With the advent of digital technology, national, European and international regulatory framework is challenged by the looming arrival of advanced artificial intelligence (AI) systems. Technologies exploiting AI have already become an ubiquitous part of our everyday life as autonomous machines are percolating through modern society. AI is also an increasingly powerful source of innovation.
It is widely believed that advanced, autonomous and automated AI systems are to become major drivers of economic growth.
Shlomit Yanisky-Ravid and Xiaoqiong (Jackie) Liu, ‘When Artificial Intelligence Systems Produce Inventions: An Alternative Model for Patent Law at the 3A Era’ (2018) 39 Cardozo Law Review 2215, 2223–2230; Ryan Abbott, ‘Patenting the Output of Autonomously Inventive Machines’ (2017) 10 No. 1 Landslide 16. Ryan Abbott, ‘Patenting the Output of Autonomously Inventive Machines’ (n 1); Ryan Abbott, ‘I Think, therefore I Invent: Creative Computers and the Future of Patent Law’ (2016) 57 Boston College Law Review 1079, 1080; Peter M. Kohlhepp, ‘When the Invention is an Inventor: Revitalizing Patentable Subject Matter to Exclude Unpredictable Processes’ (2008) 93 Minnesota Law Review 779, 779; Liza Vertinsky and Todd M. Rice, ‘Thinking about Thinking Machines: Implications of Machine Inventors for Patent Law’ (2002) 8 Boston University Journal of Science and Technology Law 574, 576. Ryan Abbott, ‘Everything is Obvious’ (2019) 66 UCLA Law Review 2, 5.
This paper aims at addressing a question whether AI-generated outputs can be considered patentable inventions and if so, who would be recognized as the inventor under the Polish patent law. Is there a genuine need to change our understanding of the notion of inventor as natural person? The author draws conclusions de lege lata and briefly outlines de lege ferenda observations.
Providing a uniform definition of AI poses a major challenge, as definitions vary significantly depending on the adopted research approach and assessment context.
Yanisky-Ravid, Liu (n 1) 2223. Matthew U. Scherer, ‘Regulating Artificial Intelligent Systems: Risks, Challenges, Competencies, and Strategies’ (2016) 29 Harvard Journal of Law & Technology 353, 362 as cited in Yanisky-Ravid, Liu (n 1) 2224.
Initially, AI researchers focused primarily on ‘emulation,’ that is, mimicking observable intelligent behaviour. However, superficial mimicry resulted in anything but intelligence, at least one equivalent to its human benchmark.
John Buyers, ibid. ibid. Robin C. Feldman, ‘Artificial Intelligence’ (2018) 21 Green Bag 2d 201, 202–203.; Madeleine de Cock Buning, ‘Autonomous Intelligent Systems as Creative Agents under the EU Framework for Intellectual Property’ (2016) 7 European Journal of Risk Regulation 310, 312.
Interestingly, computers are believed to have been autonomously creating patentable results for at least twenty years. In 1994, computer scientist Stephen Thaler disclosed an invention called ‘Creativity Machine,’ which combines an artificial neural network generating new ideas, an ‘imagitron,’ together with another network, a ‘perceptron,’ that perceives value or utility in the stream of outputs.
Aaron M. Cohen, ‘Stephen Thaler's Imagination Machines’ (July–August 2009) 43(4) The Futurist 28 <
Although S. Thaler is named as the inventor of both of the inventions, he states that in fact it was the Creativity Machine that autonomously invented the second patent's subject matter.
Abbott, ‘I think, Therefore I Invent: Creative Computers and the Future of Patent Law’ (n 3), 1085–1086. ibid. ibid.
The Act of June 30, 2000 on Industrial Property Law (ustawa z dnia 30 czerwca 2000 r. Prawo własności przemysłowej) does not introduce the notion of inventor, instead using the broad term creator with regard to all the inventive projects. However, for the sake of comprehension hereof, concepts will be used interchangeably.
The notion of inventive projects is an umbrella term for certain industrial property rights’ objects. Not only does it cover inventions but also utility models, industrial designs, topographies of integrated circuits and rationalization projects. Thus, the term does not refer to trademarks and geographical indications. This leads to the conclusion that industrial property objects’ ontological character differs, and so does their normative status.
