Class Action Filed Against Stability AI, Midjourney and DeviantArt for DMCA Violations, Right of Publicity Violations, UnlawfulCompetition, TOS Breach

---------------------------------------------------------------------------
SAN FRANCISCO, Jan. 14, 2023 /PRNewswire/ -- Stability AI Ltd.; Stability AI, Inc.; Midjourney Inc.; and DeviantArt, Inc. have created products that infringe the rights of artists and other creative individuals under the guise of alleged "artificial intelligence." The Joseph Saveri Law Firm, LLP—a leading class action firm with offices in California and New York—along with Matthew Butterick, and Lockridge, Grindal, Nauen P.L.L.P. have filed a lawsuit in the United States District Court for the Northern District of California on behalf of a class of plaintiffs seeking compensation for damages caused by Stability AI, DeviantArt, and Midjourney, and an injunction to prevent future harms. The lawsuit alleges direct copyright infringement, vicarious copyright infringement related to forgeries, violations of the Digital Millennium Copyright Act (DMCA), violation of class members' rights of publicity, breach of contract related to the DeviantArt Terms of Service, and various violations of California's unfair competition laws.

As alleged in the Complaint, Stable Diffusion is an artificial intelligence product used by Stability AI, DeviantArt, and Midjourney in their AI image products. It was trained on billions of copyrighted images contained in the LAION-5B dataset, which were downloaded and used without compensation or consent from the artists. If Stable Diffusion and similar products are allowed to continue to operate as they do now, the foreseeable result is they will replace the very artists whose stolen works power these AI products with whom they are competing. AI image products are not just an infringement of artists' rights; whether they aim to or not, these products will eliminate "artist" as a viable career path. In addition to obtaining redress for the wrongful conduct, this lawsuit seeks to prevent that outcome and ensure these products follow the same rules as any other new technology that involves the use of massive amounts of intellectual property. If streaming music can be accomplished within the law, so can AI products.

"As burgeoning technology continues to change every aspect of the modern world, it's critical that we recognize and protect the rights of artists against unlawful theft and fraud," said Joseph Saveri, founder of the Joseph Saveri Law Firm, LLP. He continued, "This case represents a larger fight for preserving ownership rights for all artists and other creators."

"AI needs to be fair and ethical for everyone," said lawyer/programmer Matthew Butterick. "But Stability AI, Midjourney, and DeviantArt are appropriating the work of thousands of artists with no consent, no credit, and no compensation. As a lawyer who is also a longtime member of the visual-arts community, it's a pleasure to stand up on behalf of fellow artists and continue this essential conversation about how we the people want AI to coexist with human culture and creativity."

Since its founding in 2000, DeviantArt had grown to be a haven for artists of all stripes. A core aspect of participating in the DeviantArt community for artists is the practice of sharing digital images of their artwork. Today, DeviantArt bills itself as "the world's largest art community," hosting millions of images. At the same time, it offers DreamUp, a product that unlawfully infringes on the rights of its own art community. To add insult to injury, a large portion of the training data for Stable Diffusion—which powers DreamUp—was made up of images scraped from DeviantArt without permission from the artists that posted them.

For more information, please see our case page www.saverilawfirm.com/ai-art-generators-copyright-litigation and our case website stablediffusionlitigation.com.

ABOUT THE FIRMS

The Joseph Saveri Law Firm is one of the country's most acclaimed, successful boutique firms, specializing in antitrust, class actions, and complex litigation on behalf of national and international consumers, purchasers, and employees across diverse industries. For further information on our practice and accomplishments on behalf of our clients, please visit www.saverilawfirm.com or call us at (415) 423-1799.
Lockridge Grindal Nauen P.L.L.P. has served clients throughout the Midwest and in Washington, D.C. for more than 40 years. It has extensive experience in local, state, and federal government relations as well as antitrust, business, campaign finance, consumer, data breach, governmental, health care, employment, environmental, ERISA, intellectual property, real estate, securities, and tribal law litigation. For further information, please visit https://www.locklaw.com/.

