer has also said the startup is embarking on a fishing expedition, suggesting that the discovery demands are designed less to obtain relevant evidence than to search for any information that might be used to complicate the plaintiffs’ claims.

That framing matters because discovery disputes can shape the direction of a lawsuit long before any final ruling on liability. If a court allows Midjourney broader access, the startup could gain leverage and potentially uncover material useful in other AI copyright fights as well.

The key issue: how broad should discovery be?

In copyright cases, discovery is often where strategic battles are won or lost. A narrow discovery order can deprive a defendant of evidence needed to argue fair use, market substitution or industry practice. A broad one can force plaintiffs to reveal internal conduct they may have preferred to keep private.

The prior court ruling reportedly allowed information about generative AI use, but only when it resulted in content meant for consumers. Midjourney is now pushing to widen that standard. The company says consumer-facing limitations are too restrictive because they exclude internal experimentation that may still be legally significant.

If the studios are using AI to generate storyboards, scene layouts, concept art or pitch materials, those uses may never reach the public directly. But in Midjourney’s view, that does not make them irrelevant. They could still show that the studios themselves view AI-generated imagery as a practical creative tool.

That is especially important because courts often examine real-world market behavior when deciding whether new technology harms existing rights holders. Internal use may not equal public distribution, but it can still reveal whether a market exists, whether licensing is customary and whether similar practices are already taking place within the industry.

Why this dispute could shape future AI copyright litigation

The Midjourney case is one of several high-profile disputes testing how copyright law applies to generative AI. The outcome may not settle the field, but it could influence how other courts think about discovery, fair use and the relevance of internal corporate AI experiments.

There are at least three reasons the case matters beyond the parties involved.

• It could define the scope of discovery in AI cases. If courts accept that internal AI use is relevant even when it is not consumer-facing, future plaintiffs may have to disclose far more about their own experimentation.
• It could affect how fair use is argued. A broader record may strengthen defendants who want to show that their conduct fits into an existing industry pattern or a transformative workflow.
• It could influence settlement pressure. The more evidence each side must produce, the harder it becomes to manage risk quietly.

In the short term, the ruling may simply decide what documents are shared and what questions are answered. In the long term, it could contribute to the legal architecture around how creative industries handle generative AI.

Timeline of the Midjourney-Hollywood dispute

The lawsuit has moved quickly by copyright standards. The following table summarizes the major milestones referenced in the current dispute.

What Midjourney says the studios are hiding

Midjourney’s filing suggests that the most valuable evidence may be inside the studios’ own creative operations. The startup argues that the studios are withholding records that would show whether they are, in effect, doing internally what they accuse Midjourney of doing commercially.

The company says those materials could include documents showing use of generative AI for developing film or television ideas, generating visual references, or assisting with preproduction workflows. It is also asking for the prompts the studios used in Midjourney, along with the outputs those prompts produced, not just the examples that allegedly yielded infringing imagery.

That demand reflects a common evidentiary problem in AI disputes. Plaintiffs typically highlight the worst examples: the outputs that most clearly resemble copyrighted material. Defendants, by contrast, often want the full context, including prompts that did not produce problematic results, because those may show how the system behaves in ordinary use.

Midjourney’s position is that partial disclosure paints a misleading picture of what users actually do with the tool. If the studios are relying on a small set of outputs to argue harm, the startup wants the court to see the broader interaction history that might show experimentation, rejection, iteration and non-infringing use.

Prompts, outputs and the problem of selective evidence

Prompt-and-output records are particularly important in generative AI cases because they reveal how a model responds to user instructions. A single prompt can produce many different results depending on phrasing, parameters and surrounding context.

If only the most troubling output is presented, a court may not see the degree of variation, the user’s intent or the extent to which a system can be steered away from protected material. Midjourney’s attempt to obtain broader prompt records appears aimed at showing that isolated examples may not tell the whole story.

In practical terms, the company seems to be asking for a more complete map of how the studios interacted with the platform, including prompts that were not alleged to be infringing. That could reveal whether the technology was used for routine creative exploration rather than copying specific characters or scenes.

Hollywood has already embraced AI in some parts of production

Even as these lawsuits unfold, Hollywood has been grappling with AI in a far less abstract way. Major studios, streamers and production houses have already begun testing machine-learning tools in visual effects, dubbing, voice work, script development, marketing and localization.

That reality gives Midjourney’s argument extra force. If entertainment companies are increasingly using generative systems to accelerate ideation and reduce production costs, then the legal debate is not simply about whether AI should exist. It is about who can control it, under what conditions, and whose creative labor gets protected or commodified in the process.

There is also a labor dimension. Writers, actors, designers and visual artists have all voiced concern that generative systems can blur the line between assistance and substitution. The entertainment industry’s public stance on AI has often been cautious or critical, even as individual business units quietly experiment with the technology.

That tension is exactly what Midjourney is trying to surface. If the studios are resistant in public but pragmatic in private, the startup’s defense could gain credibility not only in court but in the broader cultural debate over how creative work is made.

Why the lawsuit has drawn attention beyond the courtroom

This case has attracted interest because it sits at the intersection of three major forces: copyright law, generative AI and one of the most protective intellectual-property sectors in the world.

Hollywood has long fought hard to control the use of its characters and brands. AI companies, meanwhile, argue that modern systems cannot function without large datasets and that training on copyrighted works is part of how these models learn. When those positions collide, the result is a dispute that can influence both policy and product design.

The entertainment industry is especially sensitive because its value depends on identifiable intellectual property. Characters like Darth Vader or Bart Simpson are not merely visual designs; they are commercial ecosystems with decades of marketing, merchandising and audience attachment behind them. That makes the alleged reproduction of those characters in AI output a particularly potent example for rights holders.

At the same time, AI companies argue that if they must obtain permission for every work that might appear in training data, the cost and complexity could make generative systems impossible to build at scale. The fair use question therefore has enormous implications for startups, research labs and large platforms alike.

What happens next

The immediate question is whether the court will allow Midjourney to expand its discovery request. If it does, the studios may have to disclose more about how they are using AI internally and how broadly their employees have tested generative tools.

If the court rejects the request, Midjourney may be left with a narrower evidentiary record and a harder path to proving industry custom or narrowing the studios’ market-harm claims. Either way, the ruling is likely to shape the next phase of the case and potentially influence similar lawsuits elsewhere.

For now, the dispute highlights a growing reality in AI litigation: the companies suing over machine-generated content may themselves be experimenting with machine-generated content. Whether that fact helps or hurts them legally will depend on what the court decides about relevance, fairness and the proper scope of discovery.

The broader industry will be watching closely. The answer could determine not only how Hollywood handles its own AI adoption, but also how much of that adoption it must eventually reveal when it sues a generator built on the same technological wave.

Midjourney’s filing essentially argues that the studios should not be allowed to contest AI use in public while keeping their own internal experimentation out of view.

That is the battle now taking shape: not just over who copied whom, but over who must disclose what, and whether the future of creative AI will be litigated in full daylight.

Key facts at a glance

• Parties: Midjourney vs. Disney, Universal and Warner Bros. Discovery
• Main allegation: The studios say Midjourney can generate copyrighted characters without authorization
• Midjourney’s defense: Training on copyrighted material is fair use
• Current dispute: How much the studios must reveal about their own generative AI use during discovery
• Core question: Whether internal AI use should be treated as relevant evidence even if it does not create consumer-facing content

As the case moves forward, the court’s handling of discovery may prove just as significant as any eventual ruling on infringement. For both Hollywood and AI developers, the real contest may be over transparency itself.

superintelligencenews@gmail.com