by Naomi Perla*
This Contribution examines whether an artist can claim copyright protection over art they created with the assistance of an artificial intelligence program. Naomi Perla (’21) argues that such works satisfy the “original work of authorship” requirement pursuant to 17 U.S.C. § 102(a), thereby granting copyright protection to the artist. The requirements of both authorship and originality are satisfied due to the artist’s creative choices that are largely reflected in the finished pieces. Moreover, the Copyright Act is meant to expand to include new works of art so that artists are consistently incentivized to create for the benefit of the public.
The purpose of the intellectual property regime, as set out by the Intellectual Property Clause, is to incentivize artists and inventors to create and produce works that will benefit society.2 While the Copyright Act aims to further these goals by granting protection for various works, Congress also intended to create some limitations as to the types of works afforded protection.3 Accordingly, the Copyright Act stipulates that a work must be considered an “original work of authorship” pursuant to 17 U.S.C. § 102(a).4 The main purpose of this statement is to set out the originality requirement constitutionally necessary to establish copyright protection.5 The requirement of “authorship,” although not explicitly articulated by courts as a constitutional requirement distinct from originality, is an imperative component of the analysis that courts have grappled with in unique circumstances.6 The issue of authorship becomes increasingly significant as technology advances and replaces the human author in some capacity, thereby forcing courts to reinterpret the meaning of “authorship” within the Copyright Act.7
The utilization of technology specifically within the art space poses challenges for courts faced with determining whether a work of art can be considered an “original work of authorship.”8 To satisfy the originality requirement, an author must independently create the work and provide a “modicum of creativity.”9 Moreover, according to the Compendium of U.S. Copyright Office Practices (“Compendium”), “to qualify as a work of ‘authorship’ a work must be created by a human being.”10 Works of art that are created by a technological mechanism that uses random mechanical processes and requires no human intervention fail to satisfy the Copyright Act’s requirements, thereby falling into the public domain.11 When artists utilize technology to assist them in creating their art pieces, they arguably delete themselves from a part of the creative process. This raises the question of whether these art pieces are authored by a human, and whether there is sufficient creativity present to satisfy the originality requirement. More recently, artists have started to use artificial intelligence (“AI”) programs to create their art, which has exacerbated these existing concerns about authorship and originality.12 AI programs perform considerably more of the work than a standard technological mechanism, which in turn leads to less creativity being provided by the human author. As such, it is unclear whether art work created using AI programs are worthy of copyright protection.
This Contribution argues that art produced by a human creator with the assistance of an AI program has been “authored” to the extent required by the Copyright Act, and satisfies the Act’s originality requirement. Accordingly, such art falls within the definition of an “original works of authorship” pursuant to 17 U.S.C. § 102(a), and is worthy of copyright protection.
The phrase an “original work of authorship” can be parsed out into two components that must be satisfied: authorship and originality.13 The “originality” component has been clearly articulated by the Supreme Court, while “authorship” has been left somewhat unattended.14
Although the Copyright Act does not provide a definition of “authorship,” courts have interpreted the phrase broadly and with enough flexibility to reflect the everchanging creative landscape.15 That said, the definition of “authorship” is not without limits.16 Even the Ninth Circuit, which argued for an expansive definition of “authorship” in Garcia v. Google, later sought to limit the term’s scope in Naruto v. Slater.17 In Naruto, the Ninth Circuit explained that a flexible definition of “authorship” did not mean that the term had no bounds.18 Accordingly, the court denied standing under the Copyright Act to a monkey who took a selfie because “authorship” did not include animals.19 Similarly, the Compendium defines “authorship” as requiring a human creator; therefore, when there is absolutely no human intervention, authorship is not present.20
The compromise between these two approaches lies between requiring a human creator and recognizing authorship in situations where the human is contributing less than what is customary. For instance, courts have granted authorship to human creators and considered the originality requirement to be satisfied in cases where the human creator’s involvement was less than traditionally expected.21 For example, in Burrow Giles v. Sarony, the Supreme Court granted authorship to a photographer, even while recognizing that it is the camera that captures the image and transforms it into a medium that can be showcased to the public.22