ABOUT MATTHEW BUTTERICK

Matthew Butterick is a lawyer, programmer, designer, and writer. He has been professionally involved with open-source software since 1998. His books Typography for Lawyers (typographyforlawyers.com) and Practical Typography (practicaltypography.com) are relied on daily by lawyers and writers worldwide. For more information, please visit https://matthewbutterick.com.
---------------------------------------------------------------------------
 
If streaming music can be accomplished within the law, so can AI products.
The thing is, it's not all the same. Streaming music isn't about creation, it's about distribution.

What about the rights of AI to be creative?

When one day we get true, sentient AI (not the "AI" that everyone is trumpeting now), this would make it illegal for them to be creative. Think Data on Star Trek doing his paintings. The last thing you want is Skynet becoming a frustrated painter. That didn't work out too well for the world 100 years ago...
 
I think most people who use the arguments that AI art is "derivative" or "transformative" don't exactly understand why the artists are big mad. In the art bubble there is an invisible no-stealing-100%-nah-uh honor system in place, and the more famous/popular the artists are, the more strictly it is enforced. People have been accused of stealing composition, lighting, specific character poses, etc in their own "derivative" and "transformative" artworks. And once you get branded as an art thief other artists will be sure to remind any employer and patrons about your theft, which unusually scares off any potential work you might get. If an artist knows that their own work show shadows of inspiration from another's, they most often say so in the description, or link the inspiration in some way. (This does not always work and the "inspiration piece" artist might try to make them delete the art post. Twitter fights and drama ensues)
Looking at it from the outside one might say "wow these artists are such whiny babies! just share your skills!", but please understand that art and design is a competitive and time sensitive field. One art piece with the right elements posted at the right moment could earn someone a contract or project. On and beyond the mid-upper rungs of the art arena most artists operate on that honor system. It's what everyone has agreed to be fair.

AI art is steamrolling and pissing all over that.
 
Honestly? Good. AI artbros are the most obnoxious faggots I've ever encountered.

Way way way too many of those faggots are going and saying shit like "Now art is no longer gatekept behind artists" or you've got lazy asshole artists who use it to scam commissioners.

For example, two years ago I commissioned an artist to draw my sister's D&D character as a gift for her. When my turn for him to work on the piece came around, he worked on it for a month and then sent a DM saying that he was going to be moving to Brazil from Japan, and then vanished for a year and a half.

After I pestered him to either finish the art that I had paid for or give me a refund when he had returned, he said that the picture had been lost and he needed to restart it, and then a week later he gave some lazy ass AI art that looked nothing like the references my sister had drawn, and that he had lazily traced... But the dude was so retarded that he traced everything, including the extra fingers and other giveaways that it was AI generated.

So seriously, fuck AI art, fuck AI artbros, and fuck lazy artists. Maybe AI art has some use for giving artists ideas but the way it is being championed as a replacement for artists will never catch on, since it is soulless shit.

Go ahead and give me all the MATI stickers, because I legitimately am mad at AI bros and have no time for faggots who don't see the issues with AI art.
 
Last edited:
I think most people who use the arguments that AI art is "derivative" or "transformative" don't exactly understand why the artists are big mad. In the art bubble there is an invisible no-stealing-100%-nah-uh honor system in place, and the more famous/popular the artists are, the more strictly it is enforced. People have been accused of stealing composition, lighting, specific character poses, etc in their own "derivative" and "transformative" artworks. And once you get branded as an art thief other artists will be sure to remind any employer and patrons about your theft, which unusually scares off any potential work you might get. If an artist knows that their own work show shadows of inspiration from another's, they most often say so in the description, or link the inspiration in some way. (This does not always work and the "inspiration piece" artist might try to make them delete the art post. Twitter fights and drama ensues)
Looking at it from the outside one might say "wow these artists are such whiny babies! just share your skills!", but please understand that art and design is a competitive and time sensitive field. One art piece with the right elements posted at the right moment could earn someone a contract or project. On and beyond the mid-upper rungs of the art arena most artists operate on that honor system. It's what everyone has agreed to be fair.

AI art is steamrolling and pissing all over that.
That just shows why these suits are a joke - the cardinal rule of art : DONUT STEEL! : as you said, is a gentleman's agreement at best, and such can never ever be enforced through a court. You have to sign an actual contract, and absent that, it comes down to the law, which already has plenty of cases that say it takes more than just similarity of character designs for a genre to prove someone stole your shit.

