Disney v. Midjourney, explained.
The first Hollywood-studio lawsuit against a generative-image company. Walked through pleading by pleading — what's alleged, what's denied, and what to watch.
The plaintiffs — and why they all filed together
The complaint names seven plaintiffs: Disney Enterprises, Marvel Characters, MVL Film Finance, Lucasfilm, 20th Century Fox Film Corp, Universal City Studios Productions, and DreamWorks Animation. In studio-litigation terms that is a coalition: NBCUniversal, Disney, and DreamWorks together hold the copyrights for most of the most-recognizable Western fictional characters of the last fifty years.
Filing as a coalition matters for two reasons. It makes the registered-works exhibit list — attached to the 110-page complaint — simultaneously broad and unimpeachable. Mickey Mouse, Iron Man, Spider-Man, Yoda, Darth Vader, Shrek, Po the Panda, and the Minions all appear with registration numbers. And it concentrates the legal strategy: one set of lawyers, one motion calendar, one settlement authority. Defendants negotiating with a coalition of studios cannot pick off individual plaintiffs.
The complaint's core theory
The complaint pleads three causes of action: direct copyright infringement, contributory and vicarious infringement, and DMCA §1202(b) removal of copyright management information. The structure tracks NYT v. OpenAI — alleged infringement at training time, alleged infringement at inference time, and alleged knowledge-based secondary liability — but with a twist that the studios think helps them: the outputs in their exhibit are not loose paraphrases of news articles. They are Mickey Mouse.
The complaint's signature line — Midjourney is "the quintessential copyright free-rider and a bottomless pit of plagiarism" — is doing rhetorical work, but the legal claim under it is doctrinally clean. The studios attach a 50-page output exhibit showing prompt-and-image pairs where users asked Midjourney for "Yoda holding a lightsaber" or "Shrek in Times Square" and got back images that any first-year copyright student would call substantially similar to the registered works.
Why this case is structurally different from the artist class actions
Generative-image litigation began with Andersen v. Stability AI — individual artists suing image-generator companies as a class. Disney is not an artist class action. Three differences matter:
- The plaintiffs are not asking for class certification. They are seven corporate copyright owners with named, registered works. There is no Rule 23 hurdle, no commonality fight, no notice-and-claims process. The case proceeds on the merits straight through.
- The exhibit is famous. Substantial similarity is a fact question, but it gets a lot easier when the comparison is to Iron Man or Mickey Mouse rather than to a relatively unknown illustrator's portfolio.
- The damages math is bigger. The studios are asking for statutory damages per registered work. With hundreds of registered works in the exhibit and willful infringement on the table, the theoretical exposure runs into the billions before any actual-damages add-on.
Midjourney's answer — the August 2025 response
Midjourney filed its answer on August 6, 2025. The 43-page document does three things. It denies the core infringement allegations. It raises fair use as the first affirmative defense. And it pleads "unclean hands" — pointing out that Disney itself uses generative AI tools internally, including, per the answer, Midjourney.
The fair-use defense is the one to watch. Midjourney is not the first generative-AI defendant to plead fair use, but it has a harder version of the argument than text-model defendants because the copyrightable expression at issue (a character's appearance) tends to be exactly what the model is reproducing. The Anthropic and Meta cases turned on whether ingesting books to train a language model is transformative. The Midjourney case will turn on whether reproducing Iron Man's likeness in response to a user prompt that says "Iron Man" is transformative. That is a much steeper hill.
The unclean-hands argument is a sideshow that gets the press but probably not the verdict. Equitable defenses don't typically defeat statutory copyright claims, and even if Disney's internal AI use is admitted, that does not change whether Midjourney's outputs infringe.
What the schedule looks like
The current scheduling order sets the following milestones:
- Sep 21, 2026
- Non-expert discovery closes
- Nov 9, 2026
- Expert discovery closes
- Nov 23, 2026
- Motions for summary judgment due
- 2027
- Earliest realistic trial window
Discovery is where the case actually gets fought. Disney has already signaled it will seek the full training data inventory — what was scraped, when, with what filtering, and whether any takedown notices were honored. Midjourney will resist on trade-secret and proportionality grounds. Expect at least one major discovery dispute heard by Magistrate Richlin before the September 2026 deadline.
What to watch
- Will fair use survive a motion to dismiss? If Midjourney files an early Rule 12 motion on fair use and loses (or doesn't file at all), that signals Judge Kronstadt has read the exhibit and is unimpressed by the transformativeness argument as pleaded.
- The training-data discovery fight. The first ruling on the scope of Disney's training-data discovery will set a precedent for every subsequent generative-image case. Watch Magistrate Richlin's docket.
- Settlement signal. Midjourney reportedly generated $300M of revenue in 2024. The studios are not asking for "cost of license" damages. A settlement that includes both money and prospective output filters would be the most likely structural outcome — but well after summary judgment briefing is done.
- Knock-on cases. Other studios (Warner Bros., Sony) have not joined. If the case clears summary judgment with infringement intact, expect copycat filings against every image-generator on the market within months.
How this fits into the broader AI-litigation landscape
Disney v. Midjourney sits inside the same fact pattern as Bartz v. Anthropic — alleged ingestion of copyrighted material at training time, with downstream output that allegedly reproduces it. But the registered, famous, character-shaped nature of the outputs makes this a much cleaner test of the underlying legal question: does training and serving a model that reliably regurgitates copyrighted expression infringe?
If the answer is yes, every image-generation product on the market needs an output-filter audit. If the answer is no — if fair use holds at summary judgment — the studios will appeal to the Ninth Circuit, which in 2027 or 2028 will give us the closest thing to a definitive answer.
Frequently asked questions
What is Disney v. Midjourney about?
Hollywood studios — Disney, Marvel, Lucasfilm, 20th Century Fox, Universal, and DreamWorks — sued Midjourney on June 11, 2025 for copyright infringement and DMCA §1202 violations over AI-generated images of their registered characters. Case 2:25-cv-05275 (C.D. Cal.).
Who is the judge in Disney v. Midjourney?
U.S. District Judge John A. Kronstadt presides; Magistrate Judge Joel Richlin handles discovery disputes.
What is Midjourney's defense?
Fair use as the first affirmative defense, plus an "unclean hands" argument noting that Disney itself uses generative AI tools, including Midjourney, in its own production work.
When will Disney v. Midjourney go to trial?
No trial date yet. Discovery closes in September 2026, summary-judgment motions are due November 23, 2026. A jury trial is realistic in 2027 if the case is not resolved on summary judgment.
How does this differ from Andersen v. Stability AI?
Andersen is an artist class action with individual creators as plaintiffs. Disney is the first studio-led case — corporate copyright owners with registered, famous works and a unified legal team. The economic stakes are much larger and a quick settlement much less likely.
Where to go next
- Midjourney case tracker — every case naming the company
- How many AI copyright lawsuits are there — the live count
- Which AI companies have been sued — the directory
- Fair use and AI training: what courts have actually ruled
- What the Anthropic $1.5B settlement really means