Although the two components are separate, they are related in that determining whether the human creator is the author can be dependent on whether the creator has contributed the “modicum of creativity” necessary to satisfy the originality requirement.23 In Feist Publications Inc. v. Rural Tel. Serv., the Supreme Court emphasized that the bar for originality is low.24 The Court further held that if creativity could not be found in the work itself, a court may look to the manner in which the author selected, compiled, or arranged various elements to satisfy the requirement.25 Years prior to Feist, the Court utilized this form of reasoning and applied it to photographs in Burrow-Giles, concluding that the photographer’s creative contributions in selecting, compiling, and arranging the various elements in the photo were sufficient to satisfy the originality requirement, despite the large percentage of work the camera was contributing, and granting authorship to the photographer and copyright protection over the photo.26
In determining whether adequate originality is being provided by the human creator, some courts have designed unique tests for particular circumstances that help satisfy the original principals laid out in Feist: the lion’s share balancing test27 and the sufficient control test.28 The balancing test is specifically applied when a software program is utilized.29 This test compares the program’s contributions in creating the art work to the contributions of the program user. If the program’s work constitutes the lion’s share of the work and the user’s contributions were merely marginal, then the programmer would be able to gain copyright protection over the output and vice versa.30
The control analysis test looks at whether the human creator has sufficient control over the creative process, even if other entities contributed creatively as well.31 For instance, in Lindsay v. Wrecked & Abandoned Vessel R.M.S. Titanic, the court found that the director held the copyright over the film regardless of the various contributions provided by other individuals, since he was masterminding the entire process, thereby ensuring his original ideas were being produced.32 The Third and Fifth Circuits applied this analysis where technology was employed, and both courts found in favor of the individual claiming authorship and protection.33
Although courts have applied the analyses above and evaluated various forms of art created with technology, art created with the assistance of an AI program has largely been left unevaluated.
The use of artificial intelligence programs to assist in creating works of art is a novel and unique process, and the Copyright Act should expand to include these art works within its borders. Providing protection for such art is consistent with the history of the Copyright Act, which is one of gradual expansion,34 and the way in which the Act is written.35 Congress intentionally left “authorship” undefined.36 Rather than attempting to predict the ways in which art would evolve, Congress recognized it would better suit the copyright regime to leave terms ambiguous, which would allow the Act to adapt to the times and keep the copyright regime from ceasing to exist.37 The Supreme Court acknowledged the Act’s expansiveness in Burrow Giles, where the camera was recognized as merely a tool utilized by artists to create their works.38 The Court opined that “the only reason why photographs were not included in the extended list in the act of 1802 is probably that they did not exist, as photography as an art was then unknown . . . .”39 This analysis can be applied to AI programs, which, like cameras, are merely utilized by artists as a tool to create. Artificial intelligence is simply the next step in the advancement of technology.
Furthermore, courts can easily apply the originality analysis to the AI context by determining whether the contributions by the artist are sufficient to satisfy the originality requirement. When an artist operates an AI program, they must first determine the input that the AI program will use to generate the output.40 The AI program can only create based on the specific rules provided by the input, which in turn has been provided by the artist.41 Additionally, an artist using an AI program must also choose which of the output will be showcased to the public and in what medium they will be viewed in.42 For instance, an artist may choose specific art pieces that have particular characteristics, like children between the ages of one and five, as the input. The AI program will create the output based on that set of rules provided by the artist. Once completed, the artist will look at all the works created and perhaps select five to be converted to 3-D canvas for optimal viewing. All of these choices alone should satisfy the low bar of the originality requirement. However, if these choices as a whole are insufficient, originality can be satisfied by the mere selection and arrangement of the elements, as per Feist.43 Because the artist using an AI program typically selects and compiles the input, selects the output, and arranges the final product, the originality requirement is fulfilled.
