DEVELOPMENT OF LEGAL JURISPRUDENCE OF AI INCORPORATION IN PATENT LAWS
Artificial Intelligence (AI), a field of study conceived in Dartmouth College in 1956, disrupted the legal paradigm for over three decades. The legal issues and uncertainties posed by it have pushed international organizations and nation-states to gauge its possible effects and reimagine traditional structures to govern the same. Intellectual Property Rights has not been immune to its disrupting effects either. Recent events have shown that jurisdictions have adopted a conservative approach towards granting 'inventorship' status to AI. This article seeks to analyze the crossroads at which AI and Patent Law meet and evaluate whether AI should be granted inventorship. In this article, the author argues that AI should be designated as inventors for their inventions, depending on whether the invention in question is AI-generated or AI-assisted. World Intellectual Property Organization (WIPO) has acknowledged the importance of recognizing AI as inventors. WIPO believes that such a step would incentivize innovation. Part I of the article will discuss the basic requirements to be fulfilled by an invention to qualify as a patent. The following section of this article proposes 'minimal human intervention as a standard for judging whether an invention is AI-generated or AI-aided/assisted.
Basic Requirements to secure a Patent:
A patent is a legal instrument that recognizes that an inventor has some exclusive rights over their invention. It also has the social obligation of enriching society with the knowledge over which it exercises exclusive rights. In I Think, Therefore I Invent, author Ryan Abbott highlights that inappropriately acknowledging inventive activity by machines weakens moral justifications for patents by allowing individuals to take credit for work they have not done. It is unfair to other inventors because it devalues their accomplishments by altering and diminishing the meaning of inventorship. Therefore, it is necessary to recognize the true inventor to promote innovation.
Article 27(1) of the TRIPS agreement defines what patentable subject matter is and states that the essential requirements for an invention to receive a patent are- (i) Novelty, (ii) non-obviousness and (iii) industrial application of the invention in question. Human interference is not an essential criterion for granting patents. Moreover, scholars have highlighted that the patent is granted based on the novelty and utility of the end product or the process and who has put in the effort to produce that novel end product is immaterial.
The indispensable role of technology (specifically of computer programs) in the inventive process has been recognized in Ferid Allani v Union of India, where the Delhi HC acknowledged and recognized the need to innovate in AI, blockchain and other digital products. Technology has advanced to the extent that today, AI is inventing on its own. Machines, including AI, will always be instrumental in developing inventions.
To determine whether the end patent should be granted to the AI or the person responsible for creating the AI, the standard of minimal human interference should be applied. The AI can either autonomously arrive at a novel idea or be automated to create an invention. Therefore, to ascertain whether the invention is AI-assisted or AI-generated, it will be necessary to determine the level of human interference involved in arriving at the novel idea.
The non-obviousness criterion has also been identified as a challenge for inventions created by AI. It has been argued that if machines are creating inventions, then everything becomes obvious, making the practical application of the 'person-skilled in the art standard redundant. Now, a person skilled in the art would have specialized knowledge on the following – 1. Types of problems encountered in the art, 2. Prior art solutions to those problems, 3. The rapidity with which inventions are made, 4. The sophistication of technology, 5. The educational level of active workers in the field. To include machines as a part of the inventive process, he suggests the addition of the list as mentioned earlier should also include 6. Technologies used by active workers. The fact that machines are already augmenting workers' capabilities, in essence, makes it more prominent and expanding the scope of the prior art. Once inventive machines become the standard means of research in a field, the test would also encompass inventive machines' routine use by skilled persons. Taken a step further, once inventive machines become the standard means of research in a field, the skilled person should be an inventive machine. Specifically, the skilled person should be an inventive machine when the standard approach to research in a field or a particular problem is to use an inventive machine.
AI-Assisted & AI-Generated Inventions
AI, as it stands, does not have a universally accepted definition. AI can be defined as a machine with human-like thinking or a computerized system exhibiting behaviour requiring intelligence. It can also be defined as a system designed for rationally solving real-life problems or taking appropriate actions to achieve a goal in the given circumstances.
AI can further be divided into two broad categories- Artificial Narrow Intelligence (ANI) which focuses on solving a specific task. On the other hand, Artificial General Intelligence (AGI) exhibits an intelligence level compared to the human mind. We are currently living in times where AGI is not yet a reality. It may exist in the distant future, but currently, our lives are dominated solely by ANI. A case that becomes extremely important to be discussed in this context is DABUS, and an AI patented by Stephen Thaler. Although DABUS cannot be classified as AGI, it is claimed that it autonomously came up with an invention – fractal container and neural flame. DABUS has been defined as a Creativity Machine, a type of connectionist artificial intelligence. The first level of the artificial neural network trains the Machine with general information from various domains. The second level critiques the first level and evaluates whether an idea is sufficiently novel than the Machine's pre-existing knowledge base. It selects the idea that has the most utility, novelty and value.
Stephen Thaler, the owner of DABUS, filed patent applications for DABUS to be designated as the inventor before various Patent Offices in different jurisdictions. All such applications were rejected. The DABUS team claims that the ideas for invention have wholly come from the AI system and not from a human using AI to solve a particular problem. Dr Thaler had argued since DABUS came up with the idea by itself and assessed its utility and novelty on its own, it can qualify as an investor. The Patent Controller and the High Court rejected the application only due to the natural and legal personhood criteria. The decision of UKIPO was appealed. In the case of Stephen Thaler v Comptroller-General of Patents, Designs and Trademark, the High Court affirmed the decision of UKIPO. It did not grant DABUS the status of an inventor on identical grounds.
Policymakers are constantly trying to evaluate whether the prevailing patent law regime is suitable for AI-generated inventions. Daria Kim highlighted that the deliberations occur by ignoring the most vital consideration. That is the absence of a static definition to evaluate computational autonomy during the inventive process. A corollary to the above issue will be gauging the degree of human interference permissible for AI-generated inventions. Human interference, as a criterion, can be applied to distinguish AI-generated inventions from AI-assisted inventions. Therefore, an extremely important task for academia, policymakers and practitioners is to define the fine line between AI-generated and AI-aided/assisted inventions. Should DABUS be classified as AI that assists invention only because the data is provided by its inventor, despite arriving at the novel idea and its utility without any human interference? Alternatively, the author suggests minimal human interference could be the standard to decide whether an invention is AI-generated. Minimal human interference could be restricted to only providing data to the AI. The AI learns and unlearns to arrive at a novel idea. Therefore, if the definition of minimal human interference is adopted as suggested above, it would help distinguish between AI-generated and AI-assisted inventions and ultimately help inappropriately awarding inventorship to the true inventor.
Technology is advancing at lightning speed. Today, patented inventions are becoming a source of generating novel ideas and evaluating the utility and value of inventions so generated. To promote innovation and uphold human inventors' morality, it becomes imperative to recognize the actual inventor. Although we have not achieved AGI, AI today can autonomously generate new ideas and evaluate their utility. The author personally thinks that the UK High Court was presented with an excellent opportunity to expand the jurisprudence on AI-generated patents, but it chose to restrict the evaluation of such inventors on the question of personhood. Despite the missed opportunity, this case shall also be the first point of discussion as and when the jurisprudence around AI-Patent law advances. The patent regime should not wholly change but should be modified to accommodate AI-generated inventions.
Title Image: Medium
This article has been written by Arya Pachisia. Arya is aa IVth year law student pursuing B.A., LL.B. (Hons.) at Jindal Global Law School.