The UK Intellectual Property Office (UK IPO) gave a timely seminar on Artificial Intelligence on 9th January 2020. The speakers for this seminar were the head and senior examiner of the data processing group at the UK IPO.
There were no real surprises that for AI inventions to be patentable they must fit around the exclusions (set out in Section 1(2) of the UK Patents Act) as interpreted by the guidelines (see below). However, the UK Examiners did indicate how the nature and/or presentation of the AI invention could lead to very different results, stressing that for borderline cases they are keen to engage with applicants and listen to technical arguments for patentability.
The seminar follows a report released by the UK IPO into inventions relating to Artificial Intelligence released last year. See here for our earlier report.
In the examination of a patent application, the UK IPO assess the technical contribution of the claim relying on the test from the Aerotel & Macrossan decision of the Court of Appeal  EWCA Civ 1371], as well as the AT&T signposts set out in the AT&T Knowledge Ventures LP and CVON Innovations decision  EWHC 343.
The Aerotel / Macrossan test provides a framework for approaching the assessment of claims relating to computer implemented inventions, and requires that an Examiner: (1) properly construe the claim; (2) identify the actual contribution (to the art); (3) ask whether [the contribution] falls solely within the excluded subject matter; finally, a check is required to determine (4) whether the actual or alleged contribution is actually technical in nature.
Similarly, the AT&T signposts point to contributions that are not excluded subject matter and that can be technical, such as i) whether the claimed technical effect has a technical effect on a process which is carried on outside the computer; ii) whether the claimed technical effect operates at the level of the architecture of the computer; that is to say whether the effect is produced irrespective of the data being processed or the applications being run; iii) whether the claimed technical effect results in the computer being made to operate in a new way; iv) whether there is an increase in the speed or reliability of the computer; and v) whether the perceived problem is overcome by the claimed invention as opposed to merely being circumvented.
Categories of AI Inventions
The seminar outlined the three categories that AI inventions typically fall into: 1) applications of AI to particular fields or problems, 2) techniques for training the AI, and 3) the core AI itself (how the AI is implemented).
The UK IPO observed that applications of AI to particular technical fields would often be patentable, under the first or second signposts, in particular where the AI controlled a process outside of a computer, or operated at the level of the computer architecture and therefore improved the computer as a tool. Similarly, techniques for training and calibrating AI, would be treated in the same way as training and calibrating any computer system, and would likely be patentable, if the programmed AI was intended to control a technical process, or if the training relied on real-world data from sensors.
The patentability of core AI would be more problematic, however, if the programmed AI (without reference to its use or manner of training) was presented merely as a computer program or mathematical method as such, independently of any technical application. In this regard, the UK IPO did acknowledge that they closely follow the guidance of the EPO Guidelines G-II, 3.3 “Technical Implementations” which states that such mathematical methods may still contribute to the technical character of an invention, “when the claim is directed to a specific technical implementation of the mathematical method and the mathematical method is particularly adapted for that implementation in that its design is motivated by technical considerations of the internal functioning of the computer”. In such circumstances, the burden would be on the applicant to ensure that it is explained clearly in the description of the application, and in correspondence with the examiner, why there is a technical contribution. This is important as it may not be immediately apparent from the abstract nature of the claimed invention.
UK IPO Guidelines
Presently there are no UK court decisions directly relating to the patentability of artificial intelligence subject matter, and the UK IPO continue to examine patent applications on a case-by-case basis, applying the existing law relating to computer implemented inventions. In this regard, the EPO Guidelines G-II, 3.3 mentioned above provide a useful reminder of AI applications that are considered to be technical, including:
a) controlling a specific technical system or process;
b) determining from measurements a required process step in a machine implemented process;
c) digital audio, image or video enhancement or analysis;
d) separation of sources in speech signals and speech recognition;
e) encoding data for reliable and/or efficient transmission or storage;
f) encrypting/decrypting or signing electronic communications;
g) optimising load distribution in a computer network;
h) determining the energy expenditure or the body temperature of a subject or processing data obtained from physiological sensors;
i) providing a genotype estimate based on an analysis of DNA samples, as well as providing a confidence interval for this estimate so as to quantify its reliability;
j) providing a medical diagnosis by an automated system processing physiological measurements; and
k) simulating the behaviour of an adequately defined class of technical items, or specific technical processes under technically relevant conditions.
It was also mentioned in the seminar that the Manual of Patent Practice will be updated on the 1st April 2020 and is expected to include further guidance on patentability of AI inventions. We welcome this and look forward to further clarification of the UK IPO’s position.
Although there is no case law at the UK IPO relating to AI, a brief search of the patent database looking into the statuses of UK patents under the IPC classification G06N3/08, which relates to means and methods for machine learning or training neural networks, shows that of the applications so classified around 18% of published applications have been granted in total. A look at the broader classification G06N shows similar figures. In comparison, according to the statistics published by the UK IPO, 54% of published patents were granted in the UK in 2017 overall, and 47% of published patents for computer technology were granted in the UK in the same year.
The seminar was overall a positive insight into how the UK IPO is examining patents relating to AI. The main take away was that while the UK IPO (and EPO) apply a strict test to the patentability of AI related inventions, the test is clearly set out for the benefit of applicants and examiners alike in the Manual of Patent Practice and EPO Guidelines. Thus, there should be certainty at an early stages as to whether an AI related invention is patentable.
At Reddie & Grose LLP, our AI and IoT group have a broad range of experience in obtaining patent protection and providing tailored advice to companies of all sizes. If you would like information or advice for protecting your invention, please get in touch.
This article is for general information only. Its content is not a statement of the law on any subject and does not constitute advice. Please contact Reddie & Grose LLP for advice before taking any action in reliance on it.