.png)
This guest post under the RFMLR Blog Series on Evolving Landscape of Intellectual Property Rights is authored by Dr. Athira P. S., Director of the Centre for Intellectual Property Rights at National University of Advanced Legal Studies, Kochi.
Image Credits: WIPO
1. INTRODUCTION
Bloomberg, in 2016, reported an interesting incident on something known as ‘Pseudo-AI’, by looking into the practices of certain companies employing low paid workers to ‘act’ in the digital space as ‘bots’. What can be termed as a ‘Wizard of Oz’ effect, wherein the mystique surrounding Artificial Intelligence (“AI”) is thought to overpower its contemporary scope and applications, is relevant here. Some authors opine that perceiving, consequently believing, and ultimately acting on the notion that AI is more creative and smarter than human beings, may even be counterproductive.[i]
Scientists and engineers have so far considered computers and AI as tools or mediums that process human-provided inputs and act under human supervision. Furthermore, AI-assisted inventions or AI-implemented inventions have been granted patents universally. However, recently, the premise that AI entities can be as inventive and ‘smart’ as, if not more than, human inventors, has transcended science fiction to reality to a certain extent, through the filing of numerous patent applications for independent AI-generated inventions in different jurisdictions.
The World Intellectual Property Organization (“WIPO”) has discussed the relevance of AI in intellectual property over its Trend Reports as well as conversations among different stakeholders. As per WIPO, AI lacks a universally accepted definition, and it uses the phrase ‘deep supervised machine learning’ while clarifying that AI is a branch of computer science that aims to develop systems and machines capable of performing tasks that have been, so far, the exclusive domain of human intelligence. Depending on the task or area of technology, different subsets such as machine learning, deep learning, NLP, etc., may be employed.
2. THE DABUS CONTROVERSY AND RESPONSES FROM PATENT OFFICES
The Device for the Autonomous Bootstrapping of Unified Sentience ("DABUS") is an AI entity, invented by Dr. Stephen L Thaler. He classifies DABUS as a ‘swarm of many disconnected neural nets’ with interrelated memories, while others have identified it as ‘a type of connectionist artificial intelligence’ or as a ‘complicated creativity machine’. As per Dr. Thaler, DABUS is the inventor of Food Container and Devices and Methods for Attracting Enhanced Attention.
For this purpose, patent applications[ii] were filed at the European Patent Office (“EPO”) with a priority date of November 7, 2018, for certain inventions – fractal food container and neural flame for attracting enhanced attention – naming DABUS as the inventor. Here, Dr. Thaler, as the inventor of DABUS itself, was the applicant. Later, applications were also filed at the United Kingdom Intellectual Property Office (“UKIPO”), United States Patent and Trademark Office (“USPTO”) and under the Patent Cooperation Treaty (“PCT”) system.
Dr. Thaler argued that the DABUS is the inventor of these inventions and that Dr. Thaler himself must be considered as the applicant, and raised the following arguments:
Machines cannot own patents, and that as the employer and/or owner of DABUS, as well as a successor in title, Dr. Thaler is the rightful applicant.
Even when asked to modify the application and remove DABUS as the inventor, he argued that to do so would be immoral and contrary to the concept of moral rights of inventors, and skew the public notice function of the application.
The basis of the patent system being acknowledgement and incentivization of inventions, grant of inventorship to machine inventors would, in turn, reward the inventors of such machines.
The EPO rejected the application stating that the inventor must be a natural person. As per the European Patent Convention ("EPC"), a patent shall contain the designation of the inventor and states thus, “However, if the applicant is not the inventor or is not the sole inventor, the designation shall be filed in a separate document. The designation shall state the family name, given names and full address of the inventor”, and clearly indicates that the inventor is envisaged as a natural person only. The EPO further clarified that “Al systems or machines can be neither employed nor can they transfer any rights to a successor in title”, while acknowledging Dr. Thaler’s ownership over the output of DABUS, as its owner. It clarified that since a machine is designated as the inventor the application “does not meet the formal requirements under the EPC (Article 81, Rule 19(1) EPC) and further that things cannot claim rights, even if they are ‘named’.”
The applications filed at the UKIPO and the USPTO [iii] were also rejected, on technical grounds and formality requirements. In the UK, this decision was appealed to the High Court of England and Wales, Special Patents Court, where Dr. Thaler lost. The court clarified that since DABUS is not a natural person, it cannot own rights, and therefore cannot transfer them to Dr. Thaler as well. Referring to Section 7 and 13 of the Patents Act and on an ancillary note, Section 3.05 of the Formalities Manual of Intellectual Property Office (IPO), the court rejected the possibility of an AI entity being an inventor. The Formalities Manual was also updated in October 2019, though the IPO denies any connection to the DABUS position.
