Judge Pares Down Artists' AI Copyright Lawsuit Against Midjourney, Stability AI 41
Blake Brittain reports via Reuters: A judge in California federal court on Monday trimmed a lawsuit by visual artists who accuse Stability AI, Midjourney and DeviantArt of misusing their copyrighted work in connection with the companies' generative artificial intelligence systems. U.S. District Judge William Orrick dismissed some claims from the proposed class action brought by Sarah Andersen, Kelly McKernan and Karla Ortiz, including all of the allegations against Midjourney and DeviantArt. The judge said (PDF) the artists could file an amended complaint against the two companies, whose systems utilize Stability's Stable Diffusion text-to-image technology.
Orrick also dismissed McKernan and Ortiz's copyright infringement claims entirely. The judge allowed Andersen to continue pursuing her key claim that Stability's alleged use of her work to train Stable Diffusion infringed her copyrights. "Even Stability recognizes that determination of the truth of these allegations -- whether copying in violation of the Copyright Act occurred in the context of training Stable Diffusion or occurs when Stable Diffusion is run -- cannot be resolved at this juncture," Orrick said.
Orrick agreed with all three companies that the images the systems actually created likely did not infringe the artists' copyrights. He allowed the claims to be amended but said he was "not convinced" that allegations based on the systems' output could survive without showing that the images were substantially similar to the artists' work. The judge also dismissed other claims from the artists, including that the companies violated their publicity rights and competed with them unfairly, with permission to refile. Orrick dismissed McKernan and Ortiz's copyright claims because they had not registered their images with the U.S. Copyright Office, a requirement for bringing a copyright lawsuit.
Orrick also dismissed McKernan and Ortiz's copyright infringement claims entirely. The judge allowed Andersen to continue pursuing her key claim that Stability's alleged use of her work to train Stable Diffusion infringed her copyrights. "Even Stability recognizes that determination of the truth of these allegations -- whether copying in violation of the Copyright Act occurred in the context of training Stable Diffusion or occurs when Stable Diffusion is run -- cannot be resolved at this juncture," Orrick said.
Orrick agreed with all three companies that the images the systems actually created likely did not infringe the artists' copyrights. He allowed the claims to be amended but said he was "not convinced" that allegations based on the systems' output could survive without showing that the images were substantially similar to the artists' work. The judge also dismissed other claims from the artists, including that the companies violated their publicity rights and competed with them unfairly, with permission to refile. Orrick dismissed McKernan and Ortiz's copyright claims because they had not registered their images with the U.S. Copyright Office, a requirement for bringing a copyright lawsuit.
One Step Closer (Score:2)
That means we're one step closer to freeing all code (and basically everything) from copyright. If we can feed an AI a pile of code, and get it to reproduce that code in its entirety, it then isn't a copyright violation (and in fact cannot be copyrighted). The death of copyright is finally upon us! Rejoice! Rejoice!
Re: (Score:2, Informative)
No, that's the whole point, copyright is about making copies. The artists do not have proof of infringement because they haven't found a generated picture that would be considered a copy if an another artist made it.
For code, if you can as you say "reproduce that code in its entirety" then that would clearly be infringing, case closed.
Re: One Step Closer (Score:1)
For the first time in a long time, the cogent, relevant comments on an IP story are outnumbering the inane, but you still had to come along and offer "copyright is ab
Re: (Score:2)
Re: (Score:2)
Moreover, the defense to infringement via derivative work is fair use, which requires "transformative use." Andy Warhol Foundation v. Goldsmith makes it clear that "the degree of transformation required to ma
Re: (Score:1)
Unless you have never visited Earth, anything you produce would be considered a derivative work, since you are influenced by everything you have ever seen and heard.
If I paint a picture of the Golden Gate Bridge, which I have never seen in person, that would be a "derivative work" because it's based on what I saw in a tv show, movie, photograph or read about in a book - likely a combination of all of the above.
The question that needs to be answered is what percentage of derivativeness is allowable before It
Re: (Score:2)
"Copying" single words from books to create a new novel is not copyright infringement. Arranging notes that exist in other music in a different order to create a new music piece is not copyright infringement. If it was you could only create new books or new music if you invented writing or created a musical instrument in a clean room environment, without ever reading anything or hearing music. It's a quick step from saying that training a computer is copyright infringement to saying that learning is copyrig
Re: (Score:2)
Re: One Step Closer (Score:2)
Re: (Score:2)
Re: One Step Closer (Score:4, Informative)
Re: (Score:2)
Re: (Score:1)
Berne Convention? (Score:2)
Re:Berne Convention? (Score:4, Informative)
US courts still require the formality of registering a copyright before filing suit. This establishes prima facie evidence of authorship and ownership in a form that a judge can read. What changed with Berne Convention implementation is that the copyright owner can still sue for damages even if the work is registered after the infringement occurs.
Re: Berne Convention? (Score:2)
Re: (Score:2)
My mom, a lawyer, frequently wins cases due to this, she does her due diligence, the other lawyer, does the bare minimum to get by, and my mom exploits their laziness to her client's benefit. Truth. I spent 6 years clerking for her and it was amazing to me how slipshod and lazy most lawyers work seems to be.
Transformative use is fair use (Score:5, Informative)
- Campbell v. Acuff-Rose Music, Inc. 1994
- Blanch v. Koons 2006
- Mannie Garcia v. Shepard Fairey 2011.
- Cariou v. Prince 2013
Sarah Andersen, Kelly McKernan and Karla Ortiz are very unlikely to win.
Re: (Score:2)
Re: (Score:3)
Copying a style is not transformative. While AI is very good at copying a style it rarely spits out something resembling a piece of copyrighted art from the training set unless the user specifically askes for it, leaving the point that copyright claims are based on the individual outputs from AI rather than the concept as a whole.
