Midjourney’s bold demand: if you sue over AI, show your own models
The Midjourney copyright lawsuit is a generative AI legal dispute in which major studios accuse an image generator of infringing their characters, while Midjourney responds by demanding AI training transparency from those same studios in court.
What makes this case different is not that Hollywood studios are suing an AI company; it is that Midjourney is trying to drag the studios’ own AI practices into the light. In June 2025, Disney and Comcast’s Universal filed suit accusing Midjourney of enabling large-scale infringement of copyrighted characters, describing it as a “bottomless pit of plagiarism.” Warner Bros. Discovery joined three months later, calling Midjourney’s conduct “brazen theft” and seeking USD 150,000 (approx. RM690,000) per infringed work. Now the defendant is fighting back by insisting that if studios want to police AI training, they must first admit how they train their own models.
Turning discovery into a spotlight on Hollywood studios’ AI
Midjourney’s legal strategy is simple but explosive: to defend itself, it wants to show that Disney, Universal, and Warner Bros. Discovery do the same things they claim are unlawful. The company is asking the court to force disclosure of the studios’ AI business plans, research reports, training datasets, model weights, and even AI presentations delivered to their boards. In other words, it wants a complete map of Hollywood studios’ AI.
This is not a side skirmish; it is the heart of the case. Midjourney has staked its defense on the argument that training models on publicly available images is fair use and that studios rely on similar methods internally. If that is true, then the lawsuit becomes less about protecting artists and more about who is allowed to exploit AI at scale. The discovery fight is where that hypocrisy, if it exists, would be exposed.
The clash over AI training transparency and ‘unclean hands’
So far, the studios have mostly kept their AI black box shut. A magistrate judge recently allowed them to withhold most internal AI information and limited discovery to “consumer-facing” AI tools. Unsurprisingly, Midjourney has appealed to the federal judge, asking him to overturn that decision and restore its broader request.
Midjourney’s attorney argues that if the plaintiffs are “doing the very thing they seek to punish,” that evidence goes to the core of its fair use and “unclean hands” defenses. The studios’ lawyer has dismissed this as a “fishing expedition,” claiming Midjourney is trying to distract from its own conduct. But calling it a distraction misses the point: in a world where generative AI legal disputes hinge on what training is permissible, pretending that only one side’s data and models matter is not law; it is selective enforcement.
Why this double standard matters far beyond one AI image generator
This case is not only about whether Midjourney’s specific training methods qualify as fair use; it is about whether powerful rights-holders can hide their own generative AI habits while punishing others for similar behavior. Hollywood studios AI programs do not exist in a vacuum. If they are quietly training models on copyrighted works while suing Midjourney for comparable training, they are asking courts to bless a double standard in how AI is regulated and enforced.
The federal judge’s ruling on discovery could set a precedent on what AI training information must be disclosed in future lawsuits. If he sides with the studios, other plaintiffs will learn they can sue AI companies without revealing their own models. If he backs Midjourney, corporations may have to choose: either enforce copyright aggressively or keep their AI secrets, but not both. That trade-off would finally force a more honest debate about how generative AI should be trained and who gets to profit from it.
