[Guest Post] UK Assessment of Computer Implemented Inventions moves closer to EPO Practice - but by how far? (Emotional perception [2026] UKSC 3)

Now the dust has now settled on the high profile UK Supreme Court decision in Emotional Perception AI Limited (Appellant) v Comptroller General of Patents, Designs and Trade Marks (Respondent) [2026] UKSC 3. Katfriend Greg Corcoran provides in-depth analysis: 

Over to Greg:

"In Emotional Perception, the UK Supreme Court (SC) found a new test should apply in the UK for determining the patent eligibility for computer programs. The long awaited decision appears to be closer to European practice, G1/19, setting aside the established precedent of Aerotel Ltd v Telco Holding Ltd [2006] EWCA Civ 1371 (Aerotel). Yet G1/19 has not been applied carte blanche. Differences remain.

How has the SC decided, what questions remain and what implications could there be in view of the new law?

Hardware ANN

Recap of Proceedings

The question put to the SC was: is an artificial neural network (ANN) a program for a computer, as such? If so, would this prevent the grant of the Appellant’s patent application?

The applicant claimed to have invented an ANN which replaced differences a person may determine between files of information by reference to the physical characteristics of those files.

The UKIPO refused the application as the claimed invention related to a program for a computer which is excluded from patentability under s1(2)(c) UK Patents Act 1977 (PA), implementing Art. 52(2) & (3) European Patent Convention (EPC). After a reversal by the Patents Court, the Court of Appeal affirmed that an ANN is a program for a computer and the claim had no technical contribution (IPKat).

In considering the appeal, the SC identified three issues:

(1) what is the correct standard for matter excluded from patentability in the UK;

(2) under this standard, is an ANN a program for a computer; and

(3) applying that finding, what is the SC’s view of the claims of the patent application at hand.

Issue 1: Standard of Excluded Matter:

The prior standard of Aerotel for assessing patentability of an invention under Art 52 EPC was, Aerotel paragraph 40: (1) construe the claim; (2) identify the contribution;(3) is the contribution solely excluded subject matter; and (4) check if the contribution is technical.

When Aerotel was decided, the Court of Appeal viewed the exclusion from patentability of programs for computers, Art 52 EPC, as unsettled European law, Decision para 17. The court warned that when the Enlarged Board of Appeal (EBA) of the European Patent Office (EPO) decides this question the UK approach may need to be reconsidered. The court proposed a referral question for the EBA. Soon after, the President of the EPO initiated a referral, G3/08. Although the EBA affirmed the then existing EPO practice, the EBA found the referral inadmissible in the absence of split lines of case law.

G1/19, a decade later, approved the “any hardware” approach (G1/19 reason 28): if the subject matter of the claim features hardware, then article 52(2) is inapplicable. Aerotel found this principle was earlier rejected in Gale’s Application [1991] RPC 305; Decision paras 27 and 31. The court passed over Comvik GSM AB T641/00 (COMVIK) which established an approach which considered, in context of the problem-solution approach, just those features of a claim which contribute to the technical character of the claimed invention, Decision paragraph 32.

COMVIK was affirmed in Duns, T 154/04, which set out seven principles, summarised as:

(A) The four requirements of patentability: disclosure of an invention, novelty, inventive step; and industrial applicability.

(B) Technical character is implicit for an invention.

(C) Subject matter having technical character is not excluded from patentability by Art 52(2) EPC even if listed, because the exclusion is ‘as such’.

(D) The four requirements of patentability are separate and independent criteria; novelty, in particular, is not a requisite to an invention.

(E) For examining patentability, the claim must be construed to determine the technical features of the invention.

(F) A claim may have a mix of technical and ‘non-technical’ features. Novelty and inventive step can be based only on technical features that are clearly defined in the claim. Non-technical features are ignored if they do not interact with the technical subject matter of the claim.

(G) A formulation using the problem-solution for assessing inventive step.

COMVIK is applied in steps (F) and (G). The EBA finally affirmed this in G1/19 reasons 31 & 32; Decision paragraphs 39 and 43.

The ‘any hardware’ test for an invention, Art 52 EPC, has a low hurdle, justifiable by the wording “as such”, Decision para 60. In comparison, the SC found Aerotel adopts an incorrect interpretation of Art 52 EPC. Instead of checking that the whole of the claimed invention is technical; it considers the contribution of the claim over the prior art (step 2) and asks if that contribution is technical (step 4). This brings novelty and inventive step into the assessment under Art 52(2) EPC. Yet the EPO (Duns reason 9, 10 & 12) and English courts (Genentech Inc’s Patent [1989] RPC 147, 262) decided these criteria are separate and independent, Decision para 34. The four criteria of patentability, step (D) Duns, does not mention an order, only that they are independent and separate, Decision para 43. Yet the SC considers whether the claim discloses an “invention” must be considered first, ahead of and separately from the other three conditions of novelty, inventive step and industrial application (G1/19 reason 37 to 39 and decision para 61. So Aerotel, unlike Duns and the present decision, jumbles the test for invention with other requirements (Decision para 63).