Jerzy Szczotka, ‘Tytuł I. Przepisy ogólne. Art. 3’ in Tomasz Demendecki and others (eds), Piotr Kostański, ‘Tytuł I. Przepisy ogólne. Art. 3’ in Piotr Kostański (ed),
Nevertheless, the Act remains silent with regard to the definition of a creator, even though it uses notion thereof on several occasions. Neither does the Act provide explicitly inventorship requirements. It is generally derived from the wording of the relevant provisions and their ratio legis that although patent ownership can be held by legal persons, only natural persons can accomplish an intellectual creation and, therefore, be invertors.
Janusz Szwaja and Agnieszka Kubiak-Cyrul, ‘Twórcy projektów wynalazczych’ in Ryszard Skubisz (ed),
Fulfilment of the patentability requirements implies that whoever conceived the invention has creatively contributed to the subject matter of the patent and is therefore eligible for status of creator.
du Vall (n 16) 301 Krystyna Szczepanowska-Kozłowska, ‘Zagadnienia Podmiotowe’ in Ewa Nowińska, Urszula Promińska and Krystyna Szczepanowska-Kozłowska (eds), Szczepanowska-Kozłowska (n 19); Szwaja, Kubiak-Cyrul (n 16) 494.
There is a long standing consensus that the contribution shall be recognized as legally relevant insofar as it is creative.
Janina Preussner-Zamorska, ‘Autorstwo projektu wynalazczego’ in Stafan Grzybowski and Andrzej Kopff (eds), Szwaja, Kubiak-Cyrul (n 16) 493–494.
Some argue that to qualify as a joint inventor, one must contribute to the invention's conception or at least to inspiration thereof.
Janina Preussner-Zamorska, ‘Prawo do autorstwa wynalazku’ (1974) 2 Zeszyty Naukowe Uniwersytetu Jagiellońskiego 5, 62. du Vall (n 16) 302.
Other scholars argue that in order to constitute a legally relevant contribution, one's involvement must directly relate to the element of the subject matter, which is subject to non-obviousness assessment.
Stanisław Sołtysiński, Szwaja, Kubiak-Cyrul (n 16) 494.
The prevailing view appears to be that in order to qualify as a co-inventor, one must make a creative contribution, which, assessed together with other players’ inputs, will lead to creation of a solution meeting the non-obviousness requirement, hence a patentable subject matter.
Ibid.; Urszula Promińska, ‘Zagadnienia podmiotowe’ in Ewa Nowińska and Krystyna Szczepanowska-Kozłowska (eds), Andrzej Szajkowski,
Nevertheless, there appears to be a broad agreement that neither carrying out trials and calculations, dossier preparation, provision of technical support nor funding shall be considered as creative and players making such contribution must not be therefore considered co-inventors.
Szwaja, Kubiak-Cyrul (n 16) 494; Michał Staszków, Preussner-Zamorska, ‘Prawo do autorstwa wynalazku’ (n 22) 74; Staszków (n 28) 83. Preussner-Zamorska, ‘Prawo do autorstwa wynalazku’ (n 22) 63.
As demonstrated above, the assessment of the creative character of the input is indirect and based on a presumption that only a creative contribution results in an output that meets the non-obviousness requirement. Question of whether the contributor's effort satisfies creativity criteria is hence answered by drawing inference from the end result. While this deductive reasoning's usefulness cannot be denied as it facilitates listing those eligible for inventor status, it does not alone impose a conclusion that outputs generated by non-humans cannot attract intellectual property rights.
It is necessary to distinguish between the actor (the action) and the object of the action. Advanced AI systems have the capacity to generate outcomes that meet statutory prerequisites of patentability, that is, are new, involve an inventive step (are non-obvious) and are susceptible of industrial application, hence, would be worthy of patent protection had they been developed by humans. Therefore, when the creativity is assessed solely from patentable subject matter's perspective, lack of human intervention does not preclude an AI-generated output from being classified as a patentable invention. Not only is precluding a limine, the possibility of attracting intellectual property rights by non-human inventions unjustified form a legal perspective, but it is also questionable from an economic point of view as it neglects to provide incentives to players participating in the process of developing AI-generated results.
As already indicated, the assessment of the patent eligibility of the subject matter is de facto based upon the nature of the subject itself rather than upon the nature of the mental processes of the inventor by which he achieved the non-obviousness. As previously outlined, the Act on the Industrial Property Law does not include any provision explicitly defining the notion of a creator, nor does it provide prerequisites for the assessment of the act of intellectual creation. The prevailing view, however, is that only natural persons can accomplish an intellectual creation and, therefore, be invertors. This results from the fact that throughout history, human being has been assumed to be exclusive source of creativity. This assumption, however, is challenged by developments in computer technology.