The American grounds for denial of patent application were also technical and procedural – that the process of ‘conception’ of an invention was the exclusive domain of human creativity. Additionally, the USPTO denied the application holding that the datasheet was not filled duly with names of the natural persons who alone can be listed as inventors. It had issued a Notice to File Missing Parts, stating that the inventors should be identified by their ‘legal name’. While the final judgment in the appeal filed by Dr. Thaler [Thaler v. Iancu, et al. (No. 1:20-cv-00903)] is pending, indications do not seem too bright on the success of his position.
To sum up, in all these jurisdictions, the claim of Dr. Thaler was that the patent must be granted as applied because, while Dr. Thaler is a person, he’s not the inventor; and simultaneously, while DABUS is the inventor, it is not a person. This argument was not accepted. However, it is pertinent to note that all patent jurisdictions do not formally insist on a natural person being named as the inventor, for e.g., Israel,[iv] where both DABUS and Dr. Thaler are listed as applicants, while the former alone is named as the inventor.
3. AI INVENTORS : INDIAN PERSPECTIVE
The PCT National Phase Application for the DABUS inventions was filed in India, published under Section 11A of the Patents Act, 1970 on March 19, 2021. In India, even though ‘true and first inventor’ is not qualified in the legislative language, in practice, the inventor is a natural person, i.e., an individual or a group of individuals and an assignee may be a natural person or a legal person such as a company, university, etc. Section 7(3) states that if the inventor is not the applicant, a declaration of proof of inventorship by the inventor must be filed. This theme is carried forth in Sections 10 and 28 as well, which detail the significance of the declaration of an inventor as well as the mention of the inventor as such in the patent, respectively.
As in the case of DABUS, if it becomes necessary for the individual owner of the AI system to designate himself as the inventor to circumvent or overcome potential hurdles of opposition from the Patent Office, Section 64(1)(j) becomes crucial – which states that if the patent was ‘obtained on a false suggestion or representation’, it may be revoked. Sections 25, 52, and 64 also emphasize the requirement for full and correct disclosure of the background and details of the invention, including names of inventors. It is noteworthy that the Indian legal system recognizes that a patent not only confers economic rights but also that the actual creator must be credited with attribution and acknowledgement of his inventiveness and creativity. Evidence may be drawn from the Justice N. R. Ayyangar Committee Report, 1959,[v] which states the following about an inventor in a patent:
“The principle is that whether or not the actual deviser has a proprietary right to the invention, he has a moral right to be named as the inventor. Such a mention besides affording him mental satisfaction gives him prestige and increases his economic worth, advantages to which he is legitimately entitled…”
While the above excerpt clearly talks about an employee or other human inventors, the moral right of an inventor to be recognized as such in the application is vividly delineated. In this context, a question of whether moral rights may be attributed to machines is redundant; claiming another’s invention (even of a machine’s) as belonging to oneself is immoral. A system of joint inventorship, with both the AI creator and the human creator of the AI, being jointly cited as inventors may be an answer; as in such cases, the requirements of moral attribution may be met. Some authors argue that unless juristic personality is attributed to the AI creator, even the concept of co-inventorship may not be practically applied.
Even so, if as Dr. Thaler claims, the invention occurs completely outside the paradigms of human supervision or inputs, even a partial attribution to the human creator of the AI may not be justifiable, considering the lack of effective contribution in arriving at and formulating the invention. Again, another option maybe to designate the AI as an ‘employee adsimilis’, wherein the AI may be treated ‘like an employee’, on whose inventions, the owner of the AI system can claim ownership. In such cases, it is clear that the traditional tenets of patent laws, as applied to human inventors, is being extrapolated to machine inventors – and as such, they may lack traction. For instance, the requirements of ‘proof of right’ as provided under Section 7(2) and (3) of the Patents Act, 1970 (Form I as well), or duly authenticated assignment deeds, are to be fulfilled if they involve human inventors and it is not reasonable to apply them verbatim in the case of a non-human inventor.
Thus, from the Indian context, the objections inherently present in patent laws must be amended expansively so as to accommodate non-human inventors as well. Particularly, systemic requirements specifying that only natural persons may be inventors may have to be altered and devised along the lines of who the ‘applicant’ to a patent may be, which presently includes ‘legal entities’ as well. Such a scenario would accommodate the possibilities of the AI creator being identified as an inventor, while the human creator/owner of the AI system may be the applicant; the same may be facilitated by relevant legislative intervention, in line with the changing requirements of the times.
