Emotional perception reaches the top: Supreme Court delivers landmark ruling on AI patentability

The UK Supreme Court has now handed down its much-anticipated judgment in Emotional Perception AI Ltd v Comptroller-General of Patents, Designs and Trade Marks^[1], bringing to a close a case which has charted a turbulent course through the English courts. The decision has significant implications for the patentability of computer-implemented inventions in the United Kingdom. The decision can be accessed here.

CMS was pleased to represent The Chartered Institute of Patent Attorneys and IP Federation as interveners in this appeal and equally pleased that the Supreme Court found the intervention “helpful”, having agreed with much of the analysis included in the intervention. Many others provided valuable support on this collaborative project.

Background to the dispute

The invention at the heart of this case is a deep learning AI system designed to recommend songs to users based on both semantic properties (such as mood descriptors like happy, sad, or relaxing) and musical properties (such as tone, timbre, and beats per minute). The system employs two artificial neural networks ("ANNs"): one trained on semantic properties and one trained on musical properties. The innovation lies in the combination of these two ANNs, with the similarity scores of the musical ANN being adjusted to incorporate the results of the semantic ANN, and vice versa. This process produces a single ANN capable of recommending songs that are similar in both a musical and an emotional sense. The system can be implemented either in hardware form or through software emulation.

The journey through the courts

The case has seen contrasting outcomes at each stage of its judicial journey.

The UKIPO initially refused Emotional Perception's patent application, holding that the invention was a computer program "as such" and therefore excluded from patentability under section 1(2)(c) of the Patents Act 1977. The Hearing Officer concluded that the invention's contribution amounted to no more than a computer program, with no technical effect beyond the running of a program on a computer.

On appeal, the High Court took a markedly different view. Sir Anthony Mann held that the invention was not a computer program at all and therefore the exclusion under section 1(2)(c) was not engaged. He drew a distinction between the computer program used to train the ANN and the trained ANN itself, which simply applies its own weights and biases to produce outputs without relying on instructions. Even if the invention were to be characterised as a computer program, Mann J considered (obiter) that it made a technical contribution through its provision of improved song recommendations, drawing on the reasoning in Protecting Kids the World Over^[2].

The Court of Appeal reversed this decision. Lord Justice Birss, giving the unanimous judgment, adopted broad interpretations of "computer" and "computer program", holding that an ANN qualifies as a computer because it processes information, and that the weights and biases which determine how information is processed constitute a set of instructions, and therefore a computer program. Significantly, the Court of Appeal held that there was no distinction for these purposes between hardware and software implementations of ANNs. On the question of technical contribution, the Court concluded that the beneficial effect of recommending songs based on semantic similarity was subjective and cognitive in nature, rather than technical. The invention was therefore held to be a computer program "as such" and excluded from patent protection.

Our commentary on the High Court and Court of Appeal decisions can be found here and here.

The issues before the Supreme Court

The Supreme Court granted Emotional Perception permission to appeal, presenting an opportunity for the highest court to provide authoritative guidance on the patentability of AI and machine learning inventions (specifically as they relate to ANNs). The central questions before the Court included whether an ANN (and its weights and biases) constitutes a "computer program" for the purposes of section 1(2)(c) of the Patents Act 1977/article 52 of the EPC, whether there is any meaningful distinction between hardware and software implementations of ANNs, and whether the invention makes a technical contribution sufficient to take it outside the exclusion for computer programs "as such".

The Supreme Court's decision

In a unanimous judgment, the Supreme Court allowed Emotional Perception's appeal and set aside much of the Court of Appeal’s decision. The Court's reasoning addressed three principal issues: whether the established Aerotel^[3] approach to assessing patentability should be overturned in light of G1/19^[4]; whether an ANN constitutes a "program for a computer" within the meaning of article 52(2)(c) of the EPC; and whether the claimed invention is excluded from patentability as a computer program "as such".

Aerotel and G1/19

In Aerotel, the UK Court of Appeal set out the steps for applying the exclusions from patent protection contained in section 1(2) of the Patents Act 1977, namely:

1. Properly construe the claim;
2. Identify the actual contribution;
3. Ask whether it falls solely within he excluded subject matter; and
4. Check whether the actual or alleged contribution is actually technical in nature.

The Supreme Court’s decision has overturned Aerotel, in light of the EPO’s Enlarged Board of Appeal decision in G1/19. Despite concerns raised at the appeal hearing that such a decision would require the UK also to overturn the UK’s approach to inventive step in Pozzoli^[5], the Supreme Court disagreed, citing the UK’s approach as a legitimate practice in light of the reading of the EPC in its proper context. Although Pozzoli remains good law, it may require some modification in the future. Its reasons for overturning Aerotel included:

1. It misinterpreted the meaning of “invention”: the EPC, read in its proper context uses “invention” as a threshold concept, to be assessed before inventive step and novelty. Aerotel, however, treated an “invention” as subject matter that is new and inventive.
2. It rejected the “any hardware” approach expressly endorsed in G1/19.
3. It reversed the intended order of analysis required by article 52 EPC. The correct order is to decide whether the claim is an “invention” (i.e. the “any hardware” test), then assess novelty and inventive step only in relation to features with technical character.

In light of G1/19, the Supreme Court held, the approach that should be adopted in the UK is that contained in the Duns^[6] principles (other than principle G), as outlined below:

1. Article 52 of the EPC sets out four requirements: that there is an “invention”, novelty, inventive step, and industrial applicability.
2. Having technical character is an implicit requisite of an “invention”.
3. Article 52 does not exclude any subject matter or activity having technical character, even if it is related to the items listed in the provision as these are only excluded “as such”.
4. The four patentability requirements are separate and independent.
5. The claim must be construed to determine the technical features of the invention.
6. It is legitimate to have a mix of technical and non-technical features, even if non-technical features form a dominant part of the claimed subject matter. Novelty and inventive step can only be based on technical features, however; non-technical features are excluded.

