What happened

  • Complaint (June 11, 2025): Disney Enterprises, Marvel Characters, Lucasfilm, Twentieth Century Fox Film and related entities, plus Universal City Studios Productions and DreamWorks Animation, sued Midjourney, Inc. for direct and secondary copyright infringement (Case No. 2:25-cv-05275). The complaint says Midjourney's image service "functions as a virtual vending machine" generating unauthorized copies, and that it is working on a video service. It asks for damages or statutory damages, an accounting, and an injunction requiring "appropriate copyright protection measures."
  • Answer (Aug 6, 2025): Midjourney's first affirmative defense is that any copies made are fair use under 17 U.S.C. § 107. It also pleads lack of ownership, license, estoppel, unclean hands, de minimis copying and DMCA safe harbor.
  • Consolidation: Warner Bros. Entertainment, DC Comics and related plaintiffs sued on September 4, 2025 (No. 2:25-cv-08376). The parties stipulated on November 4, 2025 to consolidate, and later orders are captioned "ALL CASES," consolidated for all purposes.
  • Discovery (2026): The record shows disputes over Midjourney's training code and data. Plaintiffs moved to compel the training source code for Midjourney's v8 image model and a repository, and a magistrate judge denied that motion as moot on August 31, 2026 after Midjourney agreed to make the code available. A stipulation on training data production was filed September 29, 2026 (Dkt. 212). The parties also agreed to a sampling protocol for subscriber prompts and outputs.
  • Source code dispute (Oct 2026): On October 21, 2026 Midjourney asked the court to reconsider a tentative order that would require production of code for its "--cref" parameter and for its Explore Page and Midjourney TV algorithms (Dkt. 223). Midjourney says it has already produced training code for every model released before the complaints, tens of millions of user prompts and outputs and billions of training images and clips.
  • Motion on the pleadings: Midjourney moved for partial judgment on the pleadings under Rule 12(c) attacking the secondary liability claim. Its reply (Dkt. 172) says plaintiffs "have abandoned their material contribution theory" and argues the complaints do not plead inducement. The hearing was set for October 26, 2026. We did not find a ruling.
  • Schedule: On July 15, 2026 the court moved the non-expert discovery cut-off to January 25, 2027, expert discovery cut-off to April 5, 2027, the last day to file all motions to April 26, 2027, and mediation to November 18, 2026 (Dkt. 130).

What the court said

There is no merits ruling in the filings we read. The court has not decided fair use, ownership or liability. The orders we read are procedural: the magistrate judge pushed the parties toward a statistical sampling protocol for the "voluminous" prompts and outputs and later denied plaintiffs' motion to compel as moot after Midjourney agreed to the code review (Dkt. 61, 175). The court approved a revised pretrial schedule because "sufficient good cause has been shown" (Dkt. 130).

Midjourney's position, in its October brief, is that other generative AI fair use cases, naming Bartz v. Anthropic and Kadrey v. Meta, "none hinged on source code minutiae." That is Midjourney's characterization, not a ruling.

What it means for your firm

  • Our reading: this is a studio suit about AI image outputs, not only training. The complaint focuses on generated images of recognizable characters, which is closer to a reproduction claim than the training-only claims in the book cases.
  • Secondary liability is under attack. If the court narrows contributory infringement to inducement, the case for holding an AI service liable for what users prompt gets harder. Watch the Rule 12(c) ruling.
  • Discovery is the real story for now: training data, source code and user prompts. Clients who build or fine-tune image models should expect these requests and should be keeping training data records.
  • Fair use is pleaded but undecided here. The two Northern District of California rulings on book training do not decide an image-output case.
  • Client protection: if a client uses AI images in marketing or client materials, characters and styles of major studios are a known risk.

What to do Monday

  1. Ask any vendor that supplies AI-generated images or video whether outputs of recognizable characters are blocked and whether the contract indemnifies copyright claims about outputs, not only training.
  2. For clients who build image or video models, confirm training data records and code are preserved in a form that could be produced under a protective order.
  3. Tell marketing and communications staff not to use prompts that name studio characters, franchises or living artists in client-facing work.
  4. Pull the live docket for No. 2:25-cv-05275 before citing this case and check for the Rule 12(c) ruling after October 26, 2026.

Sources

Questions and answers

Who is suing Midjourney, and why?

Disney, Universal and Warner Bros. Discovery entities allege Midjourney's AI image service copies and distributes images of their characters without permission. They bring direct and secondary copyright infringement claims.

What is Midjourney's defense?

Midjourney's answer pleads fair use first, and also lack of ownership, license, estoppel, unclean hands, de minimis copying and the DMCA safe harbor. None of these has been decided in the filings we read.

Has a court ruled on whether Midjourney infringed?

No. As of the latest filing we read, on October 21, 2026, the case is in discovery. The court has set an April 26, 2027 deadline for all motions, and a ruling on Midjourney's Rule 12(c) motion about secondary liability was pending after an October 26, 2026 hearing.

Operational information, not legal advice. Facts come from the filings on the court docket linked under Sources, read on Oct 27, 2026. A docket moves: this brief covers only the filings we read and does not replace the live docket. The reading of what it means for your firm is ours.