📰 Key Takeaways

Midjourney is mounting a counterattack in copyright lawsuits filed by three major Hollywood studios — Disney, Universal, and Warner Bros. — asking the court to compel the studios to fully disclose their own AI usage. All three studios sued Midjourney last year, alleging that its image generation model can output copyrighted characters including Bart Simpson from The Simpsons and Darth Vader from Star Wars, and arguing this constitutes infringement. Midjourney is defending itself on “fair use” grounds, contending that training models on copyrighted images is lawful.

The dispute is currently centered on the pre-trial evidence-gathering phase. The judge previously ruled that the studios must produce records of their generative AI usage, but the scope was limited to documents related to final consumer-facing films or images. In its latest motion, Midjourney is asking the court to lift that restriction, arguing it lets the studios cherry-pick documents favorable to their case while suppressing evidence that could bolster Midjourney’s defense.

Midjourney further alleges that the documents the studios are withholding may reveal they too have been training AI models on unauthorized copyrighted content — for example, internal image-generation tools used for storyboarding or creative ideation in film and TV production. If true, this would confirm such practices are an industry norm, and the studios are no exception. Midjourney is also asking the studios to turn over every prompt ever entered on its platform along with the corresponding generated outputs, not only those alleged to be infringing. The studios deny any deliberate withholding, emphasizing that the lawsuit’s goal is not to suppress AI technology but to force Midjourney to stop producing unauthorized copies of their iconic characters.

💬 JudyAI Lab’s Take

Midjourney’s demand that the three major Hollywood studios disclose their own AI usage records is more than just courtroom maneuvering — it could drag the industry’s unwritten rule that “rights holders also train AI behind closed doors” into the legal spotlight for the first time.

This case exposes a critical asymmetry: the plaintiff may be a practitioner of the very behavior they’re suing over. Midjourney’s core argument is that if the studios are also training internal AI tools on unauthorized material (for storyboarding or creative ideation), the moral premise of the lawsuit itself deserves scrutiny. Equally telling is the court’s restriction on evidence scope — requiring only “consumer-facing” output records introduces an inherent selection bias into discovery. It’s a reminder that until the legal framework around generative AI copyright takes shape, the gray zone around training data is a condition almost the entire industry shares — not the exception of any single company.

What you can do right now: audit the AI tools you use (or the models behind them), check the source and licensing status of their training data, and keep documentation. Once the legal framework lands, clear data provenance will be the bare minimum for self-protection.

📅 Source Info

🔗 Further Reading