Ralpf D. Clifford, ‘Intellectual Property in the Era of the Creative Computer Program: Will the True Creator Please Stand Up?’ (1997) 71 Tulane Law Review 1675, 1676. Annemarie Bridy, ‘Coding Creativity: Copyright and the Artificially Intelligent Author’ (2012) Stanford Technology Law Review 5, 22.
Over the years, creativity has been described as the ability to do the unpredictable, to defy rules and break the routine.
ibid 22–23, citing Ada Lovelace, 19th century mathematician and writer, who cautioned against overoptimism about the potential of Charles Babbage's proposed mechanical general-purpose computer-Analytical Engine (‘It is desirable to guard against the possibility of exaggerated ideas that might arise as to the powers of the Analytical Engine. The Analytical Engine has no pretensions whatever to originate anything. It can do (only) whatever we know how to order it to perform’). ibid 23. David Cope,
Nevertheless, with the ongoing development of the AI technologies, we observe continuous rise of the threshold of machine creativity. Many scholars define computational creativity in terms of human consciousness. Abandoning less onerous, albeit more clear and measurable criterion of unpredictability, they adopted a belief that creativity requires consciousness. Some scholars argue that human thought is nothing but a very sophisticated algorithm and that, hence, humans and machines are in fact not as different as we are prone to admit.
Bridy (n 32) 25. Allen Newell, ‘Response: The Models Are Broken, The Models Are Broken!’ (1986) 47 University of Pittsburgh Law Review 1023, 1025 as cited in Bridy (n 32) footnote 68. Peter Kassan, ‘AI Gone Awry: The Futile Quest for Artificial Intelligence’ (2006) 12 Skeptic 30, 34. as cited in Bridy (n 32) footnote 68. Roger Schank and Christopher Owens, ‘The Mechanics of Creativity’ in Raymond Kurzweil (ed), Selmer Bringsjord, ‘Chess is Too Easy’ (1998) 101/2 Technology Review 23, 24. Ibid 23. Barbara Bottalico, ‘Cognitive Neuroscience, Decision Making and the Law’ (2011) 2 European Journal of Risk Regulation 427. Kohlhepp (n 2) 808.
Consciousness and creativity are surely fascinating phenomena. Nevertheless it appears that reflection thereon only adds to the struggle to articulate a comprehensible and workable approach to inventiveness. Therefore, from a functional point of view, ‘whether or not creative computers >think< or have something analogous to consciousness should be irrelevant with regards to inventorship criteria’
Abbott, ‘I think, Therefore I Invent: Creative Computers and the Future of Patent Law’ (n 2) 1108. ibid 1110–1111.
AI-generated outputs should not be a limine denied patent protection as they, under Industrial Property Law Act provisions, are capable of satisfying the patentability threshold. This conclusion however raises the fundamental question of who, if anyone, could be attributed with the status of a creator of a computational invention? It is crucial to comprehensively address this question since the inventor is the original owner of the entitlement to the right to a patent (Article 8(1).1 of the Industrial Property Law Act 2000) and has the moral right to be named as inventor in the patent specification (Article 8(1).3 of the Industrial Property Law Act 2000). It should be emphasized that discretion to refrain from exercising the right to authorship (inventorship) is subject to a limitation under Article 32 of the Industrial Property Law Act, which provides that where the patent applicant is not the inventor, he shall be obliged to name the inventor in the request and state the grounds on which his own right to a patent is based. As from the receipt of the patent application and throughout the patent granting proceedings, the Polish Patent Office shall invite the applicant to complete the application or to remedy any identified omissions within a fixed time limit. Failure to comply with Office's order will result in discontinuance of the proceedings (Article 42(1) of the Industrial Property Law Act 2000). Hence, where it is impossible to name an inventor, the patent cannot be granted.
Those most frequently listed as possible stakeholders are: the software programmer, the data supplier, the user of the system, and the AI system itself.
Yanisky-Ravid, Liu (n 1) 2231–2234. Emily Dorotheou, ‘Reap the Benefits and Avoid the Legal Uncertainty: Who Owns the Creations of Artificial Intelligence?’ (2015) 21(4) Computer and Telecommunications Law Review 85, 89. Abbott, ‘I think, Therefore I invent: Creative Computers and the Future of Patent Law’ (n 2) 1095. Yanisky-Ravid, Liu (n 1) 2237.
Data supplier provides the AI system with data that the system exploits and learns from.
ibid 2232–2233. ibid.
