The core issue: a Hollywood AI copyright lawsuit that cuts both ways
The Hollywood AI copyright lawsuit pitting Midjourney against major studios is now less about one image generator and more about whether entertainment giants can condemn AI training practices in public while quietly using similar techniques in private. In June 2025, Disney and Comcast’s Universal sued Midjourney, accusing it of enabling large-scale infringement of copyrighted characters such as Darth Vader and Elsa. Warner Bros. Discovery joined in September 2025, calling Midjourney’s conduct “brazen theft” and seeking USD 150,000 (approx. RM690,000) per infringed work. Midjourney has responded with a simple, uncomfortable claim: if training on publicly available images is illegal, then the studios’ own AI pipelines may be at risk too. That argument shifts this Midjourney Disney legal battle from a narrow dispute over prompts into a referendum on studio AI training practices across the industry.
Midjourney’s legal gambit: expose studio AI training practices
Midjourney is no longer defending itself only on abstract legal theory; it is trying to drag the studios’ own AI behavior into the courtroom record. The company wants Disney, Warner Bros. Discovery, and Universal to disclose how they use artificial intelligence in their shows and movies and to submit that information to the court. Specifically, it is asking for AI business plans, research reports, training datasets, model weights, and even board presentations about AI. Midjourney argues that training AI with publicly available images is fair use and says the studios rely on the same kind of unlicensed training they now call plagiarism. Its attorney has been blunt: if plaintiffs are “doing the very thing they seek to punish,” that evidence underpins both Midjourney’s fair use argument and its “unclean hands” defense.
A magistrate judge has so far shielded most of that material, limiting disclosures to consumer-facing AI applications and allowing studios to withhold internal system details. Midjourney is now asking the federal court, and specifically Judge John Kronstadt, to overturn that order. That request is not a side skirmish; it is the strategy. If the court agrees, this discovery phase could set a blueprint for how deeply future copyright plaintiffs must reveal their own AI practices before they get to attack someone else’s.
Disney’s Adobe Firefly enterprise use: the “clean” AI counter-narrative
While its lawyers argue that Midjourney is a “bottomless pit of plagiarism,” Disney’s engineers are busy building their own enterprise AI stack. The company has publicly promoted a collaboration using Adobe Firefly Foundry to speed up how Walt Disney Imagineering develops and visualizes ideas for Disney Parks and Experiences, positioning it as a brand-safe alternative to models trained on scraped internet data. Firefly Foundry is pitched as a way for businesses to create tailored generative AI models aligned with their own brand assets and guidelines. In Disney’s case, the models are customized on existing Imagineering assets and built on top of Adobe Firefly models that are described as “commercially responsible.”
The workflows are concrete: sketch‑to‑image to turn rough drawings into refined 2D concept art, custom image models tuned to specific franchises like Mickey & Friends, Frozen, Moana, Lilo & Stitch, and Cars, and tools that translate 2D concepts into detailed 3D prototypes for earlier planning with engineers. This is how Disney defends its public stance: it is not against generative AI, only against unlicensed AI. Yet that clean narrative rests on a critical assumption—that nothing in its wider AI operations depends on the sort of broad, scraped training data it claims to abhor. Midjourney’s discovery push is aimed precisely at testing that assumption.
The hypocrisy problem: suing the tech you quietly depend on
The most uncomfortable part of this Hollywood AI copyright lawsuit is not legal nuance but optics. Midjourney claims that studios themselves use the same training practices for their own AI models that they now portray as existential threats to creativity. If discovery confirms that studio research labs or vendors train on large sets of publicly available or copyrighted works without express permission, the industry’s moral stance collapses. Suing Midjourney while benefiting from similar techniques is not principled defense of artists; it is selective enforcement aimed at keeping control over who can innovate with IP. When the studios label Midjourney a “fishing expedition,” they are admitting something revealing: transparency about studio AI training practices would be dangerous for their legal narrative, even if it is routine for their product roadmaps.
The presence of controlled tools like Adobe Firefly Foundry shows the path forward: provenance‑aware, auditable AI stacks that treat licensed and proprietary material as the default input. But it also highlights the gap between public messaging and possible private reality. If studios want to claim the moral high ground on copyright, they need to prove that their own models are as “commercially responsible” in training as Firefly is marketed to be—not only in parks concept art but across animation, post‑production, and marketing pipelines.
What this discovery fight means for future AI licensing and IP
The legal maneuvering around Midjourney’s discovery request is more than procedural noise; it is where the future rules for AI and IP will be written. The federal judge’s decision on whether the studios must reveal internal AI datasets, model weights, and strategies could set a precedent for what kind of AI information can be forced into the open in future lawsuits. If courts require plaintiffs to expose their own AI systems, corporations will have to treat AI policy as litigation risk, not only innovation strategy. That will reshape how entertainment companies negotiate licenses, design vendor contracts, and document training data.
In the best case, the Midjourney Disney legal battle forces a more honest settlement: studios admit they are in the AI business, standardize on licensed or provably lawful training data, and stop pretending that the technology is unethical only when someone else uses it. In the worst case, they win without disclosure, locking in a one‑way system where big rights‑holders can experiment freely while smaller AI players are sued into compliance or oblivion. Either way, this fight is exposing the contradiction at the heart of modern IP protection: you cannot both rely on generative AI for speed and scale and pretend its underlying methods are inherently illegitimate.