This isn't new, either, back in Web 1.0 there were BBS fights over accusations of theft between game programmers because someone was insisting THEY came up with the idea of the bulky, buzzcut, space marine and Quake II ripped them off!!!
 
In the art bubble there is an invisible no-stealing-100%-nah-uh honor system in place, and the more famous/popular the artists are, the more strictly it is enforced.
On and beyond the mid-upper rungs of the art arena most artists operate on that honor system. It's what everyone has agreed to be fair.
Yeah, these people should fuck off.

Rembrandt, 1655 / Soutine, 1924
imagen_2023-01-16_091745006.pngimagen_2023-01-16_091932150.png

imagen_2023-01-16_092353011.pngimagen_2023-01-16_092401261.png

Velasquez 1653 / Picasso 1957
imagen_2023-01-16_092717363.png

Van Gogh 1890 / Dore 1872
imagen_2023-01-16_092923689.png

All those tantrums are pure jealousy. They're mad that they aren't as unique and talented as they think they are. The big masters of art through history did things nobody else did before, even if they copied others. Picasso and Van Gogh added their own style, which then was new and original on itself and became something new. That's why they are copied or inspire others until today.

These modern "artists" will never be this groundbreaking and they resent when someone with equal talent shows up. They are also not very original. Or their work is basically fanart, whose copyright is already debatable on itself. There was a good thread about how good artists are still around painting original stuff
imagen_2023-01-16_094225814.pngimagen_2023-01-16_094254856.png

These people aren't the ones complaining about AI. Jee, wonder why.
 
I think most people who use the arguments that AI art is "derivative" or "transformative" don't exactly understand why the artists are big mad. In the art bubble there is an invisible no-stealing-100%-nah-uh honor system in place, and the more famous/popular the artists are, the more strictly it is enforced. People have been accused of stealing composition, lighting, specific character poses, etc in their own "derivative" and "transformative" artworks. And once you get branded as an art thief other artists will be sure to remind any employer and patrons about your theft, which unusually scares off any potential work you might get. If an artist knows that their own work show shadows of inspiration from another's, they most often say so in the description, or link the inspiration in some way. (This does not always work and the "inspiration piece" artist might try to make them delete the art post. Twitter fights and drama ensues)
Looking at it from the outside one might say "wow these artists are such whiny babies! just share your skills!", but please understand that art and design is a competitive and time sensitive field. One art piece with the right elements posted at the right moment could earn someone a contract or project. On and beyond the mid-upper rungs of the art arena most artists operate on that honor system. It's what everyone has agreed to be fair.

AI art is steamrolling and pissing all over that.
wow these artists are such whiny babies
 
There's one detail you've forgotten about, the Supreme Court had already ruled that any intellectual property generated by an AI can't be copyrighted because it wasn't created by a person, and copyright is for people.

So basically you want to live in China, where any idea or invention you create can just be taken from you by the government without compensation or recourse, and you live under the oppression of an AI that manages your life through a social credit score system & facial recognition cameras? Because that's essentially the sum of what you just said.
You're going to wind up with precedent being like Shorpy(historical photos) where the original negatives are in the public domain but the human operator afterwards doing levels adjustment and dust removal creates a copyrightable derivative image. The real strength of AI is in a hybrid approach.
 
All those tantrums are pure jealousy. They're mad that they aren't as unique and talented as they think they are.
Of course it is jealousy. No human in this world will ever be able ingest and remix millions of images produced throughout centuries of history. Any skilled artist would understand this. Art is the human expression of sad limited mortals, and now we have a glimpse of what a limitless being could do. But should we push for artists to concede and diminish the human value in their work? And do you honestly believe people will rollover when their jobs are at risk? In fact it would be psychotic if they do. "Here, take my job that I spent countless hours of my life dedicated to honing the skills for!" said no one.

Do note that I have not yet taken a side in my original post, I was only explaining the artists' point of view which not many understand.
I noticed in your examples that the "copiers" have done so with paintings that are 2 decades or older. As per my post the "time sensitivity" is an important factor, nobody would be mad at the many parodies of the iconic terminator movie poster done now. Also if you think that people only paint "original" still life with physical medium then you're in for a ride.

