MidJourney is a generative AI application: a form of AI that trains on data to produce outputs based on that data. Keywords: ‘based on that data’. As an example, do you remember Google’s Deep Dream? The reason it would always give you things that were full of eyes was because it was trained to identify animals, so it would ‘see’ eyes everywhere looking for the dog. That was its training data, that’s what it produces. Similarly, Generative AI programs like MidJourney (and friends) must consume content to create content. But where do you get enough content to creatively make new pictures when a user asks? How does a computer learn what colors a cat or dog can be?
This is one of the main reasons Generative AI is so controversial: it’s widely believed that almost every Generative AI is training on material scraped off of the open web! This is naturally including a ton of stuff it didn’t have permission to touch via copyright protection. There is evidence: AI crawlers ignoring the robots.txt (a bit of code at the front of a website that is normally used to tell search engine crawlers what they can and cannot index) has caused noticeable issues with traffic for some websites, and has lead website owners to start projects designed to ‘trap’ or trick them since telling them ‘no’ doesn’t work. Proving it is the hard part: do we know it used any particular image, copyrighted or not, to make any particular picture it spits out? Because no individual image can be identified in the input, it’s incredibly difficult to bring a case when the output is ‘original’.
How do you prove the theft well enough to sue, or get it removed from the training data? Is removal of your work even possible?
The answer, so far, has been ‘we can’t do that! So if you want to support artists, you shouldn’t use this machine with mysterious training sources’. However, thanks to the tight copyright Disney and Universal manage, it’s entirely possible for them to figure out the machine has stolen their images. The problem for a lot of other artists is that a lot of other artists are producing on the scale of one person. Their contribution to the training data won’t be enough for someone to perfectly recreate a character they made on name alone. Universal and Disney have the combined efforts of millions of fans (many of which are producing art or taking pictures on Disney property with characters) on top of their own marketing efforts and the movies or cartoon series they themselves have produced. There are exponentially more drawings of, say, Po from Kung Fu Panda than there are fanworks of someone’s webcomic, and they’re all tagged by name. So if you ask Midjourney to make a picture of Po (not ‘chubby panda 3D art style with brown pants in front of a bamboo forest’), it will be able to do so with such accuracy that it’s impossible to say it hasn’t absorbed copyrighted content. This is their argument. It is able to produce images of things it shouldn’t be allowed to, shouldn’t be able to, because it’s been scraping official Disney properties in its never-ending data guzzling, and it recreates it perfectly for whoever asks, for whatever reason.
Is This Good?
Generative AI harvests user content posted publicly with no regard for what the artists prefer to make pictures for other users, and doing so often costs a lot of computer power. However, Disney’s aggressive pursuit of them because of copyright is also not good! Their complaint in the lawsuit seems to suggest the only reason they don’t pursue more fanartists making art of their characters is because they don’t have a case if the artist is not directly making money off the fanart, which would include artists who have Patreon but don’t sell prints (or make ad revenue, or take commissions on copyrighted characters, etc.) of individual works of art. It seems to suggest they’d love to pursue those people after all, because Midjourney, too, is not directly profiting from generating images of Disney properties, but is making money off of subscriptions, and make life a living nightmare for fandom artists. Never mind the idea of the ‘Disney Style’! The concept of copyright itself does not keep up with the times we live in! Metallica used to sue for six- or seven-digit numbers when they discovered ordinary individuals had pirated a song, and now what? Yes, finally a worthy victim of copyright lawsuits? Three giants are fighting, and depending on what the court says, the impact on small artists could be catastrophic.
https://www.cnn.com/2025/06/11/tech/disney-universal-midjourney-ai-copyright-lawsuit
