인공지능 저작물의 귀속 주체에 대한 견해들

 



인공지능이 범용성을 띤 생성형 인공지능(Generative AI)로 발전해 나가면서, 인공지능 저작물을 누구의 소유로 볼 것인가에 관한 견해들이 주목받고 있다. 여기에는 기본적으로 인공지능이 해당 저작물의 저자가 되어야 한다는 주장이 대두된 것이 출발점이라고 할 수 있다. 그러면서, 여러 학자들은 현재 AI가 생성한 창작물의 소유권이 누구에게 있는가(또는 그 누구에게라도 있는가)에 대해 고민하기 시작했다. 

이 질문에 대해서는 의외로 다양한 입장이 제시되고 있다. 

Dan Burk 등 일단의 학자들은 AI에는 새로운 법적 문제가 전혀 존재하지 않는다고 주장한다. 이들은 ChatGPT 같은 인공지능을 Microsoft Word나 Photoshop 같은 단순한 도구(tool)로 취급하는 입장이다. 

1. Dan L. Burk, Thirty-Six Views of Copyright Authorship, By Jackson Pollock, 58 HOUS. L. REV. 263, 266 (2020); Ryan Benjamin Abbott & Elizabeth Rothman, Disrupting Creativity: Copyright Law in the Age of Generative Artificial Intelligence, 75 FLA. L. REV. 1141, 1183 (2023); Andres Guadamuz, Artificial Intelligence and Copyright, WIPO MAGAZINE, Oct. 2017,https://www.wipo.int/wipo_magazine/en/2017/05/article_0003.html [https://perma.cc/3CX5- EANE]; CONG. RSCH SERV., GENERATIVE ARTIFICIAL INTELLIGENCE AND COPYRIGHT LAW 3 (2023), https://crsreports.congress.gov/product/pdf/LSB/LSB10922 [https://perma.cc/46DW- WY6W].

2. Burk, Thirty-Six Views of Copyright Authorship, By Jackson Pollock, at 266 (“There is never any question of assigning authorship to a paintbrush, saxophone, or word processor, despite their direct involvement in the act of expressive creation.”); Michael D. Murray, Tools Do Not Create: Human Authorship in the Use of Generative Artificial Intelligence, 15 CASE W. RES. J.L. TECH. & INTERNET 76 (2024) (arguing that generative AI does nothing creative and that the creative process for art is similar whether or not the artist uses AI). While that is certainly true of some simple technological uses, I think Burk’s argument understates what AI is already capable of and is likely to be even less sustainable as the technology develops. And Murray’s argument does not reflect the way the technology actually operates today. But in any event, Burk’s view doesn’t make the creativity issue go away; it just requires him to pick a human actor that is most closely connected to the creative work and treat that human as the author.

일단 무엇보다 논의의 출발점이기도 한 AI 자체가 저작자(author)가 되어야 한다고 주장하는 견해가 있다. 이러한 주장을 펴는 이들 중 대표적으로 Ryan Abbott은 인공지능이 도출한 결과물에 대해 인공지능을 발명자로 하여 미국 특허청에 출원을 하기도 했다. 

1. Ryan Benjamin Abbott & Elizabeth Rothman, Disrupting Creativity: Copyright Law in the Age of Generative Artificial Intelligence, 75 FLA. L. REV. 1141, 1201 (2023) (“Once the desirability of protecting these works is acknowledged, acknowledging AI authorship then becomes nothing more than opting for reality instead of elaborate legal fictions.”).

2. Emmanuel Salami, AI- Generated Works and Copyright Law: Towards a Union of Strange Bedfellows, 16 J. INTELL. PROP. L. & PRAC. 124, 124 (2020) (“This article proposes for the ascription of legal personhood to AI systems to ensure that AI is recognized as an author of its works and creations under copyright law.”).

3. Mala Chatterjee & Jeanne C. Fromer, Minds, Machines, and the Law: The Case of Volition in Copyright Law, 119 COLUM. L. REV. 1887 (2019) (discussing machine volition in the context of infringement, not authorship).

