POETIC AI _ ERROR / OUCHHH (TR) by Ars Electronica is licensed under CC BY-NC-ND 2.0.
Sam Berten, Associate Member, University of Cincinnati Law Review
The rise of artificial intelligence (“AI”) has left its imprimatur on nearly every facet of society, including the realm of art. Artists are starting to utilize AI to generate digital artworks and push the bounds of traditional art media. AI art, or “neural network art” is art created using mathematical algorithms. While humans create the code behind the AI that generates these digital artworks, human choice does not influence the resultant artwork.
One of the foremost AI creations to date is the Portrait of Edmond Belamy, which sold at Christie’s Auction House in 2018 for $432,500—43 times its estimate. Obvious, a Paris-based collective, consisting of Hugo Caselles-Dupré, Pierre Fautrel, and Gauthier Vernier created the painting. This “painting,” or digitally created image, was generated by an AI method called GAN, meaning “generative adversarial network.” GAN is a method of AI that has two parts: a generator and a discriminator. The coders or “artists,”
[feed] the system with a data set of 15,000 portraits painted between the 14th century to the 20th. The Generator makes a new image based on the set, then the Discriminator tries to spot the difference between a human-made image and one created by the Generator. The aim is to fool the Discriminator into thinking that the new images are real-life portraits. Then [the creators] have a result.
The Portrait of Edmond Belamy demonstrated that AI-generated paintings were marketable in traditional auction houses. Currently, researchers around the world are continuing to push the limits of how society defines art by creating new methods to merge AI and art. One such researcher, Professor Ahmed Elgammal, has created a system called “creative adversarial network” (“CAN”). The CAN network is programmed to “produce novelty,” which means CAN creates something different from what it sees in its database of paintings from the 14th century on. Many paintings generated by CAN fall into the abstract art category which Elgammal states is, “because the algorithm has grasped that art progresses in a certain trajectory. If it wants to make something novel, then it cannot go back and produce figurative works as existed before the 20th century.”
The AI trend is not stopping with Elgammal and Caselles-Dupré. Sougwen Chung, a former MIT media lab researcher, now uses her own hands, robots, and AI to “address the closeness between person-to-person and person-to-machine communication.” Memo Atken, a London-based artist, “trained a neural network to ‘see’ images that represent some essential concepts of human life.” Mario Klingemann uses neural networks, code, and algorithms to create his amazing artworks. Refik Anadol, a Turkish artist, uses machine-learning to make “AI projects that inspire audiences around the world.” Additionally, 17-year-old Robbie Barrat developed a program “that could write its own rap lyrics using 6,000 Kanye West lyrics” at his high school programming club.
III. Case Law
Copyright ownership vests “initially in the author or authors of the work.” But, in the case of an AI-generated work, who is the author? While caselaw addressing this issue is scarce, other areas of copyright law are certainly persuasive within the realm of copyright disputes.
For instance, in the “Monkey Selfie” case, Naruto, a crested macaque, took a picture of himself with a photographer’s camera. The photographer who owned the camera, David Slater, published the photograph in a book and claimed ownership of the photograph. This prompted People for the Ethical Treatment of Animals (PETA) and Dr. Antje Engelhardt to sue Slater, “claiming Naruto was the author of the photographs and that Slater . . . infringed Naruto’s copyright.” The U.S. Court of Appeals for the Ninth Circuit held that the Copyright Act implies the author of a copyrighted work must be a human being.
Under English law, when an owner’s animal “runs on to another person’s property and causes damage, the animal’s owner is liable for this damage. This is strict liability, so there is no need to prove negligence or intent . . . it may be appropriate that some forms of physical AI (e.g. robots) could have similar [sic] legal framework put in place.”
The Supreme Court of the United States has stated that “copyright law only protects ‘the fruits of intellectual labor’ that ‘are founded in the creative powers of the mind.” The Copyright Office has stated that “works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human-author” do not count as works founded in the powers of the human mind. Additionally, the U.K. has granted “copyright protection to the person that makes arrangements for the computer to create the work.”
In Graham v. Prince, Richard Prince, a well-known artist, was sued for appropriation when Prince displayed a screenshot of an Instagram post along with Prince’s comment on the Instagram post in an art exhibit. The district court held that Prince was liable for infringement because this use was not transformative under the fair use defense.