Further, the Ferid Allani case[vi] is noteworthy in this regard as it recognizes the role played by AI, Internet of Things (IoT), and Blockchain, in modern human life and expands the scope of protection of computer-related inventions under Section 3(k). For now, an AI entity may not be recognized as an inventor by the Indian Patent Office in the DABUS case; this domain must be monitored for further developments.
4. CONCLUSION
It is pertinent to note that Dr. Thaler’s team nowhere argues that the AI be recognized as the inventor fully endowed with the ability to wield any statutory rights. Instead, the argument is that to avoid the grant of recognition or inventorship to creative machines would either incentivize or force human supervisors, to claim inventorship over the respective inventions. Not only would such compulsion result in wrongful claiming, contrary to the concept of moral rights, but also might invalidate the patent itself or give rise to penal consequences, as the act of wrongful claiming may entail the same. Even if the argument is that AI lacks moral rights, human applicants may be penalized for claiming inventions that are not rightfully their own.
There have been numerous discussions at the policy level, in international, regional, and national echelons, regarding the patentability of AI-creations, including steps taken by the USPTO and UKIPO. Most authors rightly argue that this is just the first sign, the tip of the iceberg, of a revolution. As more such creations are originated and devised independently by AI entities, the legal system must change its viewpoint, as the routine tests of the conception of a patentable idea and inventorship applicable to human creators may not be sufficient for AI inventors. The test of non-obviousness, in particular, may have to be re-assessed and applied in such instances. Presently, the threshold of inventive step/non-obviousness is, what is not obvious to a human expert who is ordinarily skilled in the art, but the same cannot apply to an inventive AI system.
Hence, patent systems may need to apply different parameters of obviousness in the case of AI inventors as improvements to the corpus of knowledge by such non-human creators in particular sectors may not be obvious to human inventors or persons skilled in the art – this may particularly become relevant in the future, wherein the inventive capacities of AI systems may expand dramatically.
There has been significant literature about the veiled nature of AI reasoning,[vii] termed in some contexts as Blackbox technologies, wherein even the programmers may not be aware of the complex and complicated processes the AI entities employ in decision making.[viii] Such incomprehensibility may be relevant in assessing:
Sufficient disclosure and enablement, wherein human applicants may not be able to fully explain the path of arrival at the solution in the patent applied for.
Standards of inventive step or non-obviousness, as comprehensibility and understanding of an AI originator would be different from human originators.
These imply, again, that the patent system as we know it, would have to essentially evolve to accommodate AI-originated inventions. Thus, the Thaler Effect, as we may term it, and the impact of the DABUS case in initiating dialogue and discourses in the global techno-legal arena regarding AI-originated inventions are very significant, notwithstanding their success or failure at different patent offices. As more AI-creations come forth in a multitude of fields, moulding or augmenting existing laws may not be sufficient. The dynamic and forceful tide of independent AI-powered conception and creations might demand a sea-change of our intellectual property laws as well as of our collective understanding.
Endnotes: [i] Gary Smith, The AI Delusion 237 (Oxford University Press 2018). [ii] The applications EP 18 275 163 and EP 18 275 174 were filed in 2018. [iii] Stephen Thaler v. Andrei Iancu, 1:2020cv00903 (08/06/2020), in his official capacity as Under Secretary of Commerce for Intellectual Property and Director of the United States Patent and Trademark Office, and United States Patent and Trademark Office. [iv] Erika K. Carlson, Artificial Intelligence can Invent but not Patent—For Now, (November 2020) Engineering 6(11):1212-1213, DOI:10.1016/j.eng.2020.09.003. [v] Shri Justice N. Rajagopala Ayyangar, A Report on the Revision of the Law in India Relating to Patents for Inventions, 1959 (India), Para 432, Clause 22—Mention of inventor as such in patent, available at https://ipindia.gov.in/writereaddata/Portal/Images/pdf/1959-_Justice_N_R_Ayyangar_committee_report.pdf. [vi] Ferid Allani v. Union of India, 2019 SCC Online Del 11867. [vii] Michael O'Neill & Lee Spector. Automatic programming: The Open Issue? Genetic Programming and Evolvable Machines, Springer Nature (2020), available at https://link.springer.com/article/10.1007/s10710-019-09364-2. [viii] Yavar Bathaee, The Artificial Intelligence Black Box and the Failure of Intent and Causation, 31 Harv. J. L. & Tech. 889 (2018).