4. 이에 대한 반론은 다음의 논문에서 확인. James Grimmelmann, There’s No Such Thing as a Computer-Authored Work—And It’s a Good Thing, Too, 39 COLUM. J. L. & ARTS 403, 403 (2016) (“Copyright law doesn’t recognize computer programs as authors, and it shouldn’t.”).

하지만, 현재까지는 미국 저작권법은 이러한 입장을 명시적으로 배제하고 있다.  위의 Abbott이 제기한 특허청에 대한 출원에서 미국 특허청이 출원을 거절한 것에 이어, 저작권의 주체로 인공지능을 제시한 저작권 등록에서도 거절을 당했다.  그러한 거절 과정에서 미국이 특허법과 저작권법의 입장이 보다 명백하게 드러났다.

1. Naruto v. Slater, 888 F.3d 418, 426 (9th Cir. 2018) (holding that only a human can be an author); U.S. COPYRIGHT OFFICE, COMPENDIUM OF U.S. COPYRIGHT OFFICE PRACTICES § 306 (3d ed. 2021) (“The U.S. Copyright Office will register an original work of authorship, provided that the work was created by a human being.”); Pamela Samuelson, Allocating Ownership in Computer-Generated Works, 47 U. Pitt. L. Rev. 1185, 1199 (1986) (“[I]t is still fair to say that it was not within Congress’ contemplation to grant intellectual property rights to machines.”). The current law is a bit hard to square with the law’s treatment of corporations, which Charles Stross has called the original AIs, as authors in their own right. 

2. Charles Stross, Dude, You Broke the Future!, CHARLIE’S DIARY (Jan. 2, 2018), http://www.antipope.org/charlie/blog- static/2018/01/dude-you-broke-the-future.html (“I'm talking about the very old, very slow AIs we call corporations, of course.”). See also U.S. COPYRIGHT OFFICE, CIRCULAR 30: WORKS MADE FOR HIRE 4 (explaining that as defined in 17 U.S.C. § 101 “[i]f a work is a work made for hire, the employer or the party that specially ordered or commissioned that work is the author of that work.”). If corporations can be not just owners but authors, it’s not clear why other artificial entities couldn’t be.

보다 현실적인 주장으로는, AI를 소유하거나 운영하는 기업이 AI가 생성한 저작물의 저작자 및 소유자가 되어야 한다는 견해가 있다. 

1. Andres Guadamuz, Artificial Intelligence and Copyright, WIPO MAGAZINE, Oct. 2017,https://www.wipo.int/wipo_magazine/en/2017/05/article_0003.html [https://perma.cc/3CX5- EANE] (“Granting copyright to the person who made the operation of artificial intelligence possible seems to be the most sensible approach”); see also CONG. RSCH SERV., GENERATIVE ARTIFICIAL INTELLIGENCE AND COPYRIGHT LAW 3 (2023), https://crsreports.congress.gov/product/pdf/LSB/LSB10922 [https://perma.cc/46DW- WY6W], at 3 (“The creative choices involved in coding and training the AI, on the other hand, might give an AI’s creator a stronger claim to some form of authorship than the manufacturer of a camera [who does not retain copyright in photographs taken by the cameras they produce].”). 

2. Notably, the current generation of AI companies does not retain copyright in the output of its works. E.g., Terms of Use, OPENAI, https://openai.com/policies/terms-of-use [https://perma.cc/C72Y-UBT2] (last updated Jan. 31, 2024) (“[OpenAI] hereby assign[s] to you all [its] right, title and interest, if any, in and to Output.”); see also Terms of Use, STABILITY AI, https://stability.ai/terms-of-use [https://perma.cc/N6L4-67SY] (last updated Dec. 20, 2022).

