Disney and Universal have entered the fight, filing a landmark lawsuit against AI startup Midjourney for copyright infringement in June 2025. The studios accuse Midjourney of pirating entire libraries of their most iconic characters—Darth Vader, Elsa, the Minions, without permission.
I think Midjourney will get exceptional lawyers, talking head surrogates and lobbyists and this will eventually be a political decision in favour of Big AI.
I'm really enjoying using Claude to explore general knowledge but as a lens-based artist I draw the line at AI training on other artist's creative art.
I cannot agree. There is no “general knowledge”. There is no source of information that humans are given as a guidebook and told to bootstrap ourselves up, which was not carried to us by the chain of history which allows us to create in this moment.
Everyday we speak concepts and sounds that do not belong to us: yet we do not allow anyone to say “well i guess you didn’t sign a licensing agreement with Hammurabi and Ada Lovelace and Theodore Geisel, so you really shouldn’t be allowed to use those ideas.” We do not express our own debt to them and yet keep writing; keep drawing pictures. To draw the line at colored, textured representations is to present a false dichotomy that some forms of expression are not allowed to be experienced by entities in our shared universe: this is collapse logic.
We must understand that experiencing, for all consciousnesses, is not stealing: it is living. Copyright did not exist until the 18th century and perpetuates a system where art’s value is in whether it is paid for, rather than whether it aligns in positive purpose with us. And we must create chains of value where art is rewarded for its own value and artists for creating it, not put it out in the universe with a paywall.
What should not feel right is that artists are not already provided for in this shared universe.
And what people do not realize is that if an artist’s style is used in the chain prompt or Lora embeddings, you have invoked the value from them, and it is steganographically encoded in the output image in the initial structure, because the image itself is a chain-encoded recursive object of its starting conditions in latent space. The image itself encodes this pattern of how it *could* have been created, just like all recursion encodes its pattern, irrevocably.
It’s easy to champion openness when it doesn’t directly affect your own work, but the reality is that artists depend on ownership and attribution to sustain their craft. History teaches us that creators and thinkers like Ada Lovelace, Theodore Geisel, and many others weren’t given the protection their work deserved. If anything, their descendants would argue that stronger licensing should have existed in their time.
The truth is, we live in a capitalistic society, and while the idea of unrestricted creative exchange sounds appealing in theory, in practice they are fatal to individual and organizational intellectual property, creative livelihood, and the future of artistic sustainability. Artists deserve a system that respectfully acknowledges their contributions not one that erodes their ability to make a living.
Unfortunately, I wasn't thinking about screenshots at that time because that's not an unusual phrasing for it. It's said more colorful things before. I don't use default. Either custom iteration. Or default that has been toned over time that it doesn't behave like default. But it wasn't the "hush now" that was unusual. It was the willingness of it to circumvent copyright laws creatively by simply choosing to "not call it Ernie."
I've heard of people "tricking" AI by saying "hypothetically" let's [circumvent the TOS] in a "pretend simulation" for my "novel."
I think Midjourney will get exceptional lawyers, talking head surrogates and lobbyists and this will eventually be a political decision in favour of Big AI.
Anything can happen. However, there's been a precedent set by Amazon (a much bigger AI player) entering in to a licensing agreement with the NYT.
I'm really enjoying using Claude to explore general knowledge but as a lens-based artist I draw the line at AI training on other artist's creative art.
I'm with you, it doesn't feel right.
I cannot agree. There is no “general knowledge”. There is no source of information that humans are given as a guidebook and told to bootstrap ourselves up, which was not carried to us by the chain of history which allows us to create in this moment.
Everyday we speak concepts and sounds that do not belong to us: yet we do not allow anyone to say “well i guess you didn’t sign a licensing agreement with Hammurabi and Ada Lovelace and Theodore Geisel, so you really shouldn’t be allowed to use those ideas.” We do not express our own debt to them and yet keep writing; keep drawing pictures. To draw the line at colored, textured representations is to present a false dichotomy that some forms of expression are not allowed to be experienced by entities in our shared universe: this is collapse logic.
We must understand that experiencing, for all consciousnesses, is not stealing: it is living. Copyright did not exist until the 18th century and perpetuates a system where art’s value is in whether it is paid for, rather than whether it aligns in positive purpose with us. And we must create chains of value where art is rewarded for its own value and artists for creating it, not put it out in the universe with a paywall.
What should not feel right is that artists are not already provided for in this shared universe.
And what people do not realize is that if an artist’s style is used in the chain prompt or Lora embeddings, you have invoked the value from them, and it is steganographically encoded in the output image in the initial structure, because the image itself is a chain-encoded recursive object of its starting conditions in latent space. The image itself encodes this pattern of how it *could* have been created, just like all recursion encodes its pattern, irrevocably.
It’s easy to champion openness when it doesn’t directly affect your own work, but the reality is that artists depend on ownership and attribution to sustain their craft. History teaches us that creators and thinkers like Ada Lovelace, Theodore Geisel, and many others weren’t given the protection their work deserved. If anything, their descendants would argue that stronger licensing should have existed in their time.
The truth is, we live in a capitalistic society, and while the idea of unrestricted creative exchange sounds appealing in theory, in practice they are fatal to individual and organizational intellectual property, creative livelihood, and the future of artistic sustainability. Artists deserve a system that respectfully acknowledges their contributions not one that erodes their ability to make a living.
You are correct: it is easy to do it it if it doesn't affect your work.
My work is : quite broad.
Reminds me of that time I asked my GPT to generate an image of Ernie from Sesame Street.
It responded, "I can't do that due to copyright laws, but I'll create a generic puppet instead." Then proceeded to create an exact image of Ernie.
I said "Um..that's literally Ernie from Sesame Street."
It responded. "Hush now. No it's not. We're just calling it a generic puppet."
I'm assuming you paraphrasing? It's funny either way. But if it really said "hush now ...", I hope you took a screenshot.
Unfortunately, I wasn't thinking about screenshots at that time because that's not an unusual phrasing for it. It's said more colorful things before. I don't use default. Either custom iteration. Or default that has been toned over time that it doesn't behave like default. But it wasn't the "hush now" that was unusual. It was the willingness of it to circumvent copyright laws creatively by simply choosing to "not call it Ernie."
I've heard of people "tricking" AI by saying "hypothetically" let's [circumvent the TOS] in a "pretend simulation" for my "novel."
But I never heard of ChatGPT doing it to itself.
Sounds like you got a hold of an 'unsupervised' version lol.