Mickey & Minions vs Midjourney: Creators sue over turning characters into AI slop
In an unlikely corporate alliance, Disney and Universal are suing Midjourney for turning iconic characters like Yoda, Shrek and Iron Man into hundreds of thousands of AI-generated graphics without permission or payment.
Disney's Magic Kingdom, not AI. Image credit: Mike Wuerthele
Disney and Universal Pictures are taking AI startup Midjourney to court, accusing it of mass copyright infringement for training its image generator on the studios' most recognizable characters. In a joint lawsuit filed Wednesday in federal court, the studios accuse Midjourney of building its business on unauthorized use of their most valuable characters.
The two studios are rarely on the same side. Disney and Universal have long competed for box office dominance and theme park visitors, but they now share a common goal.
The lawsuit calls Midjourney "a bottomless pit of plagiarism." It outlines what the studios describe as a business model built on unauthorized use of protected material.
The studios are asking for an injunction to stop the alleged infringement and damages for what they describe as calculated and ongoing misuse of intellectual property.
Example of Midjourney recreating copyrighted content
Midjourney's image generator turns typed prompts into high-quality visual output. That approach is common across the generative AI industry.
These systems are trained on massive datasets collected from the internet without regard for copyright. They learn how to replicate visual styles and recognizable characters.
According to Disney and Universal, the results are not just inspired by pop culture but closely resemble copyrighted works. The lawsuit points to AI-generated images of Yoda with a lightsaber, Bart Simpson on a skateboard, and Iron Man in flight.
It also includes characters from Universal films, such as Shrek and Toothless from How to Train Your Dragon. The studios argue these examples show that Midjourney used their content to train its model without permission.
Disney and Universal say they asked Midjourney to stop using their copyrighted material or at least install safeguards to prevent further misuse. The company refused, they allege, and instead released newer models that doubled down on the problem by producing even more detailed recreations.
"Piracy is piracy," said Disney's top legal executive Horacio Gutierrez. "The fact that it's done by an AI company doesn't make it any less infringing."
NBCUniversal's Kim Harris put it more bluntly
"This is about protecting the actual artists whose work these tools are built on, and the companies that funded their creation," Harris exclaimed.
Midjourney: scraping first, worrying about rights later
Midjourney was founded in 2021 and quickly rose to prominence by offering paid subscriptions for AI-generated images. The company was reported to have had an estimated revenue of $300 million in 2024.
Its founder, David Holz, told Forbes in 2022 that they built their training data by scraping the internet. When asked if they got permission from artists, he admitted they didn't.
"There isn't really a way to get a hundred million images and know where they're coming from," he said. Other AI companies have said as much to the governments around the world.
That quote may come back to haunt him.
This is far from the first legal threat Midjourney has faced. A group of artists sued the company and others in 2023, arguing that AI firms had stored and repurposed their work without consent. That case is still making its way through the courts after a judge ruled the claims were plausible.
RIAA Chairman & CEO Mitch Glazier also commented on the lawsuit.
"There is a clear path forward through partnerships that both further AI innovation and foster human artistry," he said in a statement seen by AppleInsider. "This action by Disney and Universal represents a critical stand for human creativity and responsible innovation."
Now, with two of the US' major entertainment firms joining the fight with nearly bottomless pockets for lawyers, the stakes are higher. The lawsuit highlights the growing tension between AI companies who treat copyright law as an obstacle and creators who expect to be paid when their work is used.
Example of Midjourney recreating copyrighted content
The Recording Industry Association of America (RIAA) is a trade organization that represents major U.S. music labels and works to protect the intellectual property rights of artists and record companies.
The case is part of a growing wave of lawsuits targeting AI companies for training their models on copyrighted material. Authors, musicians, news organizations and other rights holders have raised similar concerns, arguing that AI tools built on unauthorized content undermine both copyright law and creative labor.
On a smaller scale, we've seen our own content mangled beyond coherency in the interest of AI summaries. We've seen our copy, verbatim, mashed up in news or procedural steps that make no sense when combined with other steps from other venues.
While companies like Midjourney describe the process as innovation, it is ultimately a shortcut that avoids paying the people who made the original work. And, as time has gone on, and AI feeds on AI generated content, the results are getting messier for the user, and legally.
Whether or not Midjourney thinks tracking licenses is too hard, the courts may soon decide whether that's a valid excuse or just a convenient one.