한편, 컴퓨터 생성 저작물(computer-generated works)에 대해 보호를 인정하는 몇 안 되는 법 체계 중 하나인 영국법(U.K. law)은, 그러한 저작물의 저작자는 “그 저작물의 창작에 필요한 준비를 수행한 자(the person by whom the arrangements necessary for the creation of the work are undertaken)”라고 규정하고 있으며, 이는 AI 사용자일 가능성이 크지만 반드시 사용자일 필요는 없다. 

1. U.K. INTELL. PROP. OFF., ARTIFICIAL INTELLIGENCE AND INTELLECTUAL PROPERTY: COPYRIGHT AND PATENTS (2022), https://www.gov.uk/government/consultations/artificial- intelligence-and-ip-copyright-and-patents/artificial-intelligence-and-intellectual-property- copyright-and-patents [https://perma.cc/7UUZ-VC8G]. China takes the same approach. See Li v. Liu, (2023) Jing 0491 Min Chu No. 11279 (Beijing Internet Court) (English translation on file with author). See also Annemarie Bridy, Coding Creativity: Copyright and the Artificially Intelligent Author, 2012 STAN. TECH. L. REV. 1, 20 (“[A]rtificially intelligent computer programs that autonomously generate art need not be relegated for copyright purposes to scare-quoted authorship; their works can be regarded as proper works of authorship . . . by virtue of their nexus to human creativity.”). 

2. Edward Lee, Prompting Progress: Authorship in the Age of AI, FLA. L. REV. (forthcoming 2023) (manuscript at 84-85), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4609687 [https://perma.cc/CB23-RHW5] (“Creators should receive copyrights for their AI-prompted works that show, at the least, a minimally creative selection or arrangement of elements.”). Who that creator is in the context of generative AI (the writer of the algorithm? The trainer? The owner of the computer it runs on? The person who prompts the AI?) is far from clear. 

3. Bryan Casey & Mark A. Lemley, Remedies for Robots, 86 U. CHI. L. REV. 1311 (2020) (noting a similar problem in assigning liability for AI error).

Interestingly, the U.K. takes the opposite approach with patents, denying protection for AI- generated inventions. See Thaler v. Comptroller-General of Patents, Designs and Trademarks, [2023] UKSC 49.

또 다른 입장으로는, AI가 생성한 창작물은 누구의 소유도 아니며, 공공 영역(public domain)에 귀속된다고 주장하는 견해도 있다. 

1. Matt Blaszczyk, Impossibility of Emergent Works’ Protection in U.S. and EU Copyright Law, 25 N.C. J. L. & Tech. 1 (2023) (arguing that copyright cannot provide protection to AI-generated works as a matter of normative coherence of the law).

2. Grimmelmann, supra note 9, at 411 (discussing the public domain as a possible outcome); Jane C. Ginsburg & Luke Ali Budiardja, Authors and Machines, 34 BERKELEY TECH. L.J. 343, 446-47 (2019) (arguing that where “no human participant would meet the requirements of authorship . . . the work is not a work of authorship and thus falls outside of copyright’s domain”).

3. Haochen Sun, Redesigning Copyright Protection in the Era of Artificial Intelligence, 107 IOWA L. REV. 1213, 1217 (2022) (“However, AI works generated solely by autonomous AI systems should be placed in the public domain without copyright protection.”); 

4. Michael Kasdan & Brian Pattengale, A Look at Future AI Questions for the US Copyright Office, LAW360 (Nov. 10, 2022), https://g2bswiggins.wpenginepowered.com/wp- content/uploads/2022/11/Law360-A-Look-At-Future-AI-Questions-For-The-US-Copyright- Office.pdf#page=7 [https://perma.cc/MB67-FUM2] (“If recent case law from the patent law side of the aisle is to be any guide, it is likely that courts will continue to find that AI will not be eligible as the author of a work. Therefore, any such AI-generated works would be public domain.”).

