🤚 The Open-Palm Subpoena

Midjourney, the AI image generator that trained its models on approximately everything with a pixel, has asked a federal court to force Disney, Universal, and Warner Bros. to disclose exactly how they use AI internally. The company is demanding the studios hand over their AI business plans, research reports, training datasets, model weights, and even the board presentations where executives presumably used the phrase “generative AI strategy” while a McKinsey consultant nodded approvingly.

The context: Disney and Universal sued Midjourney in 2025 for copyright infringement, noting that the platform could generate images of characters like Bart Simpson and Darth Vader. Warner Bros. filed its own lawsuit later that year, adding Superman and Batman to the list of allegedly violated intellectual properties. Midjourney’s response, in essence, has been: “Interesting. Now show us yours.”

👐 The Two-Handed Mirror

Midjourney attorney Bobby Ghajar put it with the kind of precision that makes opposing counsel reach for their antacids:

“If Plaintiffs are doing the very thing they seek to punish, that evidence goes to the heart of Midjourney’s fair use and unclean hands defenses.”

The “unclean hands” doctrine, for those who went to law school and immediately forgot, is a legal principle stating that a party cannot seek a remedy if they themselves have engaged in the same wrongful conduct. It is the judicial equivalent of “he who smelt it, dealt it,” but with billable hours.

A magistrate judge ruled in mid-June that the studios only need to disclose AI-generated content that was “consumer-facing” — meaning whatever Disney used Midjourney for in a boardroom stays in the boardroom. Midjourney is now appealing that ruling, arguing that the studios are “withholding precisely those documents that would reveal whether, behind closed doors, they are doing exactly what they are suing Midjourney for doing.”

This is, to be clear, the company that scraped the entire visible internet — every DeviantArt portfolio, every stock photo watermark, every photograph a wedding photographer posted to their business page — now asking a court to compel others to be transparent about their data practices. The audacity is immaculate.

🌿 The Gentle Awakening

But here’s the thing: Midjourney might not be wrong.

Hollywood studios have been enthusiastically adopting AI tools behind closed doors while publicly positioning themselves as defenders of human creativity. Disney has filed patents for AI-generated animation pipelines. Universal has explored AI-driven storyboarding. Warner Bros. tested AI scripts before the ink was dry on the SAG-AFTRA agreement that was supposed to prevent exactly that.

The studios want to protect their copyrighted characters from being reproduced by AI. That is a legitimate legal interest. But they also want to use AI to create new content, potentially training those tools on other people’s copyrighted work in the process. That is what the legal profession calls “a problem.”

What Midjourney has effectively done is turn a copyright infringement lawsuit into a referendum on the entire industry’s relationship with generative AI. If the court grants full discovery, we might learn that the same studios suing over unauthorized Darth Vader generations have internal tools that were trained on datasets of equally questionable provenance.

If the court denies it, the precedent is clear: major corporations can use AI however they like internally, sue anyone who does the same thing externally, and keep the evidence locked in a vault next to Walt Disney’s cryogenic chamber.

👑 The Gold-Leaf Reckoning

This case is not really about Bart Simpson. It is not about Superman. It is about whether the rules of the AI economy apply equally to a $12 billion startup and a $170 billion media conglomerate, or whether copyright law functions like a velvet rope — permeable if you’re on the guest list.

The fair use doctrine was designed to balance the rights of creators with the public interest in new works. What nobody anticipated is that both sides of a lawsuit would be doing the same thing, at the same scale, with the same technology, and that the only material difference would be that one side has a theme park and the other has a Discord server.

Midjourney scraped the internet. The studios may have scraped the internet. Midjourney built generative models. The studios may have built generative models. Midjourney made money from the output. The studios definitely made money from the output. The only party in this lawsuit with genuinely clean hands is the judge, and even she probably used ChatGPT to summarize the briefs.

“We trained on their characters. They trained on everyone’s characters. The only difference is that we admitted it on a website instead of in a shareholder letter. Discovery is just honesty with a subpoena.” — The Slap of Wisdom Intellectual Property Desk, currently generating a legally distinct image of a mouse with round ears that is definitely not anyone’s trademark