The user operates the AI system by indicating the goal to be achieved and initiating the creative process. The latter is generally limited to ‘pressing the button.’ It needs to be borne in mind, that employing another actor to invent does not make one an inventor.
W. Michael Schuster, ‘Artificial Intelligence and Patent Ownership’ (2018) 75 Washington and Lee Law Review 1945, 1961.
This simplified study of some of the actors involved in the computational creative process leads to the conclusion that AI systems create independently of programmer, data supplier and user alike. Human input is, undoubtedly, essential for the creation of an AI-generated invention, as neither do intelligent machines originate from void, nor do they operate by themselves. But the human contribution, however necessary, does not meet the threshold of creativity, and consequently, none of the human actors could be deemed an inventor under the current notion thereof.
Can the AI system itself be therefore considered an inventor? None of the existing legal systems recognizes computers as legal entities and it is generally assumed that AI will not be granted legal personhood anytime soon.
Id., p. 1114; Yanisky-Ravid, Liu (n 1) 2228; Robert C. Denicola, ‘Ex Machina: Copyright Protection for Computer-Generated Works’ (2016) 69 Rutgers University Law Review 251, 274. Promińska (n 26) 185. Yanisky-Ravid, Liu (n 1) 2231.
It should be emphasized, however, that legal persons have been denied inventor status in order to protect and safeguard moral rights, in particular the right to authorship (inventorship) of the individuals participating in the inventive process. The idea of a human-inventor defied the concept of an ‘enterprise's invention’ ( Promińska (n 26) 185. ibid.
On the contrary, computational inventions are de facto ‘inventions without an inventor.’ As already demonstrated, natural persons’ involvement in the inventive process of AI systems is generally indirect and insignificant. Thus, indicating a human-inventor with regard to AI-generated outputs is impossible, as none of the actors satisfies the threshold of inventorship. Consequently, no right to authorship (inventorship) would be violated by a legal fiction vesting patent rights in one of the stakeholders. AI systems do not hold rights that could be thereby infringed, for they are not recognized as legal entities. If, however, legal personhood was assigned to autonomous computers, they could themselves enjoy inventor status and there would be no necessity to introduce aforementioned legal fiction.
Legal fiction of ‘computational invention’ vesting the inventor status and rights thereby conferred in one of the stakeholders would introduce profoundly different rationale than legal fiction of the ‘enterprise's invention’ did. The legal justification for such fiction would not evoke to the difficulty with indicating the inventor-in-fact but to the lack thereof. The computational inventions’ inventor-in-fact is the AI system itself which, at least for the time being, cannot enjoy legal status of an inventor. Said solution, by vesting the inventor status in one of the persons who contributed to the AI-generated result would offer a reasonable incentive to actors involved in the innovation process and, at the same time, leaving aside vexed problem of computational personhood, would not undermine established legal paradigms.
Determining who ought to enjoy the default inventor status, should the Polish legislator decide to grant patents for AI-generated outputs, entails a dilemma of who among multiple actors involved in the creative process shall be granted patent rights with respect to the given invention. Assignment of exclusive intellectual property rights ought to be studied from an economic perspective. Being able to support legal reasoning by reference to economics enhances the justifiability thereof, since, naturally, an instrument that is recognized as economically efficient carries a stronger weight of justifiability. The concept of economic efficiency, that is, maximization of socioeconomic utility, does provide a useful guide in interpreting, applying and developing the law. Whilst there are a number of traditional justifications for intellectual property rights,
Peter Drahos, William M. Landes and Richard A. Posner,
Incentive appears to be a strong argument supporting the programmer's claim for patent rights as it is the developer's activity that constitutes a pedigree for computational inventions. However, as argued by Pamela Samuelson,
Pamela Samuelson, ‘Allocating Ownership Rights in Computer-Generated Works’ (1986) 47 University of Pittsburgh Law Review 1185, 1225. Abbott, ‘I Think, Therefore I Invent: Creative Computers and the Future of Patent Law’ (n 3) 1116.
Ryan Abbott concludes that ‘ownership rights to computational inventions should vest in a computer's owner because it would be most consistent with the way personal property (including both computers and patents) is treated [...] and it would most incentivize computational invention’.
Abbott, ‘I think, Therefore I Invent: Creative Computers and the Future of Patent Law’ (n 3) 1114–1115 ibid. Schuster (n 52) 1985.