Although the analysis above seems to satisfy the requirements, the Seventh Circuit’s decision in Kelley v. Chicago Park District raises some concerns. In Kelley, the Seventh Circuit concluded that the authorship and originality requirements were not met where a park exhibit was created by a human creator using flower arrangements.44 Even though the human creator selected the flowers and fashioned the initial arrangement, the court opined that the garden would evolve based on the forces of nature, and would continue to change long after the creator completed his work.45 This rationale may prove problematic in the AI context because once the creator plugs in the data set, the AI program is essentially left to its own devices. However, there is a key distinction: the AI program can only function based on the predetermined rules set out by the input provided by the creator. In contrast, the artist in Kelley does not have control over nature, even if he initially picks the flowers. Thus, unlike a garden, an AI program must abide by the guidelines set for it.
Because AI is such a novel concept, the lion’s share balancing test, which was created precisely to deal with technological programs, may be a better method for determining whether the originality requirement is satisfied.46 For instance, in Rearden LLC, the court determined that the actions taken by the actors and directors could not be considered merely marginal since their creative choices were largely reflected in the finished product.47 Accordingly, the programmer could not claim that the program’s contributions constituted the lion’s share of the work.48 When an artist utilizes an AI program, their creative choices are manifested in the finished art pieces as well. Therefore, when comparing the work of the AI program to that of the artist, the artist’s contributions are far from marginal and represent the lion’s share of the work.49
The last method that could be applied by courts to determine originality is the sufficient control test.50 An artist who chooses the input, which creates the rules the AI program must follow, and modifies the output, demonstrates a high degree of control over the process and the finished product, thereby fulfilling the originality requirement, as per Lindsay.51 All of these decisions indicate that the artist is the one masterminding the process to bring their creative concepts to life.52 However, irrespective of these factors, according to the Supreme Court in Community for Creative Non-Violence v. Reid, the author must be the one who translates the idea into the fixed medium of expression.53 The application of this concept to the AI context potentially leads to some complications, since it is the AI program that takes the ideas set forth in the input and expresses them through the output. However, this holding did not proscribe protection in many other instances where technology was utilized in a similar fashion, and should not prohibit protection in the AI context either.54 No matter which lens one views the originality requirement, the finished product reflects a myriad of creative choices, and that is sufficient to satisfy the “modicum of creativity” necessary to fulfill the requirement.
Finally, the Constitution clearly states that the purpose of the Intellectual Property Clause is to promote the progress of the arts by granting rights to authors for a limited period of time.55 This produces an economic incentive that motivates artists to create and disseminate their works to the public. Without protection, artists will lose this important incentive.56 As the Supreme Court opined in Twentieth Century Music Corp. v. Aiken, “when technological change has rendered its literal terms ambiguous, the Copyright Act must be construed in light of this basic purpose.”57 If copyright protection is denied to artists who create art with the assistance of an AI program, then the industry will diminish, leaving a gap in the economy where those works of art belong and depriving the public of their benefit.58
Congress intended for the Copyright Act to expand and accommodate technological and artistic advances.59 Nevertheless, a piece of art must satisfy certain requirements—namely the originality and authorship requirements—in order to gain protection.60 Art pieces that were created by a human artist, with the assistance of an artificial intelligence program, are generally able to satisfy these requirements, and should be included within the definition of an “original work of authorship.” Courts tend to interpret “authorship” broadly and have concluded that a human creator can be considered to have authored an art work even when their creative choices are minor, thereby satisfying the originality requirement.61 Moreover, providing protection helps incentivize artists to continue creating, which satisfies the policy rationale underlying the Copyright Act.62
* Naomi Perla is J.D. Candidate (2021) at New York University School of Law. This piece is a commentary on the problem presented at the 2020 BMI (Broadcast Music, Inc.) Moot Court Competition at Cardozo School of Law. The question presented asked whether works that have been created by a human artist and in part by artificial intelligence could fall within the definition of an “original work of authorship” pursuant to 17 U.S.C. § 102(a), thereby causing the work to be worthy of copyright protection. This Contribution presents a distillation of the side of the argument assigned to the author in the BMI competition, and the views expressed herein do not necessarily reflect the views of the author.