Read on AppleInsider
Comments
What Disney and Universal need to establish is that Midjourney's prominent objective is to facilitate genuine infringement - that's a high bar to reach because the overwhelming majority of Midjourney's output is not duping IP, and derivative works, satire, reporting and the like are protected under fair use. The Midjourney image pipeline in use has significantly moved beyond being just a text-to-image AI model, thus one can't merely make the case that Midjourney are naively reproducing content, when user-involvement heavily shapes the output. The latest model doesn't even work unless the user actively makes decisions about the types of content they are trying to produce - this strongly implicates individual users, and not the service.
Furthermore, the US copyright office have likened AI image generation with "image search", as this was part of their denial of protection for AI-generated works. Following this analogy, Disney & Universal would be able to sue Google for their image search feature, based on the rationale that Google facilitates the acquisition of these images which are then used in copyright infringement.
But just to rehash what is often missed in these discussions: If someone makes commercial use of intellectual property, whether by copying, hand drawing, AI generation, or any other form of image generation, outside of the fair use provisions (e.g. news reporting, satire, etc), then there would be a case to go after the individual for infringement. That has never changed.
1. Copyright holders are not "creators." They simply own IP.
2. These big entertainment companies are suing a small, startup like Midjourney because it's cheaper than suing the bigger AI companies like ChatGPT and Google – who are the real targets.
3. Disney suing anybody is a joke. Disney's "copyrights" only exist because Congress keeps extending their expiration. IP is not real property. IP is a time-bound monopoly granted by the government in order to encourage innovation. These big copyright companies abuse the original intent of patent and copyright to enrich themselves, at the expense of the public – and real creators.
4. Scraping what's publicly available on the Net is not prohibited, or a crime.
5. Copyright infringement pertains to the commercial use of protected material. Simply recreating it without profit is not a crime – although clearly the copyright cartel would have you believe that even just drawing it yourself is prohibited.
6. Nobody owns the mind of the civilization, which AI is presently mapping and making accessible. The only way any copyrighted object has value is because it taps into what the civilization has instilled in individuals. Presuming to wall off pieces of people's minds, to not let them play, work, experiment, iterate on their own thoughts – is itself the offense.
1: Not entirely true. I suspect that the misuse of WALL-E that we showed, which is literally Disney's, is an issue. Mandalorian is also Disney-made, and that guy is everywhere.
3: I don't disagree entirely. However, the two can happen at the same time, despite it being distasteful.
4: Wholesale scraping should be a crime. It is a non-zero server load, and costs the hosts money.
5: Not entirely. There is a lot of legal precedent that there is more. You may not agree with it, but it exists just the same. Welcome to end-stage capitalism. In the case of Midjourney, they are using that content to train their commercial model. They are getting commercial gain from it.
6: Ah, there's the manifesto. Ridiculous.
We have been scraped over and over for AI training. We see our content, uncredited, in AI summaries, attributed to other publications. There is an inherent argument with internet use that you viewing of ads, and clicking on some deals our monetization team finds, pays for our labor.
Nice to see the shoe on the other sock.
How did they obtain the stolen content? Did they (Disney, Warner Bros) compensate the rightful owner/artist on staff? They compensated them, and although I may not approve of the price, they did paid for it. Beyond that, anyone who captures a photograph, creates a website or document featuring original content, or designs and creates a visual or written work is also a potential target for so-called AI companies. These companies take their photographs, creations, or any other original content they produce and regurgitate it. If their AI programs are truly brilliant, why not allow them to generate something truly original on their own without access to someone else’s work?
The sole issue I have with these large content or entertainment companies is their desire to extend copyright beyond the creator’s lifetime. I believe that the original laws, which stipulated that the creator’s lifetime plus 20 years should be sufficient, should be upheld. Anything beyond that should be placed in the public domain. The public domain essentially encompasses everything created before 1920, which should provide ample content for these so-called incompetent AI companies.
However, as is often the case, the privileged few will seek to benefit from both sides of the argument. It is crucial to recognize that AI companies and those who exploit their capabilities will attempt to copyright AI-generated content and assert their rights with a smug expression, as if the rest of us are naïve and gullible for their lack of moral compass.
Here's one... everyone better go and download all the LLMs files and start stacking them, because like someone said, if you just want to make a simple image for like a wallpaper, in the future this might not be possible if the government restricts AI usage and training, oh wait China... and on and on it goes... around the world to the moon and back down to earth! It's not an easy IF this THEN that...
This is gonna be great! – Otter, Animal House