Listen on Apple Podcasts, Spotify and YouTube
Welcome to AccelPro IP Law, where we provide expert interviews and coaching to accelerate your professional development. Today we’re featuring a conversation with Angela Dunning, Partner at Cleary Gottlieb Steen & Hamilton LLP.
In this interview, Dunning discusses her work in the Naruto v. Slater case, legal questions around authorship and artificial intelligence, class action lawsuits related to training data for generative AI and what the future holds for legal approaches around AI.
Note, we first spoke with Angela in mid-2023 and conducted a follow-up interview in February 2024.
Listen on Apple Podcasts, Spotify and YouTube.
Interview References:
Angela Dunning’s Cleary Gottlieb Steen & Hamilton profile.
2:06 | Sarah Andersen v. Stability AI Ltd., 23-cv-00201-WHO (N.D. Cal. Oct. 30, 2023).
7:21 | Google LLC v. Oracle America, Inc., 18-956 (2021).
7:48 | Authors Guild v. Google, Inc., No. 13-4829, 2d Cir (2015).
10:44 | Getty Images (US), Inc. v. Stability AI, Inc. (2023).
12:19 | Naruto v. Slater, No. 16-15469, D.C. No.3:15-cv-04324-WHO, 9th Cir. (2018)
19:22 | Calkins v. Playboy Enterprises Intern., Inc., 561 F. Supp. 2d 1136,E.D. Cal., (2008).
19:55 | Elliott v. Google, Inc, Inc., No. 15-15809, 9th Cir. (2017).
Supplemental Materials:
Setty, Riddhi. Getty Images Sues Stability AI Over Art Generator IP Violations. (6 February, 2023). Bloomberg Law.
Guadamuz, Andres. Can the monkey selfie case teach us anything about copyright law?. (February 2018). WIPO Magazine.
TRANSCRIPT
I. GENERATIVE AI AND THE FAIR USE QUESTION
Neal Ungerleider, Host: Can you tell us a little bit about how the Andersen litigation has been affecting the field and what we can learn from that?
Angela Dunning: Andersen was the first of the copyright class actions to be filed directed to the training and output of a generative AI model or generative platform.
All defendants filed a motion to dismiss which the court granted almost entirely in full. The only claim that remained after the initial rounds of motion to dismiss was the claim against Stability AI for direct copyright infringement arising from its training of Stable Diffusion.
All claims against Midjourney and DeviantArt were dismissed. The court granted leave to amend and plaintiffs did that. They've come back and filed an amended complaint that continues to assert the direct copyright infringement claim against all parties. They added Runway AI, which specializes in generative AI video creation.
Runway is a new defendant in the suit, and all defendants just filed another round of motions to dismiss. Some of the earlier claims that set Andersen apart are now gone.
One of the original claims in that case was a violation of the right of publicity claim. The theory being that use of plaintiffs’ styles was an infringement of their right of publicity. The plaintiffs backed away from that claim in opposing the original motion to dismiss and said it's about use of our names.
The court said that if you're going to pursue that theory, you need to come forward with facts to show that the defendants have used your name to market their products in violation of your publicity rights. When plaintiffs amended, they left that publicity claim out of the complaint.
They've also abandoned the theory of vicarious copyright infringement. The original complaint asserted that every time a user created an output, that it was an infringement for which the platforms were secondarily and vicariously liable.
That claim is now gone and the plaintiffs are now pursuing just the direct infringement claim against Midjourney.
There are some interesting new claims. With respect to Midjourney, the plaintiffs now allege that their artwork constitutes unregistered trade dress for which Midjourney is vicariously liable for infringing. The theory is that when users create works, they are infringing trade dress that belongs to plaintiffs. Midjourney has moved to dismiss that claim as inconsistent with the law. The cases are evolving in the Andersen suit.
Meanwhile, across the hallway of the Northern District of California, you have other judges who have similarly dismissed all claims saved for the direct copyright infringement claim. At least in one case, the plaintiff was granted leave to amend and chose not to do so. They are pursuing just the direct infringement claim, for instance, against Meta.
That's the same plaintiff's counsel who filed Andersen and also filed suit against OpenAI, where claims have also been dismissed. We're seeing a trend of the Northern District of California granting dismissal with leave to amend and the claims ultimately whittling down or plaintiffs trying their hand at some new theories.
It seems to be heading in the direction of allegations really principally directed to a challenge that training an AI model with copyrighted content is an alleged direct copyright infringement. I think that that will ultimately come down to the fair use question.
NU: There are millions of images out there that are readily accessible on the internet, and I expect there will be future litigation on how images are being used to train AI. What are the key legal issues as you see them?
AD: For a copyright lawyer, these issues are among the most interesting and challenging. The issue really comes down to whether training an AI model should constitute fair use under the law, such that it is not copyright infringement to do that.
In some ways, what we are all seeing unfold before us is really just an escalation or a speeding up of the process of human creation. That's always been with us.
I would argue that every time one goes to a museum and takes in the artwork there, that person is learning from what they're seeing. That person, that artist, can then incorporate what they've learned from the masters–or just what you see in your everyday life–into the work that you go on to create.