Discussion of this issue demonstrates that if an AI-artist “sells or displays AI-art that is substantially similar to the underlying work, it is unlikely that the AI-artist will be able to rely on fair use.” Thus, no matter how “transformative” an AI-generated artwork may be in its process, the artist could still be liable for copyright infringement if the artwork appears too similar to another previous artwork. But, if the resultant work is “quite different in composition and presentation” from the original artwork, then there could be a plausible fair use defense.
This conclusion will likely be challenged, however, in the time to come. AI artworks are, by their nature, based on databases of over 14,000 paintings. The work generated, even if it appears to be similar to another existent painting, is, in its essence, a completely different painting that is an amalgam of pieces, trends, styles, and colors from the paintings in the AI database.
Further, the creators of the AI-code are likely to be the owners of the resultant images based on legal precedent. These owners and creators should be able to succeed on a fair use defense if a copyright infringement claim arises, but there are still a plethora of legal issues that accompany AI-generated artworks. For instance, once an AI-generated work is produced, such as the Portrait of Edmond Belamy, if the AI program creates another work similar to the Portrait of Edmond Belamy, does one infringe the other? Should there be a limit on how many pieces these programs can create? Should the AI program database be limited to works out of copyright (now in public domain), or can copyrighted works be included since the AI program will naturally transform the work?
AI cases will force the legal system to grapple with new ideas and adapt precedent to meet the needs of modern technology. Based on the caselaw, it is likely that computer programmers will own the works created by their AI, but there is still latent ambiguity as to many aspects of AI.
Caselles-Dupré stated that “if the artist is the one that creates the image, then that would be the machine . . . If the artist is the one that holds the vision and wants to share the message, then that would be us.”
Professor Elgammal said “there is a human in the loop, asking questions, and the machine is giving answers. That whole thing is the art, not just the picture that comes out at the end. You could say that at this point it is a collaboration between two artists – one human, one a machine.”
The future of technology is brimming with possibility, and the legal field will soon have to adapt to handle the multitude of questions that will certainly arise. AI has already affected the art market and auctions worldwide, and sooner rather later, it will affect the law.
 The Federal Bar Association’s “Art Law and Litigation Conference” that took place in New York, New York on February 6, 2020, inspired this article. Professor Ahmed Elgammal’s discussion of AI-generated art and his system, CAN, which is a creative AI network, was particularly intriguing and prompted this article.
 Christopher McFadden, 7 AI-artists That Are Changing Our Understanding of Art, Interesting Engineering (Nov. 10, 2019), https://interestingengineering.com/7-of-the-most-important-ai-artists-that-are-defining-the-genre.
 Christie’s Photographs and Prints Auction Preview, Is artificial intelligence set to become art’s next medium?: Christie’s The first piece of AI-generated art to come to auction, Christie’s (Dec. 12, 2018), https://www.christies.com/features/A-collaboration-between-two-artists-one-human-one-a-machine-9332-1.aspx; Naomi Rea, Sotheby’s Is Entering the AI Art Fray, Selling a Surreal Artwork by One of the Movement’s Pioneers This Spring, Artnet News (Feb. 8, 2019), https://news.artnet.com/art-world/sothebys-artificial-intelligence-1460332.
 Christie’s, supra note 4.
 McFadden, supra note 2.
 Sarah Ligon, AI Can Create Art, but Can It Own Copyright in It, or Infringe?, The Lexis Practice Advisor Journal (Feb. 28, 2019), https://www.lexisnexis.com/lexis-practice-advisor/the-journal/b/lpa/posts/ai-can-create-art-but-can-it-own-copyright-in-it-or-infringe.
 17 U.S.C. § 201.
 Ligon, supra note 16; see Naruto v. Slater, 888 F.3d 418, 420 (9th Cir. 2018).
 Id., see Naruto, 888 F.3d at 426.
 Emily Barwell, Legal Liability Options for Artificial Intelligence, Lexology (Oct. 16, 2018), https://www.lexology.com/library/detail.aspx?g=6c014d78-7f4c-4595-a977-ddecaa3a12e4.
 COMPENDIUM OF U.S. COPYRIGHT OFFICE PRACTICES III, § 306 (2017), (quoting Trade-Mark Cases, 100 U.S. 82, 94 (1879).
 Id. at § 313.2.
 Ligon, supra note 16; Andres Guadamuz, “Artificial Intelligence and copyright,” WIPO Magazine (October 2017), http://www.wipo.int/wipo_magazine/en/2017/05/article_0003.html.
 Graham v. Prince, 265 F. Supp. 3d 366, 370-73 (S.D.N.Y. 2017).
 Id. at 380-82.
 Ligon, supra note 16.
 Christie’s, supra note 4.