AI Can Create Art, but Can It Own Copyright in It, or Infringe?

Posted on 02-28-2019

By: Sarah Ligon Pattishall, Mcauliffe, Newbury, Hilliard & Geraldson LLP

A 17-year-old bet his high school programming club that artificial intelligence (AI) could outperform human beings. To prove it, Robbie Barrat developed a program that could write its own rap lyrics using 6,000...

A 17-year-old bet his high school programming club that artificial intelligence (AI) could outperform human beings. To prove it, Robbie Barrat developed a program that could write its own rap lyrics using 6,000 Kayne West lyrics.1 He is not the only one creating art using AI. Major news organizations like The Washington Post are integrating AI into their business models.2 In addition, a painting created by Obvious using AI was recently auctioned off by Christie’s for almost a half of a million dollars.3

THE RISE IN AI-ART HAS RAISED FUNDAMENTAL questions in copyright law that authors and companies are struggling to address, including questions of copyrightability, ownership, and infringement.

How Does AI Create Art?

To answer these questions, it is important to understand how AI creates art. As explained by self-taught programmer turned AI-artist, Barrat AI-artists use two neural networks that consist of algorithms called a generator and a discriminator to create AI-art.4

For example, Barrat would feed the generator paintings and the generator would create rules based on those paintings in order to produce its own version of them.5 The discriminator would then look at both the real paintings and the AI-versions to determine which painting is “real.”6 Like a game, the generator tries to trick the discriminator into believing that its new paintings are “real.”7 The fascinating part of this process is that the resulting AI-work can sometimes be beautiful and quite different from the underlying artwork on which it is based.

Can AI-Art be Copyrighted and Who (or What) is the Author?

AI-artists like Barrat input data into these neural networks, but it is the computer program that actually creates the art and, in theory, could be the author of the work. Copyright ownership “vests initially in the author or authors of the work.”8 Although Section 101 of the Copyright Act fails to define an author, recent case law suggests that the author cannot be a computer.

In the widely publicized “Monkey Selfie” case, Naruto v. Slater, a crested macaque named Naruto picked up a photographer’s camera and clicked photographs of himself.9 The photographer, David Slater, and Wildlife Personalities, Ltd. published the “Monkey Selfies” in a book and claimed copyright ownership in the photographs.10 People for the Ethical Treatment of Animals (PETA) and Dr. Antje Engelhardt sued as next friends on behalf of Naruto, claiming Naruto was the author of the photographs and that Slater and Wildlife Personalities infringed Naruto’s copyright.11

Naruto’s claim was dismissed, and the U.S. Court of Appeals for the Ninth Circuit affirmed, holding that Naruto lacked standing to sue under the Copyright Act because animals cannot sue for infringement.12 The court reasoned that “[t]he Copyright Act does not expressly authorize animals to file copyright infringement suits under the statute”, and other sections of the Copyright Act, which refer to “children” and “widow,” for example, imply that the author must be a human being.13 Although corporations can own copyrights and sue under the Copyright Act, the court noted that corporations are considered “persons” under U.S. Supreme Court precedent and, unlike animals, these entities “are formed and owned by humans.”14

The U.S. Copyright Office likewise will not grant a registration unless the author is a human being.15 It relies on old Supreme Court precedent that “copyright law only protects ‘the fruits of intellectual labor’ that ‘are founded in the creative powers of the mind.’”16 The Copyright Office’s position is that this does not include “works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human-author.”17

Thus, it appears Barrat’s rapping robot cannot be an author that owns a copyright, but this does not foreclose AI-artists like Barrat from claiming some element of authorship in the resulting work. A computer program cannot create the work without the artist’s initial input, and AI-artists arguably exercise some degree of originality in the selection of the underlying works that form a basis for the AI-created work. Some other countries already protect such works. For example, the U.K. grants copyright protection to the person that makes arrangements for the computer to create the work.18 Given that the purpose of the Copyright Act is to provide an incentive for artists to create art, it may not be long before the United States follows suit.

What is the Line Between Fair Use and Infringement for AI-Created Works?

Apart from whether AI-artists like Barrat can claim copyright ownership in the resulting work, can the likes of Kayne West stop him from selling or displaying his work? New artists often build upon existing art, but the line between inspiration and appropriation is not always clear. In the AI-world where a computer program is literally fed copies of existing art to try to recreate it, the line would seem non-existent. Prior appropriation art cases, however, suggest that some uses may be protected by fair use if the result is sufficiently transformative from the underlying work.