2. See U.S. Const. art. I, § 8, cl. 8; see also Sony Corp. of Am. v. Universal City Studios, Inc., 464 U.S. 417, 429 (1984) (describing the purpose of the limited rights granted to creators by the Intellectual Property Clause to be to incentivize further creativity).
3. See Melville B. Nimmer & David Nimmer, 1 Nimmer on Copyright § 2.03 (Matthew Bender, rev. ed. 2019) (“A flexible definition was intended that would neither ‘freeze the scope of copyrightable subject matter at the present stage of communications technology [n]or . . . allow unlimited expansion into areas completely outside the present congressional intent.’” (internal citations omitted)).
4. 17 U.S.C. § 102(a).
5. See Feist Publ’ns, Inc. v. Rural Tel. Serv. Co., 499 U.S. 340, 348 (1991) (noting that the “original work of authorship” component of the Copyright Act sets out the originality requirement).
6. See Andrien v. Southern Ocean Cnty. Chamber of Commerce, 927 F.2d 132, 135 (3d Cir. 1991) (finding authorship for an individual who created a map by utilizing a printer); Lakedreams v. Taylor, 932 F.2d 1103, 1108 (5th Cir. 1991) (finding copyright protection for t-shirt designs created by humans and transposed by printers onto silkscreen); Urantia Found. v. Maaherra, 114 F.3d 955, 958 (9th Cir. 1997) (holding a religious book, even if created in part by a celestial being, could receive a copyright for the input provided by the human author); Penguin Books U.S.A., Inc. v. New Christian Church of Full Endeavor, Ltd., 2000 U.S. Dist. LEXIS 10394, at *41 (S.D.N.Y. July 21, 2000) (“As a matter of law, it is irrelevant for copyright purposes whether Jesus wrote the Course.”); Naruto v. Slater, No. 15-cv-04324-WHO, 2016 U.S. Dist. LEXIS 11041, at *9–10 (N.D. Cal. 2016) (holding that a monkey cannot claim authorship over a photo because the authorship requirement can only be satisfied with a human creator).
7. See Burrow-Giles Lithographic Co. v. Sarony, 111 U.S. 53, 56–60 (1884) (holding a photographer as the author of his photographs).
9. See Feist Publ’ns, 499 U.S. at 346 (1991).
10. U.S. Copyright Office, Compendium of U.S. Copyright Office Practices § 313.2 (3d ed. 2017).
11. Id. (“[T]o qualify as a work of ‘authorship’ a work must be created by a human being.”).
12. See, e.g., Obvious is a collective of friends, artists and researchers, obvious, https://obvious-art.com/ (last visited Feb. 8, 2021).
13. 17 U.S.C. § 102(a).
14. See Feist Publ’ns, Inc., 499 U.S. at 347 (discussing the originality component of the phrase “original work of authorship” at length, while leaving the authorship component less explored).
15. See Garcia v. Google, Inc., 786 F.3d 733, 741 (9th Cir. 2015) (“The statute purposefully left ‘works of authorship’ undefined to provide for some flexibility[.]” (citing Melville B. Nimmer & David Nimmer, 1 Nimmer on Copyright § 2.03 (Matthew Bender, rev. ed. 2019))).
16. Nimmer on Copyright § 102 (Matthew Bender, rev. ed. 2019); U.S. Copyright Office, Compendium of U.S. Copyright Office Practices § 313.2 (3d ed. 2017) (stating that to “qualify as a work of authorship a work must be created by a human being”).
17. See Garcia, 786 F.3d at 741; see also Naruto, No. 15-cv-04324-WHO, 2016 U.S. Dist. LEXIS 11041, at *9–10 (holding that a photo taken by a monkey could not gain copyright protection because it was not created by a human).
