Patent Protection and Artificial Intelligence

by Kanerva Jalas

Intellectual property rights focus on the protection of intangible goods. Most categories of intellectual property such as patent and copyright protection require that the creator of the subject-matter is a human being, or in legal terms, a ‘natural person’. Patent protection is granted for technological innovations that fulfil certain conditions, such as novelty, and inclusion of an innovative step, while copyright aims to protect the creative expressions of the human mind. Paradoxically, computer software has proven useful for inventive creation in all fields of technology. In order to understand what the implications of this is, the technology employed for invention must be defined. Artificial intelligence (AI) refers to computer programs that make use of mathematics. The computer program follows a sequence of commands, and can be programmed to simulate the functioning of a human brain, by requiring the program to make decisions that would seem to require logical reasoning. However, many such results can be obtained by the program through mathematical and computational functions, which do not require cognitive abilities. Due to the broad definition of AI, almost all computer programs can qualify under it. However, key features of a specific software that link it to such intelligence can be said to include independence, efficiency, unpredictability, and creativity. Most AI systems work through machine learning which allows the program to learn from experience and correct its mistakes due to negative response or realisation of failure. Recently, AI programs have gained attention; the development of AlphaGo, a revolutionary computer program which defeated the world champion at the game of Go, notoriously famous for its complexity and requirement of human players.

The conditions for patent protection have similarities worldwide. This post will focus on the conditions established by the European Patent Office (EPO) and the United States Patent and Trademark Office (USPTO). The requirements established in the European area can be found from article 52 European Patent Convention (EPC): a new invention must be technical, novel, include an inventive step and possible for industrial application (commercial use). The convention has importantly established subject-matters which are not patentable. Article 52 (2) (c) EPC enumerates that programs for computers are excluded. This article implies that all inventions including any type of computer programs fall outside the scope of patent protection. Similar legislation can be found in the United States: the invention must be novel, include a non-obvious inventive step, must fall within a patentable subject matter, must be useful and must be properly disclosed. However, such an exclusion of AI from patentable subject-matter can be gone around with an inclusion of a physical part for the invention: article 52 (3) EPC includes that subject-matters which are specifically excluded in the same article can be patented where they are not the main patentable invention, but instead form a part of a technical invention including real-life effect.

Another important requirement for patenting is the inventor of the technical solution: she must contribute to the development of the invention. However, what the concept of an inventor requires remains abstract. The development of technology has resulted in the advancement of computer programs: nowadays, it is possible for software to largely contribute to the finding of a technical solution. Sometimes, such as in the case of the new art generator DALL-E 2, the software at hand is aimed to be the main content creation, and the human merely a supporting one. In the United States, the District Court found that in order for one to obtain a patent, the technical solution must be able to be used in practice, and must be conceived. According to the EPO, the inventor of the solution, as well as the applicant (for the patent) will have the right to enforce patent rights. However, the EPO states clearly that if an application for a patent does not state the inventor, it will not be granted a patent. Where the two jurisdictions differ relates to the definition of an inventor: the EPO nor the national legislations of the EPO Member States do not define the inventor to be a human. Instead, the fact that inventors are natural persons is assumed: the mention of the inventor requires human related details such as surname and residence address. In the United States, a requirement for the inventor for the purposes of acquiring a patent exists under the interpretation of the current statutory legislation. The Patent Act assumes the inventor to be a human, as they are described as an individual or a group of individuals. The Dictionary Act, however, describes the requirement in further, and in relation to all acts of Congress: where the Congress has included the words person or individual, these refer to the species homo sapiens. Furthermore, the Congress has defined the patentable subject matter to include all matters that are made by man.

However, the fast development of AI has come to challenge such provisions. The Artificial Inventor Project aims to reach the goal of appointing an AI as the inventor of a technical solution and acquiring a patent. Such an event would require the existence of AI-generated content: output that is produced by AI with little to no human interference or choice in the creation of the output. Due to this process, AI systems can be described as ‘black boxes’: data or input is fed to the software which starts a process of which even the creator of the AI software is not fully aware of. As a result, inventions that are produced as AI generated output can be argued to have been invented by the computer program, instead of a human. In 2018, Dr Stephen Thaler, the creator of an AI neural network applied for patents in the United Kingdom for the two inventions created by his AI: a food container and an emergency flare. The application was extended to other areas, including the jurisdictions of EPO and USPTO. However, both patent officers as well as those from other countries such as Japan and South Korea have rejected the application on the grounds that AI cannot be designated as the inventor of a technical solution. In December 2021, the EPO issued a decision responding to the appeal of patent rejection for DABUS. The legal board of EPO based their decision on article 81 EPC, requiring that the inventor must be an individual with legal personality, an AI not qualifying for such. The office, however, does not exclude AI generated content to be excluded as patentable subject matter: where such would be wished to be patented, as the creator of the AI can designate themselves as the inventor. A similar decision has been issued by the USPTO: in April 2020 the office denied the possibility of DABUS as the inventor, and reiterated that patents can only be issued where the inventor is a human being. In their decision, the office referred to legislation passed by the Congress where language only referring to natural persons are used. On this basis, USPTO stated that allowing AI to be designated as an inventor would contradict the purpose of such legislation.

So far, all the five largest intellectual property offices (EPO, JPO, USPTO, KIPO and CNIPA) have denied DABUS its inventorship. Arguably, such decisions stem from the unaddressed situation of granting a non-human entity legal rights and personality. However, a change could be upon such a development: recently, in July 2021, a patent for the food container by DABUS was published in South Africa’s patent journal. The neural network was designated as its inventor.

Sources:

European Patent Convention, 17th ed [2020].

An Act to amend and consolidate the law relating to patents (The Patents Act) [1970].

1 The United States Code U.S.C [2018].

Solvay SA v Honeywell (Fed.Cir 2014).

DALL-E 2 <https://openai.com/dall-e-2/> accessed on the 2nd of January 2023.

Pihlajarinne T and others (eds), Online Distribution of Content in the EU (Edward Elgar, 2019).

Kur A and others, European Intellectual Property Law: text, cases and materials (2nd ed, Edward Elgar Publishing, 2019).

Tia Matulionyte, ‘AI as an Inventor: Has the Federal Court of Australia Erred in DABUS?’ (2022) JIPITEC 99.

Schwein R L, ‘Patentability and inventorship of AI-Generated Inventions’ (2021) 60 Washburn Law Journal 561.

Chen J X, ‘The Evolution of Computing: AlphaGo’ (2016) 18 Computing in Science and Engineering 4.