This approach is to be supplemented by the Enlarged Board’s further reasoning in G1/19 regarding the need for an “intermediate step” between the “any hardware” test under section 1(2) and the assessment of inventive step, to determine which features in a “mixed” invention (i.e. one with both technical and non-technical features) contribute to its technical character.

The Supreme Court provided the following analysis in relation to the “intermediate step”:

1. This is not simply an aspect of EPO practice (like its problem/solution approach to inventive step). It is a necessary part of the process of applying (at least) articles 52-56 of the EPC because the “any hardware” invention threshold is too low to enable the invention to be tested for novelty and inventiveness.
2. The subject matter of an invention is patentable only to the extent that it is technical in character and not excluded by article 52(2). This in turn requires an analysis of the subject matter of an invention to ensure that only eligible features are considered in the assessment of inventive step.   
3. The requirement is to carry out this analysis by reference to the technical character of the invention, viewed as a whole. Only when that has been identified can it be seen whether any particular feature of the invention does or does not contribute to it, or interact with it. Although the discussion of the intermediate step is couched in terms of the technical solution of a technical problem, the EPC does not mandate the EPO’s problem and solution approach. That is simply a point of EPO practice, meaning it is open to the UKIPO and to the UK courts to adopt any appropriate method of identifying the technical character of the invention, viewed as a whole, as the necessary first part of carrying out the intermediate step.

The Court noted that whether its conclusions have any repercussions for and give rise to any need to adjust UK methodologies currently used for assessing novelty or inventive step is not something about which it received submissions; that would be left to the UK IPO hearing officer.

Is an ANN a “program for a computer”?

The Supreme Court held that the Court of Appeal’s definitions of “computer” and “program for a computer” (outlined above) were too broad. It defined a “program for a computer” as a set of instructions capable of being followed by a computer (of any kind – whether digital, with a CPU, or otherwise) to produce desired manipulations of data. As an ANN is an abstract model/entity, rather than a physical object, it could not be a “computer”.

The Court continued: “These manipulations of data are “predictable” in the sense that they are determined by the instructions which the computer is “programmed” to follow. They may well not be predictable in the sense that the results of the manipulations can be foreseen in advance. Much of the utility of computers derives precisely from their ability to perform calculations and produce results that far exceed the capabilities of any human brain.” The ANN, in the Supreme Court’s view, “constitutes, in essence, a set of instructions to manipulate data in a particular way so as to produce a desired result. In other words, an ANN is a program for a computer.”

Applying the above to the facts, the hardware that implements the ANN is the computer, and all the features of the ANN (not just weights and biases) constitute a set of instructions to a computer to carry out computations on data presented to it, so are a program for a computer.

Is the claimed invention a program for a computer “as such”?

Although ANNs of the kind with which this appeal was concerned are “programs for computers” within the meaning of article 52(2)(c), the claimed invention is not a program for a computer “as such”. The Supreme Court explained:

“Although the claimed method involves an ANN which is a program for a computer, it also involves technical means because the ANN can only be implemented on some form of computer hardware. In addition, the claims refer to a database for storing data files, a communications network and a user device - all of which require or constitute hardware. That is sufficient to show that the subject matter of the claims has technical character and is not to a computer program “as such”.”

The Supreme Court declined to conduct the "intermediate step" analysis required by G1/19, which filters out non-technical features before assessing inventive step. The Court noted that this step had not been addressed in the courts below and that it had received no significant submissions on how it should be conducted in relation to the claims at issue. The case was therefore remitted to the UK IPO to consider whether the invention satisfies the requirements of novelty and inventive step. 

CMS comment

The Supreme Court's decision represents a significant shift in UK patent law. By abandoning the Aerotel approach and adopting the EPO's "any hardware" test, the Court has brought the UK test for excluded matter towards the EPO test for excluded matter. This will be a welcome development for patent applicants who can now expect greater consistency between patents obtained via the EPO and those obtained directly from the UK IPO. Validity of EP(UK) patents is also strengthened. Applicants who have received “no search” decisions from the UK IPO on pending applications can consider requesting their application be searched and examined using the new test. Importantly, the familiar Pozzoli approach to assessing inventive step remains valid although the Supreme Court notes “Whether the conclusions reached on this appeal … give rise to any need to adjust the methodologies currently used in the UK for assessing … inventive step … will have to be left to a case where it arises”.

However, uncertainty remains. The Court declined to provide guidance on the "intermediate step" required by G1/19, which filters out non-technical features before assessing inventive step. Further guidance from the UKIPO and the courts will be needed on this point. For applicants in the ANN and wider AI field, the practical effect is nonetheless positive: claims will not be automatically excluded merely because they involve computer programs, and the key question will be whether such inventions can demonstrate novelty and inventive step once non-contributing features have been filtered out.
 

^[1] Emotional Perception AI Ltd v Comptroller-General of Patents, Designs and Trade Marks [2026] UKSC 3

^[2] Re Protecting Kids The World Over (PKTWO) Ltd [2011] EWHC 2710 (Pat)

^[3] Aerotel Ltd v Telco Holdings Ltd [2007] RPC 7

^[4] Bentley Systems (UK) Ltd/Pedestrian Simulation (Decision G1/19) [2021] EPOR 30

^[5] Pozzoli SpA v BDMO SA [2007] EWCA Civ 588

^[6] Duns Licensing Associates (Decision T 154/04) [2004] EPOR 10