This article is the first in a series of articles on artificial intelligence (AI) written by Carpmaels & Ransford’s dedicated AI team. This article provides an overview of the current legislative environment, while future instalments in the series will look in depth at some of the key industries at the forefront of AI technology.
The Covid-19 pandemic has forced businesses and institutions across the world to adopt new technologies at unprecedented rates. Technology has enabled businesses to continue operating and essential services to keep running. For the pharmaceutical industry in particular, AI has played a vital role in the race to develop new vaccines. However, innovation in many fields of technology is supported and encouraged by the availability of intellectual property rights to secure investment and ensure return on that investment. Difficulties in IP acquisition for AI inventions therefore have the potential to threaten technological progress, and development of the framework for protecting AI inventions is required.
At the beginning of 2021 the UK’s AI Council, an independent government advisory body, published its AI Roadmap which called for increased public sector investment in AI technology as well as further support for UK AI start-ups. Challenge-led innovation is promoted in the Roadmap, which calls for people to work across boundaries and existing organisational structures, and to build new relationships, networks, and common languages to develop entirely new AI solutions. The Roadmap also proposes that continued investment in basic research into the fundamentals of AI is required. According to the Roadmap, AI systems often rely on a narrow disciplinary base, which needs to be broadened to cover an array of domains in an intelligent way.
One of the issues surrounding AI inventions is the determination of inventorship, since it is possible that patentable technology could be derived by AI-based machines. In September 2021, we reported on the UK Court of Appeal’s rejection of an appeal in a case concerning the rejection of patent applications filed at the UK Intellectual Property Office (UKIPO) on the basis that an AI-based machine known as “DABUS” (Device for the Autonomous Bootstrapping of Unified Sentience) was listed as an inventor. The European Patent Office (EPO) similarly refused two patent applications filed by the same applicant on the grounds that they do not meet the legal requirement of the European Patent Convention (EPC) that an inventor designated in the application has to be a human being, and not a machine. Both EPO decisions have been appealed and the EPO’s Boards of Appeal is due to hear both cases in December 2021.
However, the UKIPO recognises that the IP system must enable society to protect innovation even if that innovation is supported by inventive and creative AI. An open consultation has therefore been launched by the UKIPO to allow it to understand in more detail the issues surrounding AI and intellectual property with respect to patents and copyright protection. In particular, the UKIPO is consulting on:
- Copyright protection for computer-generated works without a human author. These are currently protected in the UK for 50 years. But should they be protected at all and if so, how should they be protected?
- Licensing or exceptions to copyright for text and data mining, which is often significant in AI use and development.
- Patent protection for AI-devised inventions. Should we protect them, and if so, how should they be protected?
It will be interesting to see the outcome of the UKIPO’s consultation and the decisions from the EPO’s Boards of Appeal, since the UKIPO could be the favoured office for protecting AI-based inventions if the UKIPO makes changes to allow AI-based machines to be identified as inventors, for example.
What is the approach to patenting AI inventions at the EPO and the UKIPO?
In the most recent Guidelines for Examination at the EPO, there is little change to the approach to examining AI-related inventions. According to the EPO, AI inventions are generally of an abstract mathematical nature since they are based on computational models and algorithms and as such are not eligible for patenting. The Guidelines do infer, however, that the use of technical means may give rise to a technical character.
In order to have technical character, software, or computer implemented inventions, must produce a “further” technical effect when run on a computer. A “further” technical effect is a technical effect going beyond the “normal” physical interactions between the software and the computer on which the software is run. Such a further technical effect should be explicitly identified and described in a patent specification. By contrast, the normal physical effects of the execution of software are not sufficient to confer technical character on the software. For example, simply claiming the implementation of a conventional process with software based on known software methodologies will not be viewed favourably.
The Manual of Patent Practice of the UKIPO includes an explanation on how AI inventions might be handled at the UKIPO. According to the UKIPO, inventions involving AI are generally computer-implemented such that their assessment in respect of excluded matter will mirror that of any computer-implemented invention. Helpfully, the UKIPO has stated that if an AI invention is directed to a specific technical process outside of a computer or if it forms part of the internal workings of a computer, then it will likely not be excluded. In contrast, where an AI invention is directed to an excluded process, and there is nothing more to it, it will likely be excluded by the UKIPO, which is true also for an AI invention directed to a non-specific purpose.
Patenting various categories of AI invention
AI inventions may be divided into three different categories, each of which have varying requirements for demonstrating the technical character necessary for patentability. The following discusses the potential challenges and possible solutions when applying for patent protection for AI inventions within those three categories.
- Core AI technology
New configurations of AI processes and methods may be patentable so long as a technical effect resulting from these new configurations can be identified. For example, the application of neural networks to various fields of technology, such as medical applications or image/video processing, has technical character. The technical character is likely to be in the processing of technical data (e.g. data generated using medical applications or image/video processing). Processing abstract data without an identification of a technical use may not be patentable.
- Training
Training a neural network, even if the neural network being used is known, might be patentable. An example of this would be the modification of the training steps of a neural network for identifying regular heartbeats that results in a neural network which enables fewer false positive identifications. Focusing on the modifications to the training or the specific form of the input training data might be helpful when drafting patent applications in this category.
- Application of AI to specific fields and sectors, or use of AI in devices
The application of known AI methodologies to a specific field or sector may be patentable but is likely to be challenging to pursue. For example, it may be difficult to establish an inventive step when applying a known AI methodology to a new data set from a different field for which the AI methodology was conceived. There may be approaches to claim drafting in such situations, which focus on the source of the input data, or the way in which the input data is generated or pre-processed, to improve the eligibility position.
Other forms of IP
Obtaining a patent for AI inventions is not without its challenges and enforcing a patent can be equally difficult. Furthermore, there are commercial reasons why in some circumstances patents may not be the best form of protection for the methodologies and large data sets typically employed in AI based systems. It is therefore worth considering other forms of IP protection that are available, including trade secrets or confidential information, copyright, and database rights.
Summary
Protecting the value created by AI technology can be challenging but by considering IP strategies early on in the development process, it is possible to mitigate these challenges. Focusing patent claims on the innovative technical aspects of how an AI system is built or deployed will help to ensure eligibility, while additional forms of IP such as copyright can help reinforce protection. The Carpmaels & Ransford team has a wealth of experience in this field, working with innovators to create effective and comprehensive IP strategies that will protect innovations in the long term.
Over the coming months, we will be publishing a series of follow up articles looking at the different issues that should be considered when protecting AI technology. As well as looking in further detail at the various ways in which AI technology can be protected, we will also be considering individual technology sectors, and how the fundamentals of AI technology can be best protected. Consideration will also be given to how innovators can collaborate whilst still protecting their investment.
