Controversial Creation: AI Debate Strikes at Foundations of IP Law

[1]

In recent years, the legal community has hotly disputed the applicability of intellectual property (IP) law to artificial intelligence (AI) creations of materials normally granted IP protection.[2] This blog post does not endeavor to show the most prudent path forward; it instead seeks to place the debate within the legal and philosophical underpinnings of IP law. The existence of the controversy, to begin with, is a testament to the United States’ broadly utilitarian justification for IP protection.[3]

Legal theorists justify IP rights under three primary modes of reasoning: personality-based, Lockean labor-based, and utilitarian.[4] Personality theorists posit that people have “moral claims” to their individual attributes and endeavors as an extension of their own personhood.[5] They see one’s intellectual creations as a facet of this personal domain to which they hold moral claim and therefore consider IP worthy of legal protection.[6] The Lockean perspective considers individuals deserving of control over the value they produce through the efforts they exert on the world.[7] This school of thought views IP as a form of such value creation. Utilitarianism seeks to “maximize social utility.”[8] The utilitarian IP justification operates on an incentives-basis, wherein the prospect of IP ownership rights forms the conditions necessary for individuals to create socially beneficial IP and thus advance social utility.[9]

The U.S. Constitution explicitly grants Congress the “[p]ower . . . to promote the Progress of Science and the Useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries.”[10] With the aim of “[p]rogress” constitutionally instantiated, the philosophical underpinnings of United States IP law are decidedly utilitarian.[11]

Commentators debating the wisdom of IP protection for AI-formulated creations have sought to contextualize such prospective IP within the three aforementioned justifications.[12] Opponents of AI-created intellectual property dispense with the proposition with relative ease under the personality- and Lockean labor-based perspectives.[13] This is because both theories pre-suppose a human deserving of exclusive rights consequent to that person’s actions directly creating the work in question.[14] With AI-created works below a certain level of human involvement, no one is so positioned.[15] The analysis is, however, more opaque under utilitarian bases because it relies, not on a priori principles of justice for the inventor, but on the policy aim of promoting social welfare. Moreover, AI’s indifference to human motivations and emotions evades traditional utilitarian incentive considerations.[16]

Commentators have conceptualized the creation of an idea as a ratio between human and software contributions.[17] With AI becoming more sophisticated and the ratio tipping more towards a majority software input, the theoretical protection of AI output could potentially disincentivize purely human IP progression.[18] On the other hand, a blanket lack of protection could potentially dampen creative production involving AI programs, leading to a less-than-optimal number of both AI applications and subsequent AI-generated works.[19]

Courts in the United States have expressly denied patent[20] and copyright[21] protection to emergent works of AI. The Federal Circuit and D.C. District Court based their decisions on statutory interpretation of “individual” in the Patent Act[22] and “authorship” in the Copyright Act,[23] dismissing policy arguments and utilitarian justifications as outside of its domain.[24] The district court in Thaler v. Hirshfeld acknowledged that “[a]s technology evolves, there may come a time when artificial intelligence reaches a level of sophistication such that it might satisfy accepted meanings of inventorship. But that time has not yet arrived, and, if it does, it will be up to Congress to decide how, if at all, it wants to expand the scope of patent law.”[25] Whether Congress will respond, and how it might assess the social utility of AI-created IP, remains to be seen.