The big masters of art through history did things nobody else did before, even if they copied others. Picasso and Van Gogh added their own style, which then was new and original on itself and became something new. That's why they are copied or inspire others until today.
We can credit Picasso and Van Gogh for their efforts, so that more people can learn about art and history. Man I wonder how I could ever credit the [fantasy][elf][4k][artstation trending][forest] pieces.

Anyhow. "Nobody cares". Just as what the twitter thread you included have said. But it only makes sense for many artists to be sad that things are this way.
 
This stuff always struck me as well-laundered copying and pasting.
That's algorithmically not what it is. I'm not even sure if something like that is possible, but the end result of such a thing would be an obvious collage.

Honestly? Good. AI artbros are the most obnoxious faggots I've ever encountered.

Way way way too many of those faggots are going and saying shit like "Now art is no longer gatekept behind artists" or you've got lazy asshole artists who use it to scam commissioners.

For example, two years ago I commissioned an artist to draw my sister's D&D character as a gift for her. When my turn for him to work on the piece came around, he worked on it for a month and then sent a DM saying that he was going to be moving to Brazil from Japan, and then vanished for a year and a half.

After I pestered him to either finish the art that I had paid for or give me a refund when he had returned, he said that the picture had been lost and he needed to restart it, and then a week later he gave some lazy ass AI art that looked nothing like the references my sister had drawn, and that he had lazily traced... But the dude was so retarded that he traced everything, including the extra fingers and other giveaways that it was AI generated.

So seriously, fuck AI art, fuck AI artbros, and fuck lazy artists.
You get screwed over by someone you commissioned a conventional artpiece from that disappeared for nearly two years and then only afterwards phoned in your commission, but you're mad at AI art?

You do realize that he was going to phone it in, either way, right?

Of course it is jealousy. No human in this world will ever be able ingest and remix millions of images produced throughout centuries of history. Any skilled artist would understand this. Art is the human expression of sad limited mortals, and now we have a glimpse of what a limitless being could do. But should we push for artists to concede and diminish the human value in their work? And do you honestly believe people will rollover when their jobs are at risk? In fact it would be psychotic if they do. "Here, take my job that I spent countless hours of my life dedicated to honing the skills for!" said no one.
If you're prompting for anything more than a single subject, it's absolutely uncanny and severely unnerving-- which is expected from a soulless automaton performing something that requires a soul. Artists that are threatened by it don't even know what it is, AND they suffer from being equally soulless to the point they consider a complex algorithm "competition".
 
Last edited:
But stable diffusion is open source... What are they going to do, confiscate everyones harddrives?

The important question is, if they win their suit, what can anyone who has it actually legally do with it that even matters? Stable Diffusion gets reduced from something that was going to revolutionize the graphic design industry to a little toy that software pirates can play with on their home PCs for fun.

Even researchers won't be able to use it, they'll have to rebuild it and ensure the training data is all used with permission.

That's algorithmically not what it is. I'm not even sure if something like that is possible, but the end result of such a thing would be an obvious collage.

The end result includes other mathematical operations - transformations, convolutions, etc, but it's still fundamentally derived from the training inputs. Sometimes it's really obvious what the training inputs were, sometimes it's not, but "it's not obvious how my software used your copyrighted image without permission to derive an output" is probably not enough on its own to settle whether or not I actually need your permission.

Example of obviousness:
Original photo:
1673892143329.png
AI asked to produce "Karl Marx leaning on a desk":
1673892165427.png

That just shows why these suits are a joke - the cardinal rule of art : DONUT STEEL! : as you said, is a gentleman's agreement at best, and such can never ever be enforced through a court. You have to sign an actual contract, and absent that, it comes down to the law, which already has plenty of cases that say it takes more than just similarity of character designs for a genre to prove someone stole your shit.

Among other things, the plaintiffs are alleging that DeviantArt providing their art to be used to tune art-generating software is a violation of a written contract. Virtually none of the images on the internet are free of written T&Cs, which is precisely the problem with doing any kind of bulk scraping of content and then doing well, pretty much anything other than academic research with it.