It's the same with literary work. Every book I've ever read informs how I write and what I know.
I think in the context of these generative AI cases, the question comes down to: Should a model be able to train on the millions of images that are publicly available to everyone and that everybody learns from every day?
That issue will likely not be adjudicated to resolution anytime soon. It's going to take some time to work its way through the courts, including in the case that I'm currently working on. I think there are some clear guideposts in the Supreme Court and other jurisprudence that we've seen to date. The Supreme Court Google LLC vs. Oracle America, Inc. decision is an important one.
The question there was whether copying thousands of lines of code was nevertheless transformative and fair use. The whole purpose of it was to allow developers to go on and create new programs, new applications to increase the body of knowledge and utility available to users everywhere.
Similarly, in the Authors Guild, Inc. v. Google, Inc. case, copying whole books to make them searchable and accessible to the world was deemed fair use. One could argue here that these AI tools–including ChatGPT and the art platforms–are taking what has come before, processing that, and then making the power of that information available to all of us to use and create brand new things. That's a really exciting possibility and prospect.
I'm looking forward to seeing where the technology goes, and really exploring what that means for human potential more broadly.
NU: Are there arguments to be made that we need open access to material to better train generative AIs? Is there a legal framework you would propose for how AIs use that material?
AD: Congress and the White House are certainly thinking whether training AI on publicly available materials should be something that we're incentivizing and affirmatively making lawful.
In the UK, there has been legislation under consideration for some time that would immunize any use of publicly available data and images for training AI.
This is because these tools are with us. They're going to be deployed. We want to make sure countries are trying to figure out how they can appropriately encourage the development of these tools in a safe and fair way.
I do think there's an argument to be made that access should be open. To some extent, it's open right now. I can go on the internet, search through untold webpages and take in as much knowledge and data and information as my head can hold. I try to learn from that, grow from that, and create new material based on what I've learned.
In some ways, that's really what the AI tools are doing. We do need to be mindful of safeguards and the societal impacts that these tools may have to the extent we can. But the information is publicly available to every human.
NU: What can copyright and trademark holders do to protect themselves and their IP in an age when digital creative editing tools (including generative AI) are easier to access than ever?
AD: It's a tough question. There are a variety of tools popping up that advertise to copyright owners the ability to mask, in some ways, their works and inhibit the ability of AI to train on those materials.
There are certainly concerns that have just arisen recently in the Getty Images (US), Inc. v. Stability AI, Inc. case around trademark hallucinations coming up where copyrighted works were used to train an AI model.
Then output reflects a hallucination of the trademark, or watermark, in the work. We’re going to see litigation around that piece.
One of the big challenges that AI poses is that generating new works has become a whole lot easier to do virtually overnight. Many works can now be created quickly, so we are going to see a proliferation of content.
It will only serve to complicate what is already becoming a more complicated environment. I think that this is a challenging era that we're embarking on. No one is going to remain on one side of these issues. Every content creator will want to figure out how to use these new tools to both expand what they can do and create new markets.
At the same time, they will want to protect–to the greatest extent they can–the work they generate. They will want to protect themselves from both infringement by others and from liability to others who may claim that they benefited from other people's work.
I think we will see that these issues are not one-sided. It will not be artists or content creators against platforms or AI tools.
—
II. NARUTO V. SLATER
NU: Can you tell us about your work on Naruto v. Slater?
AD: The case of Naruto v. Slater is probably one of the greatest cases I've ever had the pleasure to litigate. I don't know if it will ever be topped. Naruto at the time was a six year old crested macaque from Indonesia.
He filed a copyright infringement lawsuit in the Northern District of California against David Slater, who was a British nature photographer who had gone to Indonesia to photograph these crested macaques in the wild where they are incredibly endangered.
He spent several days in the forest there having no luck getting the macaques to cooperate. Ultimately he set up a lead with a button that they could click to operate the shutter of the camera. Sure enough, they figured it out. They loved the sound, and they started clicking selfies.
Slater then flies back to the UK and has no idea what he's captured on this film. He is excited to see it. Slater then comes across what is now a very famous photograph of Naruto hamming it up for the camera.
He thinks “I've got the next cover of National Geographic.” This is a ticket to my being able to afford school for my kid. He is so excited. Slater self-publishes those photographs on a platform provided by Blurb, which is a San Francisco based company that I had the pleasure of doing some work with early on in their life. He publishes this book and seeks a copyright registration for these photographs.
The first impediment that he encounters is that the US Copyright Office refuses to grant a copyright. Because the macaques clicked the shutter, they–not him–would be the author of the photographs. He can't claim authorship and get the copyright. The monkeys are non-humans and cannot own a copyright either, the Copyright Office says.
PETA–People for the Ethical Treatment of Animals–then turned around and filed this lawsuit in the name of Naruto and claimed that yes, he in fact was the author. Naruto is the owner of the copyright and Mr. Slater and Blurb have committed copyright infringement by allowing the publication of these photographs.