20. Compendium (Third) § 306 (explaining that to “qualify as a work of authorship a work must be created by a human being”).
21. See Andrien v. Southern Ocean Cnty. Chamber of Commerce, 927 F.2d 132, 135 (3d Cir. 1991) (finding authorship for an individual who created a map using a printer, where the printer did most of the creative work instead of the human); see alsoLakedreams v. Taylor, 932 F.2d 1103, 1108 (5th Cir. 1991) (finding copyright protection for t-shirt designs created by humans, but where the printers did a substantial amount of the work by transposing the designs onto silkscreen).
22. See Burrow-Giles Lithographic Co. v. Sarony, 111 U.S. 53, 56–60 (1884) (granting authorship to the photographer because of the creative work he contributed in selecting and arranging the items within the photograph).
24. See Feist Publ’ns, 499 U.S. at 346 (noting that “even a slight amount [of originality] would suffice”).
25. Id. (discussing the ability for one to gain copyright protection through the mere selection, arrangement, or compilation of various elements).
26. See Burrow-Giles Lithographic Co., 111 U.S. at 56–60.
27. See Torah Soft Ltd. v. Drosnin, 136 F. Supp. 2d 276, 283 (S.D.N.Y. 2001) (concluding that because defendant user only contributed a marginal amount of the work by inputting a single word into the program, no copyright protection should be provided): Rearden LLC v. Walt Disney Co., 293 F. Supp. 3d 963 (N.D. Cal. 2018) (holding no copyright for the programmer on the output created by the software program since the input contributed by the user was not merely marginal); Digital Drilling Data Sys. LLC v. Petrolink Servs., No. 4:15-CV-02172, 2018 U.S. Dist. LEXIS 83158, at *20–21 (S.D. Tex. May 16, 2018) (providing copyright protection for the plaintiff, the program creator, over the schema which is created by the program, since the elements of the schema already existed in the program and was considered the lion’s share of the work).
28. See Lindsay v. Wrecked & Abandoned Vessel R.M.S. Titanic, 1999 U.S. Dist. LEXIS 15837, at *15–18 (S.D.N.Y. Oct. 13, 1999) (holding the director was the author of the film due to the level of control he had over the entire process to ensure his original ideas were being produced, regardless of the fact that he did not actually film every part of the movie); Brod v. Gen. Publ’g Group, Inc., 32 Fed. App’x 231, 234 (9th Cir. 2002) (finding the author’s contribution to the photos and the creation of the book were sufficiently original “even though he did not trigger the shutter”).
29. See Torah Soft Ltd., 136 F. Supp. 2d at 283 (software program that compiles information); Rearden LLC, 293 F. Supp. 3d at 963 (software program that converts human expression into CGI); Digital Drilling Data Sys. LLC, No. 4:15-CV-02172, 2018 U.S. Dist. LEXIS 83158, at *20–21 (software program that benefits the oil industry).
30. For instance, in Torah Soft, the user’s input of a single word into the computer software program was determined to be merely marginal, since it was the system that compiled all the information. See Torah Soft Ltd., 136 F. Supp. 2d at 283. In contrast, when the creative choices in the input contributed by the user was largely reflected in the output, the court in Rearden LLC v. Walt Disney Co., held that contribution by the user was not merely marginal. See Rearden LLC, 293 F. Supp. 3d at 963.
31. See supra note 28.
32. See Lindsay, 1999 U.S. Dist. LEXIS 15837, at *15–18.
33. See supra note 21.
34. See Notes of Committee on the Judiciary, H.R. Rep. No. 1476, 94th Cong., 2d Sess. 51 (1976), reprinted in 1976 U.S.C.C.A.N. 5659, 5664 (“The history of copyright law has been one of gradual expansion in the types of works accorded protection. . . . Authors are continually finding new ways of expressing themselves, but it is impossible to foresee the forms that these new expressive methods will take.”).