5. The phrase “fall into the public domain” was coined by nineteenth century poet Alfred de Vigny, who analogized a work’s expiration of copyright protection with it falling “into the sink hole of public domain.” PAUL TORREMANS, COPYRIGHT LAW: A HANDBOOK OF CONTEMPORARYRESEARCH 154 (2007). It’s an odd phrase, as if the expiration of copyright protection were some sort of accident or step down. It isn’t. The public domain is a central part of the copyright balance. 

6. Benjamin Ely Marks, Copyright Protection, Privacy Rights, and the Fair Use Doctrine: The Post-Salinger Decade Reconsidered, 72 N.Y.U L. REV. 1376, 1377 (“The control granted by the federal copyright statute, however, is neither perpetual nor absolute . . . [r]ather, the author’s property interest in his or her intellectual or artistic production is balanced against the public’s interest in free and immediate access to materials essential to the development of society.”).

7. WIPO Copyright Treaty, Dec. 20, 1996, 36 I.L.M. 65 (emphasizing the need to “maintain a balance between the rights of authors and the larger public interest, particularly education, research and access to information”). But cf. Anupam Chander & Madhavi Sunder, The Romance of the Public Domain, CALIF. L. REV. 1331, 1334 (2004) (“But we are also concerned that the increasingly binary intellectual property debates—in which we must choose either intellectual property or the public domain—obscures other important interests, options, critiques, and claims for justice that are embedded in property rights.”). Perhaps we should speak of works “rising into the public domain” instead.

이러한 입장은 샴크리슈나 발가네쉬(Shyamkrishna Balganesh) 교수의 이론, 즉 저작권의 필수 조건은 인간의 인과적 개입(human causation)이라는 주장과도 궤를 같이 한다.

1. Shyamkrishna Balganesh, Causing Copyright, 117 COLUM. L. REV. 1, 9, 18-23 (2017) (discussing the role of mechanical intervention in copyright protection and postulating, but not resolving, the prospect of AI creativity).

중간 지점에 해당하는 해석도 있을 수 있다. 즉, 저작권의 중심을 인간의 창의성이 얼마나 프롬프트 구성에 투입되었는지에 두는 것이다. 미국 저작권청(Copyright Office)은 이론적으로는 이러한 입장을 취하고 있다. 

1. Copyright Office Statement of Policy, 88 Fed. Reg. 16190, 16192 (Mar. 16, 2023) (“In the case of works containing AI-generated material, the Office will consider whether the AI contributions are the result of ‘mechanical reproduction’ or instead of an author’s ‘own original mental conception, to which [the author] gave visible form.’ The answer will depend on the circumstances, particularly how the AI tool operates and how it was used to create the final work.”).

그러나 실무에서는, 사용자가 상당한 수준의 프롬프트 엔지니어링(prompt engineering)을 수행했음에도 불구하고 저작권 등록을 거부한 사례가 존재한다. 

1. Letter from U.S. Copyright Office Review Board to Tamara Pester (Sept. 5, 2023), https://fingfx.thomsonreuters.com/gfx/legaldocs/byprrqkqxpe/AI%20COPYRIGHT%20REGISTR ATION%20decision.pdf [https://perma.cc/Z273-7UTZ]. To be sure, the Copyright Office’s ruling depended on the artist’s unwillingness to disclaim the parts of the work created by AI.

Shyamkrishna Balganesh would presumably endorse this approach, since it looks for human agency in creation.

이러한 ‘프롬프트 기반(prompt-based)’ 접근법은 AI가 기여한 창의성은 무시하지만, 사용자가 기여한 창의성은 여전히 보호한다. 단, 해당 프롬프트 또는 일련의 프롬프트가 창작적 선택(creativity of choice)의 수준에 이를 만큼 구체적일 경우에 한정된다. 

이것이 현재 저작권청의 가이드라인과 더불어 실무에서 나타나는 흐름인 것으로 보인다. 또한, 내가 아래에서 지적하겠지만, AI 생성 저작물이 점점 더 경제적 가치를 갖게 됨에 따라, 그 소유자를 특정하려는 강력한 유인이 생겨날 것이다.