Circumventing patent allocation by internalizing the inventive process in order to become a right-holder would be indeed the expected response from software companies, should patent rights for computational inventions be granted to AI users. Ronald H. Coase argued, however, that companies will externalize whenever the costs of internalization is greater than the cost of externalization.
Ronald H. Coase, ‘The Nature of the Firm’ (1937) 4 Economica 386, 395 as cited in Schuster (n 52) footnote 277. Schuster (n 52) 1995. ibid 1989. ibid 1985.
Grounds for the EPO decision of 27 January 2020 on EP 18 275 163 [1–5]; Grounds for the EPO decision of 27 January 2020 on EP 18 275 174 [1–5].
Following the oral proceedings, both applications were refused by the Receiving Section of the EPO, which based its holding on two primary grounds. Firstly, The Receiving Section noted that the application does not meet the formal requirements of Article 81 and Rule 19(1) of the European Patent Convention (EPC).
Grounds for the EPO decision of 27 January 2020 on EP 18 275 163 [19] and Grounds for the EPO decision of 27 January 2020 on EP 18 275 174 [20]. Grounds for the EPO decision of 27 January 2020 on EP 18 275 163 [21] and Grounds for the EPO decision of 27 January 2020 on EP 18 275 174 [22]. Grounds for the EPO decision of 27 January 2020 on EP 18 275 163 [22–28] and Grounds for the EPO decision of 27 January 2020 on EP 18 275 174 [23–29]. Grounds for the EPO decision of 27 January 2020 on EP 18 275 163 [29] and Grounds for the EPO decision of 27 January 2020 on EP 18 275 174 [30]. Grounds for the EPO decision of 27 January 2020 on EP 18 275 163 [30] and Grounds for the EPO decision of 27 January 2020 on EP 18 275 174 [31]. Grounds for the EPO decision of 27 January 2020 on EP 18 275 163 [31–33] and Grounds for the EPO decision of 27 January 2020 on EP 18 275 174 [32–34]. For a detailed analysis of EPO's reasoning see: Martin Stierle, Artificial Intelligence Designated as Inventor – An Analysis of the Recent EPO Case Law (2020) 69(9) GRUR International 918–924.
Likewise, patent applications indicating DABUS as the inventor filed by Thaler with the Intellectual Property Office of the United Kingdom (UKIPO)
UK IPO patent decision BL O/741/19 of 4 December 2019, < USPTO decision of 22 April 2020 on Application No. 16/524,350, < David Flint, ‘Intelligence: The Artificial Way’ (2020) 41 Business Law Review 4, 151–152. WIPO, Revised Issues Paper on Intellectual Property Policy and Artificial Intelligence (WIPO/IP/AI/2/GE/20/1 REV.) 21 May 2020 < ibid [issue 2, 17(i)]. ibid [issue 4, 20(i)]. ibid [issue 2, 17(viii)].
Without doubt, increasingly advanced AI systems challenge well-established notions of inventorship and inventor. The author argues that existing legal framework suffices for recognizing AI-generated outputs as patentable inventions. What subsequently sparks controversy, however, is query as to whoever, if anyone at all, meets the threshold of inventorship and can be thus named an inventor. As already demonstrated, it is questionable whether any of the actors involved in the computational creative process satisfies the standard. Thus, the author argues, that de lege lata, computational results do not enjoy patent protection due to the impossibility of indicating a human inventor. As already indicated, where it is impossible to name an inventor due to lack thereof, the patent cannot be granted.
Nevertheless, denying computational inventions patent protection encourages AI-systems’ operators to maintain the secrecy of otherwise patentable technical solutions. It can be argued that lack of patent protection and excessive trade secrecy resulting therefrom slows down the dissemination of technological advances, hence doing a disservice to the social welfare.
Luigi Alberto Franzoni and Arun Kumar Kaushik, ‘The Optimal Scope of Trade Secrets Law’ (2016) 45 International Review of Law and Economics 45, 45. Schuster (n 52) 2001–2002; Ryszard Markiewicz, ‘Sztuczna inteligencja i własność intelektualna’ in Alicja Adamczak (ed),
The author argues that legal fiction of ‘computational invention,’ by vesting the inventor status and rights thereby conferred in one of the stakeholders having particular interest in patent monetization, would offer a reasonable solution in a rapidly changing creative domain; yet, it would not undermine the established legal paradigms. The pending question of whoever ought to be assigned with patent rights shall be given serious consideration by policy makers. It is further necessary to address the issue of how to protect AI-generated outputs.