AI art - lawsuit explanation

[Adam Thornton 0]

If any of you have been following the AI art debate (the industrial-scale copyright infringement of art on the internet), a new 16 minute video has been posted by Corridor Crew on YouTube.

The video explains the new class action lawsuit that has been filed in the USA.

You can watch it here: https://www.youtube.com/watch?v=gv9cdTh8cUo&ab_channel=CorridorCrew

[ksfabian]

@Adam-Thornton-0 Wow! Thanks for sharing!!

[Melissa_Bailey]

@Adam-Thornton-0 just watched it last night -- very interesting!

[davidhohn]

@Adam-Thornton-0 Agree with others -- an interesting video.

The following might seem like a "nit-pick" but it is an important distinction. What's happening in AI generated images isn't "plagiarism" but rather "copyright infringement"

Check out this article the defines the differences:
https://copyrightalliance.org/differences-copyright-infringement-plagiarism/#:~:text=While both plagiarism and copyright,illegal%2C while plagiarism is not.

TLDR: Both Copyright and Plagiarism involve using ideas or creativity that isn't your own. But plagiarism isn't illegal. Copyright infringement is.

[davidhohn]

After watching the video I think the lawyer and the plaintiffs are making a mistake. Both appear to be shifting away from the fundamental copyright infringement made during the data scraping stage, and instead focusing on to the "latent images" and the "derivative images" made within the AI algorithm.

I say this because I don't think "derivative images" will be a convincing argument for the courts. After watching copyright cases over the years, courts have become (in my opinion) far too lenient when applying the concept of "transformative use" to an infringement which was eventually deemed permissible under the "fair use defense"

Based on what has been considered "fair use due to transformative use" in the past (Cariou V Prince being a particularly blatant example https://en.wikipedia.org/wiki/Cariou_v._Prince) I would expect courts to allow the AI Images to continue.

This is as compared to the Google V Authors Guild (https://en.wikipedia.org/wiki/Authors_Guild,_Inc._v._Google,_Inc.) in which huge data scraping occurred BUT it didn't significantly impact the authors or publishing business.

I don't think that will be the case with AI Images. The industrial-scale copyright infringements here WILL directly affect artists, image makers and other creatives.

[Adam Thornton 0]

@davidhohn Aha! Good point. Thanks for saying so, David. I've edited my post now.

[davidhohn]

@davidhohn Update to my post above:
I wrote that I think the plaintiffs are making a mistake (based on the video) by focusing on latent images in the AI algorithm.

I just listened to Sarah Anderson (one of the plaintiff's in the class action lawsuit) being interviewed on the tech podcast HardFork
https://open.spotify.com/episode/7j23TCIS7aDlvEaAHuZLwo

and she doesn't appear to be focusing on that at all. Instead the focus (based on her interview) is on the copyright infringement in the datasets. This is the stronger argument in my opinion.

So my takeaway: The Corridor Crew video above is (sort of) right -- but also kind of wrong.

What you are seeing is his opinion and may not reflect the arguments in the actual class action lawsuit.

[carlianne]

@davidhohn I'm so glad that they are focusing on the copyright infringement in the dataset as well! I agree that I don't think this case will win if they focus on the art being generated, the use of copyrighted images in the dataset is a much clearer violation of use.

I am not sure though if there will be an issue if the dataset itself is open source and therefore considered free use since it's not directly monetized? The companies using the dataset can claim it's not their fault (they didn't make the dataset) and the creators of the dataset can claim they aren't using it for profit. Unless I'm mistaken on how the dataset is set up.

Interested in your thoughts

[davidhohn]

@carlianne said in AI art - lawsuit explanation:

I am not sure though if there will be an issue if the dataset itself is open source and therefore considered free use since it's not directly monetized?

Do you mean "Fair Use" rather than "free use"?
Fair use is defense argument used in a case in which a copyright violation HAS taken place but the infringer feels it should be allowed.

There are a number of factors go into determining if in fact a Fair Use defense excuses the "prima facie" (i.e. accepted as correct) infringement.

One of those is monetary impact. In some cases this means the infringer is making money from the copyright infringements, but a "fair use defense" argument can be stuck down even if the infringer isn't making money themselves, but is preventing the copyright owner from generating income from their own work.

An example would be an artist sells prints of their work.
An infringer makes prints of the artists work and hands them out for free. Now, the people who would normally pay the artist for the prints don't anymore because they are getting prints for free.

The companies using the dataset can claim it's not their fault (they didn't make the dataset) and the creators of the dataset can claim they aren't using it for profit. Unless I'm mistaken on how the dataset is set up.

I'd have to look into how the AI Image companies and Dataset companies are set up. My impression is that they are very closely linked. But regardless, the AI companies know very well that the datasets contain copyrighted images (they actively promote this fact to their users!) In my opinion making them as culpable for infringements as the dataset companies.

[Adam Thornton 0]

@davidhohn I nominate David to represent SVS members in court! You clearly know so much about this stuff!

[Geoffrey Mégardon]

Could adding a license to our artwork help preventing further undesired use of our work? We could have a Creative Common license modified so that it states clearly that we do not allow our artworks to be used in AI training or NFT markets.

I think, one problem that AI training is highlighting is that artists distribute their artwork without End User License Agreements (EULA), we just rely on copyright laws which aren’t that strong it seems.

Of course, there is also the question of how to link an EULA to an artwork

[Melissa_Bailey]

Adding to the discussion...
Another lawyer on YouTube has weighed in on the A.I. art lawsuit: A.I. Versus The Law by Legal Eagle. One of his areas of expertise is copyright law.