35. See, e.g., Garcia v. Google, Inc., 786 F.3d 733, 741 (9th Cir. 2015).
36. Melville B. Nimmer & David Nimmer, 1 Nimmer on Copyright § 2.03 (Matthew Bender, rev. ed. 2019) (suggesting that Congress made a deliberate choice to leave the term authorship undefined).
37. Id; see Garcia, 786 F.3d at 741.
38. See Burrow-Giles Lithographic Co. v. Sarony, 111 U.S. 53, 56–60 (1884) (granting copyright protection for photographs regardless of the technological and novel nature of cameras and photography).
39. Id. at 58.
40. See Discover the Story Behind the Next Rembrandt, The Next Rembrandt, https://www.nextrembrandt.com/ (last visited Feb 8, 2021) (describing the process of creating the Next Rembrandt as entailing the creation of a database of all of Rembrandt’s original works, which was then inputted into the AI program).
43. See Feist Publ’ns, Inc., 499 U.S. at 348 (determining copyright protection can be granted in situations where the creator merely selects, compiles, or arranges the various elements).
44. See Kelley v. Chi. Park Dist., 635 F.3d 290, 302 (7th Cir. 2011) (denying copyright protection for a park exhibit created by a human).
46. See Torah Soft Ltd., 136 F. Supp. 2d at 283 (utilizing the lion’s share balancing test to determine whether the program user should gain copyright protection for contributing a marginal amount of work); Rearden LLC, 293 F. Supp. 3d 963 (applying the balancing test to determine program creator could claim copyright protection over the output); see also Digital Drilling Data Sys. LLC , No. 4:15-CV-02172, at *20–21 (employing the lion’s share test when determining copyright protection over the schema created by a software program).
47. See Rearden, 293 F. Supp. 3d at 963.
50. See supra note 28.
51. Lindsay, 1999 U.S. Dist. LEXIS 15837, at *15–18.
53. See Cmty. for Creative Non-Violence v. Reid, 490 U.S. 730,737 (“As a general rule, the author is the party who actually creates the work, that is, the person who translates an idea into a fixed, tangible expression entitled to copyright protection.”).
54. See Burrow-Giles Lithographic Co., 111 U.S. at 56–60 (granting protection for photographs).
55. See U.S. Const. art. I, § 8, cl. 8. (“To promote the progress of science and useful arts, by securing for limited times to authors and inventors the exclusive right to their respective writings and discoveries”).
56. See, e.g., Sony Corp. of Am. v. Universal City Studios, Inc., 464 U.S. 417, 429 (1984); Weissmann v. Freeman, 868 F.2d 1313, 1325 (2d Cir. 1989) (finding the incentive interests an important factor in determining copyright protection due to its effect of causing continued research); see also Mazer v. Stein, 347 U.S. 201, 219 (1954) (holding copyright protection for statuettes used as the base of a lamp in part due to the incentive to further research, disseminate ideas, and create art).
57. Twentieth Century Music Corp. v. Aiken, 422 U.S. 151, 156 (1975) (discussing the advent of the commercial radio and the granting of copyright infringement to the broadcast of copyrighted musical compositions).
58. See Daryl Lim, AI & IP Innovation & Creativity in an Age of Accelerated Change, 52 Akron L. Rev. 813, 815 (2018)(“Excluding AI-generated work for protection leaves an economic lacuna, which U.S. IP policy is loath to do. Allowing AI-generated works to fall into the public domain reduces the incentive to invest in the growth of the industry.”).
59. E.g., Garcia v. Google, Inc., 786 F.3d 733, 741 (9th Cir. 2015) (citing Melville B. Nimmer & David Nimmer, 1 Nimmer on Copyright § 2.03 (Matthew Bender, rev. ed. 2019)).
60. See Feist Publ’ns, Inc. v. Rural Tel. Serv. Co., 499 U.S. 340, 111 (1991) (setting out the requirements for copyright protection).
61. See supra note 6.
62. See supra notes 57–59.