I know of an AI project where every single piece of data is provided under written permission by its owners to be used for that purpose. So it's not a trivial issue that is just now coming out of the blue. Plenty of people have known for years that this could get you in trouble, but the Stable Diffusion group just didn't give a shit and treated images from the internet like they were free for them use in any way they wanted without permission, and now they've got a lawsuit on their hands. So now we're going to find out if a court thinks copyright covers using images to build a model or not.
 
If the defence is clever, they'll demonstrate the dire negatives of Stable Diffusion live in court. I've played with Stable Diffusion for hours on end after finally getting it to run on PC. As a fun timesink, 10/10, it pushed everything else I was doing out of the way for about a week. As an artistic tool, though? 2/10, and that's being generous. For comparison I'd rate the Daz3D Dollhouse 8/10 if you're not adversed to stealing digital assets / not adversed to spending a small fortune, and things like the gelbooru search engine are 10/10, you can always instantly find fresh ideas there given very few prompts.

What Stable Diffusion is very good at:
(1) Boring yet beautiful landscapes devoid of characters and indeed character.
(2) Utterly cursed images. More amusing than genuinely disturbing, yet once in a while it will accidentally make something horrifying out of the most innocent prompt.

Problems with dream.py that aren't just bitching about the 8GB+ GPU hogging:
(1) It doesn't do humans well. Right off the bat, what is the point? If you're not trying to make a headshot portrait, or with a huge struggle a cowboy portrait (cut off just below the waist, preferably with hands hidden behind back or above head) forget it. There is no magical combination of prompt text that will trick this AI into understanding limbs. It shits itself over fingers and toes, and as often as not adds random arms, legs, hands etc. At best you accidentally luck out with a seed that generates something approximating human and can zero in from there; spoiler: the results aren't worth it.
(2) It is horribly obvious that much of the weighting is draw from social media. Wanna pump out an insta's worth of bland art-ho still life bullshit? dream.py has you covered. The other source is blatantly Artstation, which is better, but in terms of seeing something new, forget it. Much of that stuff might as well be created by AI to begin with anyway; not one new idea to be found, it's all horribly safely colouring in between the lines.
(3) Blandness is the core issue that kills dream.py artistically. It doesn't appear to understand verbs, so in the unlikely event you luck out and get it to generate a passable scene, you cannot simply edit the action. It also chooses the least energetic poses imaginable. Great if you want characters to mill about aimlessly, but beyond that it really struggles.
(4) 512 by 512 px is a joke, isn't it? How is that sending swim's poor GTX Titan into a tizzy? Oh, yeah, I wasn't going to bitch about this...

Problems with img2img.py that aren't identical to the above, or bitching about the computer dork pedantry of the parameter phrasing:
(1) What is perspective? This AI doesn't know. Unless the seed image is 90 degrees flat on, it misinterprets the visual prompt as having deformed body proportions. Often amusingly, to be fair, but not helpfully so.
(2) "You're supposed to feed it with stick figures..." yeah, no. Fuck off. The whole point of an AI enhanced image is to make the existing one better. Or remixed into the style of artist you feed the prompt. Whilst the AI makes an attempt - and indeed, the very best outputs are of this nature, try Hieronymus Bosch, Caravaggio, and even Larry Elmore which I was notably impressed at - it just doesn't work as intended. It either scratches at the input picture S=<0.5 or gives up attempting to interoplate entirely S>=0.7 which means you're pretty much stuck with S=0.6 and tough shit if it doesn't work with that.
(3) Related to perspective, this AI flattens and simplifies backgrounds unless you obsessively tell it not to do so, and even then it just ignores you most of the time. One of the hardest things to do is place characters into an environment where they appear naturalistic; img2img.py doesn't even bother trying. Even if you seed it with a beautifully balanced picture, nope, get that background out of here, and have some additional limbs while you're here.
(4) I can see the variance. After a short while I began to spot the patterns the AI pulls from, to the point of being able to predict them. Extending the output iterations eventually resulted in a loop, even.

As it stands, it was fun, often amusing, but I'm not inclined to go back to it. There is no danger this is going to replace anyone as an artisan, let alone an artist, even with a custom ultra user-friendly interface. I'd be very interested to see the hidden base prompts novelai uses to lock down output images into a pleasingly generic genre output, but beyond that curiosity, I probably won't use Stable Diffusion again.

