A class action copyright infringement lawsuit was filed against Stability AI (Stable Diffusion), Midjourney, and DeviantArt, Inc. (DreamUp) on behalf of artists who allege their works were used to train #AI generative art algorithms.
THREAD 🧵 to come 👇 once I read!
THREAD 🧵 to come 👇 once I read!
It was only a matter of time before the issue of copyright infringement by AI apps was tested by the Courts.
In our Nov 2022 @Law360 article, “A Look at Future AI Questions” we asked whether & when AI generated work is an infringing derivative work:
g2bswiggins.wpenginepowered.com/wp-content/upl…
/2
In our Nov 2022 @Law360 article, “A Look at Future AI Questions” we asked whether & when AI generated work is an infringing derivative work:
g2bswiggins.wpenginepowered.com/wp-content/upl…
/2
The class action plaintiffs in this ND Cal suit, which is - as far as I know - the first asserting these types of claims against generative art AI apps are artists Sarah Andersen, Kelly McKernan, and Karla Ortiz.
/3
/3
The named Defendants in this class action IP lawsuit are Stable AI, the company that created Stable Diffusion, as well as MidJourney and Deviant Art (DreamUp app). All of these are AI image generation apps, similar to DALL-E.
/4
/4
The causes of action in the Stable Diffusion Class Action include claims of direct copyright infringement, vicarious infringement, DMCA violations (for the removal of copyright management info), California right of publicity and related unfair competition claims.
/5
/5
In the main, the Complaint alleges that the defendants AI apps use the works of the plaintiffs without permission to train AI through machine learning to create new AI-generated works, some of which constitute infringing derivative works.
/6
/6
The 46-page Complaint, which includes a ToC, leads off stating that “AI IMAGE GENERATORS ARE 21ST-CENTURY COLLAGE TOOLS THAT VIOLATE THE RIGHTS OF MILLIONS OF ARTISTS.” It details how Stable Diffusion operates, as well as the assembly / use of massive training data sets.
/7
/7
In the section containing its thesis, the Complaint explains the crux of its claim:
The allegations here focus on the ability of some generative AI apps to create art “in the style of” a particular artist based on art by that artist included in its training data set.
/8
The allegations here focus on the ability of some generative AI apps to create art “in the style of” a particular artist based on art by that artist included in its training data set.
/8
It’s interesting that this “in the style of” feature appears to be the focus of the claims.
This is narrower than what may be possible to allege in terms of copyright infringement/derivative work.
There are also questions about whether it is copyright infringement to do so.
/9
This is narrower than what may be possible to allege in terms of copyright infringement/derivative work.
There are also questions about whether it is copyright infringement to do so.
/9
Also interesting is the allegation of what the derivative work is:
“ Derivative Work”…refers to the output of AI Image Products as well as the AI Image Products themselves—which contain compressed copies of the copyrighted works they were trained on.”
/10
“ Derivative Work”…refers to the output of AI Image Products as well as the AI Image Products themselves—which contain compressed copies of the copyrighted works they were trained on.”
/10
In describing the set of plaintiffs, the Complaint notes that each has a number of registered copyrighted works included in the Training Data.
/11
/11
The support for citing the number of works included in the AI Training Data for each of the plaintiff artists is:
haveibeentrained.com
/12
haveibeentrained.com
/12
Because this is a Class Action, the Complaint also includes class definitions and FRCP Rule 23-related allegations.
(Aside from substantive copyright related defenses, compliance with the class action requirements is another route to defend against these claims)
/13
(Aside from substantive copyright related defenses, compliance with the class action requirements is another route to defend against these claims)
/13
The factual allegations are long and detailed.
The Complaint notes they offering these apps for free rather than for a paid license had contributed to their rapid adoption.
Also noteworthy is the detailed description of how Stable Diffusion works.
/14
The Complaint notes they offering these apps for free rather than for a paid license had contributed to their rapid adoption.
Also noteworthy is the detailed description of how Stable Diffusion works.
/14
Significantly, the Complaint alleged that the very point of the complex mathematics behind Stable Diffusion is “to reconstruct copies of the training data with maximum accuracy and fidelity to the Training Image. It is meant to be a duplicate.”
/15
/15
The other bottom line point made from this technical description is allegation that the resulting images are derivatives from the original copyrighted images.
/16
/16
Another point that the Complaint emphasizes about the operation of Stable Diffusion is the use of highly complex interpolation techniques and “conditioning” where the resulting image may not look like a copy of any single copyrighted source image.
/17
/17
Nonetheless, the Complaint alleges, “resulting image is necessarily a derivative work, because it is generated
exclusively from a combination of the conditioning data and the latent images, all of which are
copies of copyrighted images.”
/18
exclusively from a combination of the conditioning data and the latent images, all of which are
copies of copyrighted images.”
/18
It turns next to where the Training Data comes from, alleging “Though the rapid success of Stable Diffusion has been partly reliant on a great leap forward in computer science, it has been even more reliant on a great leap forward in appropriating copyrighted images.”
