Supreme Court Rules Against Andy Warhol in Prince Copyright Case: What this Means for AI Art

The Supreme Court just ruled that Andy Warhol’s portraits of Prince infringed upon the copyright held by the photographer, Lynn Goldsmith, who took the original photo. In this video, we’ll review highlights from the Supreme Court’s opinion and discuss how it might affect the legality of content created with generative AI tools.
BECOME A CHANNEL MEMBER: / thelegalpad
Chapters:
00:00 - Introduction
00:43 - Relevant Facts
03:25 - District Court's Decision
03:59 - 2nd Circuit Court of Appeals' Decision
04:26 - Supreme Court's Majority Decision
06:13 - Dissenting Opinion
07:30 - Implications on AI-Generated Content
08:21 - Transformative Use Analysis
08:51 - Will the Court's Decision Open the Floodgates For Copyright Infringement Claims?
In the district court, the Andy Warhol Foundation argued that Warhol's artwork constituted "fair use" because Warhol sufficiently transformed the original photo of Prince and thereby gave it new meaning. The judge agreed that Warhol’s Prince artworks transformed the photos of Prince from a "vulnerable, uncomfortable person" into "an iconic, larger-than-life figure.”
Photographer Lynn Goldsmith appealed the district judge’s decision to the Second Circuit. On appeal, the court overturned the district judge’s ruling, finding that the district court had placed too much weight on the subjective meaning of the work.
The Warhol Foundation then appealed to the U.S. Supreme Court, which agreed to decide the case. In a 7-2 decision, the Supreme Court ultimately held that Warhol’s use of Lynn Goldsmith’s photo of Prince without her permission was copyright infringement. Justice Sotomayor wrote the majority. We'll review highlights from the Supreme Court's majority opinion in this case.
Justice Kagan and Chief Justice Roberts dissented from the rest of the Justices. They felt that Warhol's Prince artworks were sufficiently transformative of the original photograph to qualify as "fair use" ratyher than copyright infringement. Justice Kagan, who wrote the dissenting opinion, was very critical of the majority’s holding. We'll review highlights from Justice Kagan's dissenting opinion in this case.
Next, I'll share my thoughts on the implications of the Court’s opinion in this case on the use of generative AI tools.
We’ve seen generative AI tools that convert text to images, like Stable Diffusion and Dall-E. We’ve seen examples of people telling such AI tools to create artwork in the style of a particular artist, such as Picasso, Van Gogh, or Andy Warhol.
Does the Supreme Court’s opinion makes these types of works less likely to qualify as fair use? How much of the original work must a content creator change in order for their new work to qualify as "transformative" for purposes of fair use? Will this case open the floodgates of litigation to people claiming that AI-generated images used their copyrighted work without permission? Watch to find out.
SUBSCRIBE: / @legalpadlaw
WAGONER LAW FIRM: wagonerlawfirmpllc.com/
#copyright #supremecourt #prince
NOTICE AND DISCLAIMER: The information in this video and on this Channel is not legal advice or any other form of professional advice. It's for informational/educational purposes only. This information is not intended to create, and receipt or viewing of it does not create, an attorney-client relationship. Nothing in this video or on this channel should be taken as legal advice or any other form of professional advice for any individual claim, case, or situation. You should seek advice from an attorney or other professional regarding your specific situation. Contacting the Wagoner Law Firm does not create an attorney-client relationship, nor does your receipt of an initial case evaluation. The only way to create an attorney-client relationship with the Wagoner Law Firm is through a mutual written agreement, which you will be asked to sign. Please do not send any confidential information to the Wagoner Law Firm until an attorney-client relationship has been established. This description may contain one or more affiliate links, meaning that I will earn a financial benefit if you click through and use an affiliate link (at no additional cost to you).

[@hudekhoustonartist]

This situation creates quite a conundrum for the creative process of all artists. The only way to know whether a particular transformative use is a fair use is to get sued and let judges decide.

[@tanveerahmed2327]

I recently came across this case in a question during a course and I had to decide which party I favored and why. I went in favor of Goldsmith and so did the Supreme Court! Glad to see that I was right in my assessment! P. S. I passed that course with flying colors ❤

[@damnitstroubleman]

Outta all the arts and disciplines, visual art gotta be the hardest to design concrete laws around. It's a medium that's 50% artist intention and 50% viewer interpretation.

[@ThePigman6]

It is critical to get the facts correct. False information does not lead to sound analysis. Patty Smith?

[@seanrowe6973]

Thanks for your overview of this SCOTUS case. Super helpful as I'm producing a Duke Law Podcast episode on this and hadn't considered the AI angle. Thanks again!

[@LegalPadLaw]

Of course. Thanks for commenting. Send me a link to your podcast once it's published. I'm fascinated by the collision between copyright law and generative AI tools that's happening right now.

[@craigshields8806]

It's a bit confusing because you don't always say "foundation" but are you saying that the artist Andy Warhol got permission from the photographer in '84 to make that work, then his foundation gave license to Condé Nast for a slightly different work in 2016 (he died in '87) but never got permission from the artist? Is she also suing Condé Nast? Thx

[@richarcher1626]

The foundation was paid by Conde Nast to use the Prince image. The photographer initially licensed the photo to be used for an article in Vanity Fair, in which Vanity Fair had Warhol create a silkscreen based on the photo. That license was a one-time use for that article. Conde Nast didn't sell work based on her photo without a license - the warhol foundation did that.