[1] Howard Lake, Artificial Intelligence? (photograph), in flickr (Dec 29, 2014), https://www.flickr.com/photos/howardlake/37725529692/in/photostream/ [2] See e.g., Mauritz Kop, AI & Intellectual Property: Towards an Articulated Public Domain, 28 Tex. Intell. Prop. L.J. 297 (2020). [3] For discussion of the utilitarian basis of United States IP law, and the non-compatibility of AI IP protection under the alternative justifications, see infra notes 7-15 and associated text. [4] Adam Moore & Ken Himma, Intellectual Property, Stanford Encyc. of Philosophy, https://plato.stanford.edu/entries/intellectual-property/#LockJustInteProp (Aug. 18, 2022). [5] Id. § 3.1. [6] See id. [7] See id. § 3.3. [8] Id. § 3.2. [9] See id. Under this theory, IP rights, specifically patents, work as an incentive system by awarding a monopoly—an exclusive right over the invention—to an inventor, who in turn discloses their innovation to the public. Alan Devlin & Neel Sukhatme, Self-Realizing Inventions and the Utilitarian Foundation of Patent Law, 51 Wm. & Mary L. Rev. 897, 910–20 (2009) (explaining the grant of patents and ensuing monopoly for the inventor is a means to an end of overall increased happiness). [10] U.S. Const. art. I, § 8, cl. 8. [11] See Tabrez Y. Ebrahim, Artificial Intelligence Inventions & Patent Disclosure, 125 Penn St. L. Rev. 147, 199 (2020) (detailing utilitarian constitutional basis of IP); see also Nina I. Brown, Artificial Authors: A Case for Copyright in Computer-Generated Works, 20 Colum. Sci. & Tech. L. Rev. 1, 15 n.81 (2018) (recounting broad usage by courts of utilitarian theory for IP protection to exclusion of natural rights theories). [12] See e.g., Ana Ramalho, Will Robots Rule the (Artistic) World? A Proposed Model for the Legal Status of Creations by Artificial Intelligence Systems, 21 J. Internet L. 11, 19 (2017). [13] See e.g., Bruce E. Boyden, Emergent Works, 39 Colum. J. L. & Arts 377, 391 (2016) (noting lack of support under personhood or labor-based justifications on account of no human creativity or labor). [14] See Artificial Intelligence and Intellectual Property 53 (Hilty et al., eds., 2021) (noting “anthropocentrism” of deontological theories of IP protection, including labor theory and personality theory). [15] See id. at 56 (explaining failure of labor and personality theories “once human impact or guidance falls below a certain critical level”); see also Boyden, supra note 12, at 391. But see David L. Schwartz & Max Rogers, “Inventorless“ Inventions? The Constitutional Conundrum of AI-Produced Inventions, 35 Harv. J.L. & Tech. 531, 565 (2022) (arguing rights-based theories do not foreclose patents for AI inventions because human AI creators may be positioned to deserve “fruit produced by [] trees [they] plant[]”). Other observers argue against naming the human programmer as the inventor because doing so would allow someone to “invent some code that is then used on thousands of machines that end up producing an invention that the original programmer doesn’t even know exists.” See Samantha Handler, Federal Circuit Panel Balks at Accepting AI as Inventor, Bloomberg L. (June 6, 2022, 1:11 PM), https://www.bloomberglaw.com/bloomberglawnews/ip-law/X151AB9C000000?bna_news_filter=ip-law#jcite. Another commentator suggests, given AI’s arguably current and inevitable intelligence, AI property rights should be allocated to “AI users” – parties that purchase or license AI software and utilize it for invention – instead of the software developers themselves. W. Michael Schuster, Artificial Intelligence and Patent Ownership, 75 Wash. & Lee L. Rev. 1945, 2004 (2018) (recognizing that software companies are most likely not market participants because they do not have infrastructure to monetize their AI software, unlike those who purchase or license software for that reason). Justified by the goals of the Coase Theorem, such a conclusion works to solve the ownership issue by allocating rights directly to those who value it most in order to maximize economic efficiency. Id. Doing so may prevent software companies from internalizing many of the costs associated with analyzing and monetizing AI inventions and may discourage patent trolling. Id.; see also Who Owns the Rights to a Patent? The Employer or Inventor?, ADLI Law Group P.C., https://adlilaw.com/who-owns-the-rights-to-a-patent-the-employer-or-inventor/ (last visited Sept. 24, 2023) (canvasing the employee and employer relationship in terms of who can claim property rights over an invention made by an employee in different contexts). [16] Cf. Tim W. Dornis, Artificial Creativity: Emergent Works and the Void in Current Copyright Doctrine, 22 Yale J.L. & Tech. 1, 34–35 (2020) (noting AI’s lack of human motivation requires that human actors behind production of autonomous AI be focus of a utilitarian IP rights scheme). [17] Franklin Graves, Thaler Pursues Copyright Challenge Over Denial of AI-Generated Work Registration, IPWatchdog (June 6, 2022, 4:22 PM), https://ipwatchdog.com/2022/06/06/thaler-pursues-copyright-challenge-denial-ai-generated-work-registration/id=149463/ (outlining spectrum of creation from entirely human created works to entirely machine-generated works). The author proposes four categories on the “Creation-Generation Spectrum,” namely, entirely human-created works, human creation with machine, human creation with artificial intelligence (AI) or machine learning (ML), and entirely machine-generated works – each with increasing software intervention, respectively. Id. Entirely human-created works are medium- and tool-free creative expressions (e.g., speaking, dancing, and performance art). Id. Human creation with machine examples include a book written in Microsoft Word, a composite made with Adobe Photoshop, and software code written in Xcode or Visual Studio. Id. Human creation with AI/ML examples include Adobe Sensei, some NFT collections, AI-powered word processors, or graphic design generators. Id. The author proposes the final category – entirely machine-generated works – include Dr. Stephen Thaler’s “Creativity Machine” and “DABUS.” Id. [18] See Daniel J. Gervais, The Machine as Author, 105 Iowa L. Rev. 2053, 2105–06 (2020) (positing increased productions in literary and artistic fields made by AI could decrease human creativity); see also Emily Ostertag Whittle, Rage Against the Machine: Copyright Protection and Artificial Intelligence in Music, 20 Va. Sports & Ent. L.J. 56, 70 (2021) (noting utilitarian theory’s goal of net social welfare is not met by copyright protection of AI-generated music because computers are not motivated by extrinsic factors and do not need incentivizing to do what they are programmed to do). [19] See Tim W. Dornis, Artificial Creativity: Emergent Works and the Void in Current Copyright Doctrine, 22 Yale J.L. & Tech. 1, 34–35 (2020) (arguing for balanced approach to IP protection, despite AI’s lack of human qualities for incentivization); see also David L. Shwartz & Max Rogers, “Inventorless” Inventions? The Constitutional Conundrum of Ai-Produced Inventions, 35 Harv. J.L. & Tech. 531, 565 (2022) (concluding taking a rights-based approach to granting IP protection in AI works recognizes the uncontroverted fact that AI systems only come into existence as product of human efforts). [20] See generally Thaler v. Hirshfeld, 558 F. Supp. 3d 238 (E.D. Va. 2021), aff’d, Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022), cert denied, 2023 U.S. LEXIS 1763 (Apr. 24, 2023) (denying Dr. Thaler’s patent applications listing his AI machine “DABUS” as sole inventor with application amendment where he attempted to convey property rights from legally incapacitated AI machine as assignor to himself as assignee). [21] See generally Thaler v. Perlmutter, 2023 WL 5333236 (D.D.C. Aug. 18, 2023) (affirming the United States Copyright Office’s rejection of copyright protection for “A Recent Entrance to Paradise,” a completely AI-generated output of Dr. Thaler’s “Creativity Machine,” based on lack of human authorship). [22] 35 U.S.C. §§ 100(f)-(g). [23] 17 U.S.C. § 102(a). [24] Thaler, 558 F. Supp. 3d at 248 (dismissing Dr. Thaler’s policy arguments by rejecting his attempt to provide evidence to overcome the Patent Act’s plain language meaning). [25] Id. at 249.