Artificial intelligence (AI) is the latest technology to disrupt the application of intellectual property law concepts. This article is Part 1 of 3 considering IP law challenges in the age of AI and focusses on inventorship in patent law.
What is AI?
AI refers to computers which are ‘intelligent’ in the sense that they can learn through processing large sets of data and independently generate rules. This allows an AI system to deal with information it has not previously encountered by applying those rules. For example, an AI system can learn to recognise faces in this way (for a brief explanation of AI see The Economist, ‘Rise of the machines’ (2015) available here).
AI can be divided into two categories: ‘applied’ (sometimes referred to as ‘specific’) AI and ‘general’ (sometimes referred to as ‘strong’) AI. Applied AI refers to specialised systems which are designed to solve a certain type of problem or perform a particular task, such as medical diagnosis. In contrast, general AI refers to a system which could in theory handle any task, possessing the flexibility and levels of intelligence akin to humans. This type of AI has not been realised (yet) and therefore this article is written primarily with applied AI in mind.
Who is the inventor of AI-created inventions in patent law?
AI systems are already capable of creating inventions independent of human intervention. For instance, applied AI systems can be used to produce innovative designs or methods. AI was responsible for the cross-bristle design of the Oral-B CrossAction toothbrush which had increased efficiency over the prior art.
UK jurisprudence defines an inventor as the person(s) responsible for the inventive concept of an invention (see IDA and Ors v University of Southampton and Ors  EWCA Civ 145). It is therefore unclear how patent law might accommodate AI-created inventions where the inventive concept is produced by artificial intelligence and not a human being.
In answer to the question who is the inventor of AI-created inventions, UK patent law could: (i) require a human individual to be named as the inventor; (ii) allow AI systems to be named as inventors; (iii) determine that there is no recognised inventor, rendering the invention unpatentable; or (iv) determine that there is no recognised inventor but nonetheless allow inventions to be patented.
Option (i) raises difficulties because it would require a new test to identify the inventor in situations where AI is responsible for the inventive concept of an invention. Aside from the difficulties that this may raise, it would lead to the undesirable situation of differing interpretations of the concept of an inventor within patent law.
Ryan Abbott, in a 2017 article titled ‘I Think, Therefore I Invent: Creative Computers and the Future of Patent Law’ argues for option (ii), that AI systems should be recognised inventors on the basis that this would encourage innovation and the commercialisation of inventions. However, the more commercially and practically significant question remains: who is entitled to exploit the patent? The default rule under section 7(2) Patents Act 1977 is that it is the inventor. However, an AI system is currently not recognised as a legal entity, it cannot licence or assign rights.
Option (iii) would appear to be least optimal. Denying patentability for AI-created invention would be likely to increase the use of trade secrets and thereby restrict publicly available information on technological developments.
Option (iv) therefore emerges as the straightforward approach. It it avoids the artificial task of identifying a human inventor for AI-created inventions. Furthermore, option (iv) goes straight to the crucial question of ownership.