Disney, Warner Bros and Universal could be forced to come clean on AI

Here’s something you don’t see every day: three of the biggest movie studios in Hollywood—Disney, Warner Bros. and Universal—have found themselves in a bit of an AI-related legal kerfuffle, and the whole thing is getting oddly revealing. The tangled mess centres around an ongoing copyright argument with Midjourney, which if you haven’t kept up, is one of those buzzy AI image generators that can whip up ‘original’ images based on user prompts (and, not coincidentally, on lots of other people’s artwork).

The Copyright Row: Who Started What

Long story short, Disney and Universal fired the starting pistol on this dispute over a year ago, filing lawsuits against Midjourney. Their claim? Midjourney was allegedly building its AI by trawling through buckets of copyrighted images—think film art, posters, conceptual sketches—without permission. Because of that, users could then generate images in the distinctive style of the big studios, which, unsurprisingly, the studios weren’t too keen on. Warner Bros. joined the dogpile later with similar accusations.

Midjourney’s not exactly rolling over. Their stance is that what they’re doing should be considered 'fair use', and they’ve gone one further by making a rather pointed accusation right back: they reckon the studios themselves have got their own hands dirty using AI trained on other people’s copyrighted material.

Fighting to See Behind the Curtain

This brings us to the juicy part. In June, a judge actually agreed to let Midjourney request info from Disney, Warner and Universal about their own use of AI—specifically, what the court called 'consumer-facing' AI tools. But that’s not enough for Midjourney. They’ve filed an appeal, pushing the court to force the studios to share a much fuller picture.

Midjourney’s lawyer, Bobby Ghajar, is arguing that if these film giants are developing their own internal AI tools for things like storyboarding or cooking up new ideas, and if those AIs are being trained in the very way they’re complaining about—on unlicensed, third-party copyrighted content—then surely, that’s relevant to the case. Here’s how Ghajar put it:

"If Plaintiffs are developing image-generating AI models—trained on unlicensed, third-party copyrighted data—for internal use in storyboarding or ideating content for film or TV, that evidence would equally demonstrate that it is an industry custom, even among the studios themselves, to download and train AI on unlicensed copyrighted content."

What Exactly Is Midjourney After?

• Internal AI business plans
• Research reports relating to their use of AI
• Details of AI model weights and training data
• Board meeting minutes where AI is discussed
• Basically, the full rundown of how AI is used to create and market film and TV content

The upshot: the studios say they’ll show data on products or tools they make available to the public, but that anything internal is off limits. They’re being particularly cagey about letting people peek at their in-house systems and secret recipes for using AI—possibly because, well, those recipes might look a lot like the ones they’re suing over.

The judge in June actually sided with the studios on this point, arguing that the wide-ranging data Midjourney wanted was not really relevant to the original copyright claim. For now, at least, the court says the defence can’t go rooting around in the studios’ private AI stash—but given the appeal, the story’s not over.