/19
/19
The Training Data is from a project called Large-Scale AInOpen Network (LAION), which the Complaint alleged pulls data from copyrighted data sets like Getty Images, Shutterstock, etc.
/20
/20
I will turn next to a review of the specific Claims here, starting with the copyright infringement claims.
/21
/21
But first, I need to 🛌 😴 💤 More tomorrow!!!
. @threadreaderapp unroll
Here we are on the morrow, continuing.
Before we get into reviewing the claims, a few overview comments. Because it’s already apparent that this issue (and this lawsuit) is a polarizing one.
/22
Before we get into reviewing the claims, a few overview comments. Because it’s already apparent that this issue (and this lawsuit) is a polarizing one.
/22
On the one hand, AI supporters have said you can’t fight against AI and we need to embrace it as a tool for innovation. On the other hand, many artists are saying everything these apps do is theft.
/23
/23
In my view, the extreme poles laid out above aren’t really helpful here. Of course AI is here to stay and I personally believe best approaches are to embrace it and use it. Though I think this will require big changes to how we educate, work and live.
/24
/24
But it may well be that original artists and creators are owed some compensation and some control over their work.
Depending on the facts and how this complex issue is analyzed and decided under the law.
/25
Depending on the facts and how this complex issue is analyzed and decided under the law.
/25
This also highlights the value of data sets.
I’ve read of AI learning by watching terabytes of videos of humans doing tasks on YouTube.
These data sets are going to be VERY valuable.
And this dispute is about who should profit and share@profits from its monetization.
/26
I’ve read of AI learning by watching terabytes of videos of humans doing tasks on YouTube.
These data sets are going to be VERY valuable.
And this dispute is about who should profit and share@profits from its monetization.
/26
I’ve also seen the “If there has to be a significant portion of a music arrangement in another to call it plagiarism, I don't see how these lawsuits will proceed” point. Be made. I don’t think that’s helpful here either.
/27
/27
Why?
Copyright infringement requires copying. In the music cases there is often not direct evidence of copying. So having more than a similarity of small bits can be circumstantial evidence of copying. And without that no copying.
/28
Copyright infringement requires copying. In the music cases there is often not direct evidence of copying. So having more than a similarity of small bits can be circumstantial evidence of copying. And without that no copying.
/28
Here. It’s likely the use of the copyrighted art in the data set (copying) will be undisputed.
It’s a “derivative work” argument: creating a “derivative work” is an act of infringement.
/29
It’s a “derivative work” argument: creating a “derivative work” is an act of infringement.
/29
In this theory of infringement you use the copyrighted work as your source, the output doesn’t have to be substantially similar.
The legal question is whether the output of the AI *is* a “derivative work” under the law. (And whether defenses like fair use apply)
/30
The legal question is whether the output of the AI *is* a “derivative work” under the law. (And whether defenses like fair use apply)
/30
So the two big legal themes teed up by these allegations are (a) whether and when the process of training AI off of copyrighted materials is infringing; and (b) when and whether I can use your data, monetize it and profit from it without sharing compensation.
/31
/31
On to the specific legal Claims alleged here:
1. Direct Copyright Infringement
Based on the unlicensed use of copyrighted images as source images to train AI by the makers of the AI apps. Focused on creation of derivative works & copying of original work in the data set.
/32
1. Direct Copyright Infringement
Based on the unlicensed use of copyrighted images as source images to train AI by the makers of the AI apps. Focused on creation of derivative works & copying of original work in the data set.
/32
2. Vicarious Copyright Infringement
This focused on individuals using the text prompts to cause the AI apps to create “fakes” that are passed off as the original artists works.
This is a much narrower claim.
/33
This focused on individuals using the text prompts to cause the AI apps to create “fakes” that are passed off as the original artists works.
This is a much narrower claim.
/33
3. Violation of the DMCA
This focuses on the removal of certain copyright management information associated with the copyrighted works, like name, title, artist and copyright notices.
/34
This focuses on the removal of certain copyright management information associated with the copyrighted works, like name, title, artist and copyright notices.
/34
4. Violation of California Right of Publicity &
5. Violation of Common Law Right of Publicity
This focuses on the use of plaintiff artists names to advertise the “in the style of” feature.
/35
5. Violation of Common Law Right of Publicity
This focuses on the use of plaintiff artists names to advertise the “in the style of” feature.
/35
6. Unfair Competition under California Law - based on the alleged copyright infringement and violations of the DMCA
/36
/36
7. Breach of Contract
Against DeviantArt for violations of their own privacy policy
/37
Against DeviantArt for violations of their own privacy policy
/37
Wheeeeeee. Well that’s all (for now) folks!! Now we can all get on with our Sundays.
Should be an interesting case to follow for those interested in #AI IP law, art and business!
/38
/END
Should be an interesting case to follow for those interested in #AI IP law, art and business!
/38
/END
. @threadreaderapp unroll
• • •
Missing some Tweet in this thread? You can try to
force a refresh