I litigated the case for Blurb. We filed a motion to dismiss. In the Ninth Circuit, where the case was pending, there had been a decision issued in an earlier case filed by the world's porpoises, dolphins and orcas called Cetacean.
In that case, those animals filed suit under the Marine Mammal Protection Act and the EPA and a variety of other statutes passed specifically for the benefit of animals. The Ninth Circuit held that, absent an affirmative explicit statement from Congress in the statute granting animals standing to sue, they would lack statutory standing.
However, those claims can be filed by humans. For instance, a marine biologist or fisherpersons who may be affected can, but the animals themselves lack standing.
That really was dispositive in the case that Naruto, the crested macaque, filed against Blurb and Slater. There were significant issues that we delved into because the counsel who represented Naruto argued that the Copyright Act does not actually define “author,” which is true and that there was no reason that an animal like Naruto couldn't have standing.
We pointed out in our papers that the Copyright Act has a ton of language that is very human specific. It speaks of heirs and widows and other inheritance concepts that are uniquely human.
When you combine that with the basis for the Copyright Act–the constitutional provision to promote science in the useful arts–no benefit is served to society by granting copyright to an animal who can't actually distribute or publish the work.
There was a very far ranging and interesting discussion at the district court level. Then, when the monkey appealed, the Ninth Circuit had some pointed questions around Naruto's financial interest in the copyright. My response to that question was, he has none. He's a monkey. That seemed to make sense to everyone.
The upshot of that case–and I don't know whether I'll ever have as much fun in a lawsuit–was a circuit-level published decision reaffirming that there is a human authorship requirement for copyright. That has served as a significant precursor to the discussion that we're now having around AI where the Copyright Office has come out strongly on the side of requiring a human author before any copyright will be issued.
I never got a chance to ask Naruto, the crested macaque, anything about the case because they didn't present him for a deposition. My guess is that any macaque who smiles that warmly would have been in favor of having his picture widely available and would have granted a worldwide, royalty free license. Alas, I will never have the opportunity to test that theory in court.
PETA was represented by sophisticated counsel at Irell and Manella, and I believe they were pursuing claims that they believed were appropriate to highlight the sentience of these macaques and how smart and aware they are. I do not agree that federal litigation was the proper vehicle to bring those issues to light.
The lawyers did advance a number of arguments specifically around the fact that the copyright office doesn't define authorship or who can be an author. I think those arguments are going to come back and we're going to be hearing those again.
—
III. CAREER PATHWAYS
NU: What made you interested in copyright and IP as a specialty?
AD: My background is in literature and creative writing. At some point, I realized I wasn't as great a poet as I maybe wanted to be. My brain also worked as a lawyer–a puzzle solver.
So I went to law school, and absolutely loved it. Over the years, I had an amazing opportunity to touch some really fantastic cases. I litigated a case for Playboy called Calkins v. Playboy Enterprises Intern., Inc. The case involved allegations Playboy had infringed the copyright in a photograph taken by the high school yearbook photographer of the centerfold model.
I litigated that case to summary judgment of fair use because it was news reporting and transformative use of this photograph of the centerfold from high school to tell her personal story. That was one of the early cases that hooked me.
I was thrilled to be able to represent Google in a case called Elliott v. Google, Inc. that challenged the Google trademark as “generic”. As billions of us use the word “google” every day, it became important to make sure that the case law reflected the law in this area. So long as a brand serves a source identifying function (which the world famous “Google” mark does), it doesn't matter how any of us use that term in everyday speech.
The Ninth Circuit issued a great decision. Bringing that decision around was a huge moment to participate in.
NU: Can you give our listeners some background on Cleary and your role there?
AD: Cleary Gottlieb is an international law firm specializing in litigation, corporate transactions, and antitrust competition. Cleary is headquartered in New York, but with offices all around the world.
I'm super excited to be part of building out the offices here in the Bay Area in California.
NU: You were at your previous firm, Cooley, for a long time. For our listeners who may be considering switching firms, can you tell us what helped you make the decision and what advice you can give?
AD: Cooley is a great firm. I have many great friends there and I'm working with colleagues from Cooley to this day. I do not wish to disparage Cooley in any way. For me, the opportunity to help Cleary grow its California offices, which opened just a couple of years ago, was a really exciting opportunity.
The firm has a tremendous dedication to excellence, client service, and camaraderie and collegiality. This was really attractive to me. I think their international expertise, the cross-border transactions, the international competition and antitrust issues are going to be hugely important.
Listen on Apple Podcasts, Spotify and YouTube.
This AccelPro audio transcript has been edited and organized for clarity. This interview was recorded on May 31, 2023 with a subsequent follow-up recorded on February 20, 2024.
AccelPro’s expert interviews and coaching accelerate your professional development. Our mission is to improve your day-to-day job performance and make your career goals achievable.
Please send your comments and career questions to questions@joinaccelpro.com. You can also call us at 614-642-2235.
If your colleagues in any sector of the IP law field might be interested, please let them know about AccelPro. As our community grows, it grows more useful for its members.