Take prompt suggestions from the court regarding a scene with human figures interacting. Stable Diffusion will shit itself live on the court monitors, and case dismissed.
It’s not just human figures AI is shit at, it’s trash at architecture. It looks fine at a glance, but then you’ll notice doors in the middle of walls, arches of differing heights on the same level and all sorts of shit that’s just wrong. The building blocks are there, but it doesn’t know how to arrange them.
 
You get screwed over by someone you commissioned a conventional artpiece from that disappeared for nearly two years and then only afterwards phoned in your commission, but you're mad at AI art?

You do realize that he was going to phone it in, either way, right?

Him being a scumbag is irrelevant and just a symptom of the greater issue.
 
The end result includes other mathematical operations - transformations, convolutions, etc, but it's still fundamentally derived from the training inputs.
By your own description, it's literally "transformative".

The only difference between Stable Diffusion and (most) artists is that the former doesn't have a soul, but that's it.

also I can't stand that damned picture, the fingers are Lovecraftian abominations, the face is wrong, it's just straight up icky-- if you're worried about this coming for your job, you have no soul

Him being a scumbag is irrelevant and just a symptom of the greater issue.
The "greater issue" is that you got scammed by a guy that ghosted you for nearly two years and then phoned in your work.

He could have done that without Stable Diffusion. Stable Diffusion didn't make him a flake.
 
The "greater issue" is that you got scammed.
Fair enough about me being scammed, but no, the greater issue is that a lot of scumbags are using it to scam their clients. So far only two types of people use Stable Diffusion and call it art:

1) Faggots who couldn't cut it as artists / never bothered to try / hate artists for whatever reason - The kind of people sobbing about how "you don't understand it man" in this thread
2)Lazy asshole artists who think that its a good replacement for the work they had done in the past.
 
Last edited:
By your own description, it's literally "transformative".

Moving something three pixels to the right is literally a transformation. Applying a blur filter is literally a transformation. The problem is, applying linear transformations, Gaussian filters, convolutions, eigendecompositions, and other mathematical transformations to electronic data given to you under license does not absolve you of having to comply with the terms of the license. That's how contracts work. Lots of images on Flickr and DeviantArt specify "no derivative works" & require attribution. A judge may very well determine that the StableDiffusion AI is itself a derivative work, and is therefore a violation of the CC Attribution-NoDerivs 2.0 license. Maybe he won't, though, maybe he'll rule that creation of an AI model doesn't have to comply with any terms under which you obtain data.

However, if a judge rules that StableDiffusion is not a derivative work, nor can the output of the model be considered derivative works, we'll see an updated Creative Commons license with an additional clause barring the use of the image for training an AI. And then it won't be ambiguous at all.

All that's ambiguous right now is whether licenses that were written before people really knew about the possibility of them being used to fit billion-parameter models can be interpreted as potentially baring this use of the licensed content. If it were obvious, there wouldn't be two open lawsuits right now concerning exactly this.

The only difference between Stable Diffusion and (most) artists is that the former doesn't have a soul, but that's it.

No, actually, your brain doesn't internally have a fancy nonlinear function that's fitted to a bunch of data points. They don't work anything at all the same.
 
Last edited:
  • Informative
Reactions: IAmNotAlpharius
Fair enough about me being scammed, but no, the greater issue is that a lot of scumbags are using it to scam their clients.
This is why you look at their portfolios and arrange a way to not pay in full before you see a sketch of what they're intending.

Applying linear transformations, Gaussian filters, convolutions, eigendecompositions, and other mathematical transformations
You keep saying the word, but you keep denying my point.

Or, I don't know, do you think if you got a CAD model under NDA, you could distribute it on the internet after transforming it a bit?
>take a base and edit it a bit
Even now, that's not how Stable Diffusion works.

No, actually, your brain doesn't internally have a fancy nonlinear function that's fitted to a bunch of data points.
The "fancy nonlinear function that's fitted to a bunch of data points" is an inferior mathematical representation of natural human creative processes-- that's my point.
 
Back