In Graham v. Prince, a photographer sued the well-known appropriation artist Richard Prince for infringement when Prince took a screenshot of a third-party’s Instagram post of Graham’s photograph that Prince commented on and later displayed the post, along with Prince’s comment, in an art exhibit featuring a collection of screenshots of social media posts with nonsensical comments by Prince.19 Prince moved to dismiss on the basis of fair use, claiming that his work was transformative because a “reasonable viewer” would interpret his art as conveying a number of messages that are distinct from the underlying art.20

The district court disagreed, holding that Prince’s reproduction of the photograph in its entirety with “de minimis cropping” and “a cryptic comment” is not transformative as a matter of law.21 Unlike Prince’s prior appropriation art that the U.S. Court of Appeals for the Second Circuit previously held was insulated from infringement under the fair use defense, “Prince has not materially altered the ‘composition, presentation, scale, color palette, and media’ originally used by” Graham.22

So, if an AI-artist sells or displays AI-art that is substantially similar to the underlying work, it is unlikely the AI-artist will be able to rely on fair use. If, however, Barrat’s rapping robot, for example, produced lyrics that are quite different in composition and presentation from the original Kanye West lyrics, then Barrat could plausibly assert fair use, bearing in mind that the defense is typically more difficult to successfully assert earlier on in a dispute.23 Thus, risk averse AI-artists who wish to avoid the murky waters of fair use or high costs of discovery can rely solely on public domain works, obtain a license, or commission the underlying works.

The key takeaway from these cases is that AI-artists should not only document the creative process when selecting and inputting the underlying art, AI-artists should also consider evaluating the resulting AI-work to determine whether it is sufficiently transformative before releasing it to the public to mitigate any potential claims of infringement.


Sarah Ligon focuses her practice at Pattishall, McAuliffe, Newbury, Hilliard & Geraldson LLP on trademark prosecution and litigation in addition to copyright, right of publicity, and marketing and advertising law. She began her career at a Chicago-based intellectual property firm where she also worked as a summer associate. While in law school, she worked as a law clerk at the Tribune Company, where she handled a variety of intellectual property law matters, as well as worked on issues involving marketing and advertising, media, internet, sweepstakes, and privacy law. Prior to that, she worked in the branded entertainment industry as a group marketing manager, where she managed numerous entertainment marketing and promotional campaigns on behalf of national and international brands.


To find this article in Lexis Practice Advisor, follow this research path:

RESEARCH PATH: Intellectual Property & Technology > Copyright > Copyright Counseling & Transactions > Articles

For more information on the requirements for the granting of a copyright, see

> AUTHORSHIP AND OWNERSHIP OF COPYRIGHT

> Intellectual Property & Technology > Copyright > Copyright Counseling & Transactions > Practice Notes

For a detailed discussion on the categories of works that may be entitled to copyright protection, see

> COPYRIGHT FUNDAMENTALS

> Intellectual Property & Technology > Copyright > Copyright Counseling & Transactions > Practice Notes

For a review of the exclusive rights that are controlled by an owner of a copyrighted work, see

> EXCLUSIVE RIGHTS OF COPYRIGHT OWNERS

> Intellectual Property & Technology > Copyright > Copyright Counseling & Transactions > Practice Notes

For an analysis of the fair use defense to a claim of copyright infringement, see

> FAIR USE CONSIDERATIONS

> Intellectual Property & Technology > Copyright > Copyright Counseling & Transactions > Practice Notes

1. Dave Gershgorn, “A West Virginia teen taught himself how to build a rapping AI using Kanye West lyrics,” Quartz (March 17, 2017), https://qz.com/920091/a-west-virginia-teen-taught-himself-how-to-build-a-rapping-ai-using-kanye-west-lyrics/. 2. Corinna Underwood, “Automated Journalism – AI Applications at New York Times, Reuters, and Other Media Giants,” Emerj (last updated November 29, 2018), https://emerj.com/ai-sector-overviews/automated-journalism-applications/. 3. Jimmy Im, “This portrait made by A.I. just sold for $432,000 — that’s 40 times the original estimate,” CNBC Make It (October 25, 2018), https://www.cnbc.com/2018/10/25/portrait-made-by-artificial-intelligence-sold-for-432k-at-christies.html. 4. Joe Dworetzky, “Q&A: Robbie Barrat on training neural networks to create art,” The Stanford Daily (June 12, 2018), https://www.stanforddaily.com/2018/06/12/qa-robbie-barrat-on-training-neural-networks-to-create-art/. 5.Id. 6.Id. 7.Id. 8. 17 U.S.C.S. § 201. 9. Naruto v. Slater, 888 F.3d 418, 420 (9th Cir. 2018). 10. Id. 11. Id. 12. Id. at 426. 13. Id. 14. Id. at n. 9. 15. COMPENDIUM OF U.S. COPYRIGHT OFFICE PRACTICES III, § 306 (2017). 16. Id. (quoting Trade-Mark Cases, 100 U.S. 82, 94 (1879)). 17. Id. at § 313.2. 18. Andres Guadamuz, “Artificial Intelligence and copyright,” WIPO Magazine (October 2017), http://www.wipo.int/wipo_magazine/en/2017/05/article_0003.html. 19. Graham v. Prince, 265 F. Supp. 3d 366, 370-73 (S.D.N.Y. 2017). 20. Id. at 380. 21.Id. at 380-82. 22. Id. at 381 (quoting Cariou v. Prince, 714 F.3d 694, 706 (2d. Cir. 2013)). 23. Id. at 377 (“Due to the fact-sensitive nature of the inquiry, courts generally do not address the fair use defense until the summary judgment phase.”).