Midjourney pushes back: demands studios reveal their own AI use in court

Three Hollywood studios—Disney, Universal, and Warner Bros.—sued Midjourney for copyright infringement, arguing that the company’s image generation models were trained on copyrighted characters without authorization. Midjourney, which can generate recognizable depictions of Bart Simpson or Darth Vader from text prompts, has been in the discovery phase of litigation since the suits were filed.

In July 2026, Midjourney filed to expand the scope of discovery. A judge had already ordered the studios to disclose their use of generative AI, but limited that disclosure to consumer-facing applications. Midjourney is now seeking a broader ruling that would require the studios to reveal all AI activity—internal tools, storyboarding, effects pipelines, and model training—on the argument that the withheld documents would show whether the studios themselves train AI on unlicensed material. The legal strategy is a version of the fair use defense: if the studios do the same thing they are suing Midjourney for, that is relevant evidence.

For designers who use AI image generation tools, the case has practical implications beyond its headline. The outcome will help define where the boundary sits between permissible and infringing AI training practices, and whether generating likenesses of fictional characters constitutes fair use when the output is a commercial product. A ruling against Midjourney would likely push the industry toward more restrictive licensing requirements for training data, which could change what generative image tools are able to produce and at what cost. A ruling for Midjourney would provide clearer legal grounding for AI training on web-scale data, though it would not settle the broader question of whether individual images generated from that data infringe specific rights.

The case is among several running in parallel that will collectively shape the legal framework for AI and creative work through the rest of this decade.