These people are suing tools not copyrighted works.
Copying (Score:2)
It's nearly impossible for the exact work to be recreated by an AI image generator, as the image only exists as a set of weights, or probabilities, in-between pixels. The odds of the neural network recreating the exact image from the initial noise seed are astronomical.
Re: (Score:2)
No one said "exact". We're discussing when something is or isn't transformative. You absolutely can generate works which would be considered to not be transformative enough with the correct AI inputs.
Re: (Score:2)
Well, it's a great way to kill Linux and open-source if AI is fair use.
It implies using an AI to suck up all of the GNU, Linux, and other open-source projects you can use it to spit out your own Linux-like OS as fair use. And then license however you wa
Re: (Score:3)
Uhhh, that's quite literally the whole point that lets Linux exist in the first place?
Linux is an re-implementation of UNIX / POSIX, doing the same things with different underlying code .
Anyone is free to re-write and re-implement from scratch the GNU userland, or even the Linux kernel in code that is not exactly copied from the original source. They can even *gasp* have it be closed source. Or BSD licensed. Or MIT licensed. It's up to the person / team / company that rewrote the NEW code how they want t
Re: (Score:2)
AI models cannot understand intent of things like a semaphore and how multiple code elements use it to interact. Sure it could generate some example code on how to use the semaphore,
Re: (Score:2)
- Harper & Row Publishers, Inc. v. Nation Enterprises 1985
Hinged on public interest and actually ruled against it being fair use in this specific case.
Your first citation and it already weakens your case, I'm not going to bother with the rest and going to just assume they are equally irrelevant.
If only they'd copied RIAA copyright songs instead (Score:2, Insightful)
What DeviantArt, etc... did was certainly unethical, regardless of the legality. They violated the trust of their users to build a product which won't benefit artists, and won't produce copyrightable work.
But I remember when the filesharing lawsuits were going on that the RIAA and MPAA were prevailing in lawsuits even when they couldn't prove that copying had occurred. Making available copyrighted works was enough to get a verdict.
Now, DeviantArt and Midjourney aren't even denying they copied the work
Re: (Score:3)
Making available copyrighted works was enough to get a verdict.
*Copyright infringing. If it were merely being copyrighted that made distributing it bad then you wouldn'ty be able to share that Creative Commons licensed song you or someone else made, since it'd still be a "copyrighted work" due to copyright in the US being automatic upon an eligible work being put into a fixed tangible medium.
/pedanticMode
Re: If only they'd copied RIAA copyright songs ins (Score:1)
There are AIs for music as well. They are used for making songs sung by singers who never sung them. They are used to make songs sung by people who don't sing. They are also used to generate ambient background music.
So far, the RIAA has not sued any of them. And they know why. They have no case. (Granted, they didn't have any legal grounds against ThePirateBay either and that didn't stop them from "winning".) Parody is protected (ever since Buster Keaton lost against Disney), and original works are o
Re: (Score:3)
Re: (Score:2)
Artists study other artist's works and styles and then produce new and original works of art. What makes AI doing this different and legally actionable?
Someone please enlighten me? thanks.
Re: (Score:2)
The fact that they a.) copied the artists work without permission, and b.) it negatively affects the artists' ability to profit from their work.
What everyone seems to miss is that there are already artists whose works have been so convincingly copied by AI that they're being accused of using AI for their own work! They have been banned for submitting "AI" generated works because the AI generator output matches their originals so well.
The difference between AI and traditional artists is that traditiona
Re: (Score:2)
The difference between AI and traditional artists is that traditional artists, generally speaking, cannot create an (comparatively) unlimited number of works in another artist's style
Sure they can, just not as quickly.
Artists in general have ethics, and AI doesn't.
AI is not a social agent. Whether AI has ethics is about as meaningful a question as whether cutlery has a sense of direction.
Also: Good artists copy, great arists steal. -- <del>Picasso</del> Banksy
they a.) copied the artists work without permission
You didn't get permissions to read this post. It negatively affects my ability to profit from it.
To get an AI to copy the style of an artist, you first have to train the AI in the style of that artist. In addition to teaching the AI what a hand looks like, you a
Re: (Score:3)
Seems to me artists are the best positioned to make good use of AI. Regular people don't know how to generate with AI beyond writing a prompt. But AI art is extremely complex, with LoRAs, embeddings, depth maps, noise samplers, masking, compositing, etc etc
Re: (Score:2)
To someone trained in the visual arts, AI feels like trying to finger paint with mittens on.
The issue is that it's far more work to get the AI to produce specifically what the artist had in mind than to just paint it himself. For most artists, reproducing a particular style is the easy part; the difficult part is the inspiration, i.e. figuring out what we should paint.
The last sentence says it all... (Score:3)
Orrick dismissed McKernan and Ortiz's copyright claims because they had not registered their images with the U.S. Copyright Office, a requirement for bringing a copyright lawsuit.
Re: (Score:2)
I wonder if that has ever made it to the supreme court - That you have to take an extra step seems kind of, extra.
You don't have to register a copyright to have ownership/inherent copyright of a work you create. I don't think they are arguing that these artists are not the creators.
Maybe they can register after the fact, now that they need protection? But, I would assume their lawyers would have done that (or are now doing that) if that is allowed.
Re: (Score:2)
you cannot sue in the USA until after copyright registration:
https://copyrightalliance.org/... [copyrightalliance.org]
Full Question: Why should I register my work if copyright protection is automatic?
Answer: Since copyright protection is automatic from the moment a work is created, registration is not required in order to protect your work. However, there are numerous benefits to registration and therefore it is highly recommended, if feasible. These benefits include:
1.Bringing an Infringement Action: It enables you to file a laws