23/05/2024
Following a recent Judgment from the UK High Court, Emotional Perception AI Ltd v Comptroller-General of Patents, Designs and Trade Marks [2023] EWHC 2948, the United Kingdom Intellectual Property Office (UKIPO) has updated their guidelines for examining patent applications relating to artificial intelligence (AI), and in particular artificial neural networks (ANNs).
The update represents a shift in how the UKIPO examines patent applications that relate to ANNs, and confirms that, at least for now, the computer program exclusion should not be engaged for inventions involving ANNs.
However, it is worth noting that the UKIPO has appealed the Judgment, and there is every chance that it may be overturned in due course; until such time, the UKIPO is bound by the Judgment. Innovators in the AI space will therefore be watching with a keen eye to see how the UKIPO’s approach to examining inventions involving ANNs develops over rest of the year.
Further discussion of the Judgment can be found in a previous Reddie & Grose article.
Background
Under UK law, patents may be granted for computer programs and computer implemented inventions (CIIs), including those relating to AI, providing certain conditions are met. To assess the patentability of CIIs in the UK, the UKIPO refers to certain tests (Aerotel/Macrossan and the AT&T signposts). Provided the invention meets the requirements of these tests, a patent may be granted.
Summary of Guideline Updates
Artificial Neural Networks (ANNs)
Following the guidelines update, inventions relating to ANNs are not considered to involve a computer program, and therefore the computer program exclusion should not be invoked. This is true of both ANNs implemented as physical hardware as well as ANNs which may be emulated using software.
However, the guidelines quite clearly set out that to qualify as an invention involving an ANN, an applicant must “either claim an ANN itself or include claim limitations to training or using an ANN” [1]. If an applicant fails to claim the invention in this manner, the Emotional Perception judgment does not apply and the computer program exclusion may be engaged. This could occur in situations where a claim is directed to generic machine learning models or training of said models without being limited to an ANN use case. Ultimately, the decision of whether to limit a claim towards an ANN or the use thereof will be case-specific and depend on the risk of the computer program exclusion being engaged by a claim that is not limited as such.
Once it has been established that an invention relates to an ANN, and is therefore not subject to the computer program exclusion, the usual hurdle of establishing that the invention makes a technical contribution must be overcome. Following the Emotional Perception Judgment, the court confirmed that moving data outside a computer system using a trained ANN can provide the requisite technical effect needed to meet the UKIPO’s tests.
By virtue of the various use cases for ANNs, it is worth noting that other exclusions from patentability may still apply to inventions relating to ANNs even if the computer program exclusion does not. In particular, without any application, ANNs will be considered as purely abstract mathematical models and therefore fall foul of the mathematical method exclusion. Similarly, if an ANN is implemented such that its functions do not extend beyond a method of doing business, the business method exclusion will be engaged.
Core AI Inventions Involving ANNs
The discussion of core AI in the UKIPO guidelines has also been updated following recent decisions to clarify that core AI inventions involving ANNs are also not subject to the computer program exclusion. Particular example scenarios have been provided including:
- Optimising a neural network;
- Avoiding unnecessary processing using a neural network; and
- Active training of a neural network.
Other Updates
Although the majority of changes to the guidelines are directed to the examination of inventions involving ANNs following the Emotional Perception Judgment, some amendments have been introduced to confirm further instances where an AI invention makes a technical contribution.
In the decision BL O/296/21 (Imagination Technologies), the combination of a general-purpose computer and a deep neural network (DNN) accelerator was found to have a technical contribution as the invention improved the efficiency of image data processing.
Similarly, in the decision BL O/0007/23 (Teledyne FLIR Commercial Systems), the hearing officer confirmed that the use of a neural network (trained with synthetic images) to improve the performance of object identification and classification in images substantiated a technical effect as image object detection was “a specific task external to the computer”.
These decisions reinforce well-established UK case law and are prime examples of instances where patents have been granted in the UK for inventions that involve AI.
Conclusion
In view of the rapid advances in AI over the last decade, it is perhaps unsurprising that the UKIPO’s approach to inventions involving AI has been reviewed. At present, the UK approach diverges somewhat from the European approach. Following the Emotional Perception Judgment, the UKIPO has removed an obstacle to obtaining granted patents for inventions relating to ANNs. However, the Emotional Perception Judgment has been appealed, and there is every chance that the UK approach will change again. Until then, innovators will have to watch this space.
Here at Reddie & Grose we have a large number of attorneys in our electronics and software team with interest and experience in artificial intelligence. If you need help protecting your innovations in this field then please do get in touch.
[1] Guidelines for examining patent applications relating to artificial intelligence (AI) – GOV.UK (www.gov.uk)
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.