[@2short1968]

👍👍👍👍👍..."Pops"#truschoolsports🥊🥊

[@jaypaladin-havesmartswilll5508]

There will have to be a new US addendum to copyright laws but our politicians are so divisive with their simple and ugly rhetoric that I don't think we will see a new law or addendum any time soon.

[@Unknown17]

Even in such places as our church choir, one often has to get permission of a composer to perform the work. Sometimes they will ask for as little as a dollar. All Warhol did was tint the damn photo. To me, it's pretty easy to see he ripped her off. I admire Warhol, but to fail to give credit to the photographer here was crass, and he--or the foundation--knew it was legally risky, too. He wouldn't have been pleased if someone slightly altered one of his works and published the image. Professional courtesy--if nothing else--demands permission be sought. I'm with the majority here.

[@lukeknowles5700]

You display a massive ignorance about what Warhol actually did when you claim "all he did was tint the damn photo." All the subtle shading differences that were in the photo were removed, the background color was applied to all the various shade colors of the face and eyes and a new halo effect of teal, blue and green was added around the hair line. Look at the right eye shading, too. Your failure to understand the actual amount of hard work Warhol did to transform the photograph is numbingly ignorant.

[@Unknown17]

@@lukeknowles5700 Apparently the Supreme Court agrees with me that copyright was infringed. This work was not changed enough to constitute an original work of art. If I buy a house, you can't put it up for sale because you added two bushes, repainted the garage, and put up a mailbox. Even though it was a lot of work for you to change it in this way, the underlying structure still belongs to me. But I suppose the justices were just "numbingly ignorant," too, and didn't get to hear arguments from both sides, poor fools that they are. So, um, you're wrong, and you can shut up now.

[@lukeknowles5700]

@@Unknown17 Why do you say "apparently"? At lease learn English since you know nothing of art.

[@LegalPadLaw]

Thanks for the comment. I agree that, at the very least, Vanity Fair should have credited the photographer for the underlying photograph when they used it on the cover of the magazine. On the other hand, I feel like Warhol did in fact sufficiently transform the underlying photo to qualify as fair use. Thanks for the comment!

[@Unknown17]

@@LegalPadLaw But think of it this way. Remember the infringement suit against the Beatles for copying the tune from "He's So Fine" with their "My Sweet Lord"? The Beatles changed the INSTRUMENTS, they changed the VOCALS, they rewrote the lyrics entirely, and they even changed the THEME of the song from a simple boy-and-girl love song to an esoteric exploration of the afterlife, but they still lost the suit! And that's because they STARTED with someone else's tune and only changed some of the window dressing. This case is exactly consistent with that one, in my view.

[@rationalsatanist1811]

this proves that andy wurhol was not an artist.

[@LegalPadLaw]

Hot take.

[@pokepress]

AI art is kinda a different ball of wax since it generates content from scratch in a way where a large number of source works each have a very small influence on certain aspects of the output. What Warhol did was more akin to the img2img tab in Stable Diffusion, or certain models in ControlNet.

[@ttt5205]

That would be heavily underestimating the rate of overfitting. No one is able to confirm whether a generated image is significantly different from one or several of the original training images. For all we know half of this stuff is just a ripoff. Plenty of studies have already come out with results that don't look good for AI.

[@pokepress]

@@ttt5205 Overfitting is definitely a problem, but it is a fairly rare occurrence unless you're trying to trigger it. In many cases the output ends up being more distinct than art created directly by hand.

[@ttt5205]

@@pokepress a study has already found a significant rate of overfitting even when comparing the output to only a small part of the training data. You have absolutely no proof for your latter claim, I'd suggest being more careful with statements such as those.

[@pokepress]

@@ttt5205 We'll probably have to agree to disagree on this one. I took a quick look at a few articles out there, and while I don't claim to be immune to selection or confirmation bias, figures range from under 2% to under a tenth of a percent, and that's when the researchers were specifically trying to reproduce specific images-I'd suspect that rates are much lower during normal use if you did a reverse image search and tried to find the most similar images. Ultimately, courts will have to decide what level is significant enough to warrant liability, but my initial opinion that this is a significantly different from the Warhol case stands.

[@ttt5205]

@@pokepress Though I understand it would be difficult to understand the implications of those rates because they seem low, this is mainly the fact due to the limitations of these studies. AI (not the generative AI) is also used in order to compare generated images with original training data, this AI however, does not have the luxury of storing trained image data in the form of noise as it will be unable to compare the images. Therefore, any comparison of generated images can only be done to a tiny fraction of the generative AI's training data. This means that the vast majority of copies found in the training data will go unnoticed as it is simply impossible to store hundreds of millions of the training images to compare them in the first place. If you can find an overfit rate of ~1.8% whilst only having a sample of about 0.6% of the training data to compare it to, that's a major discovery and an awfully high rate of overfitting, imagine what the overfit rate would be if we had the resources to compare them to every single image in the training data. It wouldn't be surprising if generative AI infringes on its original training data the majority of the time it generates something. There are other limitations that could drive the overfit rate in the studies, such as limitations of the AI model comparing images, as well as this model having trouble finding infringements on multiple images to the point where it cannot point out a singular image. Overfitting isn't some problem or bug that can be fixed or trained away from generative AI, its inherent to diffusion models, it won't be getting better over time. It might in fact be getting worse as currently AI companies try to limit their datasets in favor of a more curated set.

[@girloninternet1188]

Andy Warhol wasn't an artist. Thank you SCOTUS for confirming this.