Although the SC could not ignore that an approach to the EPC that is settled in English law by the Court of Appeal and EBA decisions are not necessarily binding on the UK Courts, e.g. such as when the court is convinced that the EBA has erred (UK Ltd v Merck &Co Inc [2008] EWCA Civ 444; [2009] 1 WLR 1186, para 48) the SC decided to account for EPO practice. Adaption to EPO practice will be disruptive, yet avoidance of disruption cannot be sufficient reason to maintain a different approach inconsistent with EP practice, Decision paras 46 & 48

There is a limit: nonacceptance of the EPO problem-solution approach in its established interpretation of Art 52(2) & (3) EPC. In Actavis Group PTC EHF v ICOS Corp [2019] UKSC 15; [2019] Bus LR 1318; [2019] RPC 9, para 93 the SC decided that this EPO approach is a matter of practice rather than mandated by the EPC and approved the approach to inventive step applied in UK courts as in Pozzoli SpA v BDMO SA [2007] EWCA Civ 588; [2007] FSR 37, Decision paras 33 and 65.

The SC decided to adopt the same interpretation of Art 52 EPC as contained in the Duns but without its last step, (G), because that would necessitate the problem-solution approach. This inserts an intermediate step between the two hurdles of eligibility of patentability, Art 52 EPC, and of inventive step, Art 56 EPC. The intermediate step determines which features of a “mixed” invention contribute to technical character, see G1/19 reasons 37 to 39; Decision paragraphs 43 and 66.

Issue 2: under this standard, is an ANN a program for a computer?

The SC decided an ANN is a program for a computer.

A computer as a ‘machine which processes information’ would be too broad a definition because  it would include devices not used to perform computations such as a basic oven, a smoke detector or an old-fashioned jukebox, Decision para 76. Yet it is unreasonable to confine the definition of a computer to a digital computer, tying a computer to a specific technology and architecture. The EPC is written to encapsulate concepts independent of technological change. A computer can apply any technology from analogue to quantum, Decision para 77. Similarly, computer instructions are not tied to a specific computer architecture or form of instructions. The fundamental characterisation of a computer and a program for such a device is for performance of calculations for producing results that exceed the capabilities of a person, Decision para 79. Thus, an ANN is a program for a computer, even if its functioning is determined by training data not a programmer, Decision para 95; T702/20 reason 16.1.

In response to arguments that exclusion from patentability applies differently in hardware ANNs or software emulations, the SC found these as different embodiments. A hardware embodiment with fixed weightings does not detract from equivalence to a software emulation, even when the weights and biases are frozen after training. The choice of embodiment does not relate to the inventive concept as such, T208/84 reason 16; Decision paragraph 83. An ANN is a computer because of its functionality not because of the technology of its implementation.

In any ANN data presented as input is processed by computations in the network to produce an output, Decision para 86. Whatever the implemented type of machine, an ANN is essentially a set of instructions to manipulate data in a particular way to produce a desired result, Decision para 87.

Issue 3: how does the SC find for these issues for the patent application at hand?

In applying the “any hardware test” the SC found the claims feature an invention because they are technical in referring to hardware features for implementing the ANN.

The SC did not assess the new intermediate step, applying (F) Duns, to assess which features of the claim contribute to its technical character. Any decision by the SC would be precedential, but without contributions of stakeholders i.e. the parties, the UKIPO and judges of lower courts.

Since the intermediate step is new, the SC supplied some guidance:

1) Filtering the features determines the technical subject matter of the invention, ensuring all eligible features of the claim are accounted for in the assessment of inventive step, Decision paras 101 & 105.

2) The technical character of the invention is by analysis of the invention as a whole, Decision para 106.

3) The analysis excludes features of the invention from subsequent consideration that do not contribute to the technical character of the invention, when viewed as a whole.

Although non-technical features are non-inventions if viewed alone, the contribution of such features to the technical character of the invention should be assessed in context of the invention as a whole. Likewise, technical features should be assessed for their contribution to the technical character of the invention as a whole, G1/19 reasons 32 & 33, Decision para 102.

Only those features contributing to the technical character of the claim, as a whole, are considered in inventive step assessment, Decision para 107. Although the intermediate step practiced by the EPO is couched in the context of the problem-solution approach, it is open for the UKIPO and UK courts to adopt any appropriate method to identify the technical characteristics of the invention as a first part of the intermediate step. The SC suggests this could be identifying an inventive concept of the claim as in Pozzoli.

The SC declined to develop further the intermediate step or to apply it to the claims of the case. The basis of the law is different from that used in the lower courts and features in none of the pleadings, Decision para 115. The SC considered the appeal raised a question of fundamental importance for AI but it was inappropriate to develop the new law further. Further comments would create binding precedent. Instead, the SC reverted proceedings to the UKIPO hearing officer where development of new practice can start.

Summary and Observations

These proceedings are a test case. If the hearing officer refuses the application either on the intermediate step or inventive step, an appeal to the courts seems likely. If that appeal fails, future cases and their decisions can be expected.

Central to the intermediate step is what makes an invention - as a whole - technical, accounting for the contributions of each claimed feature (whether or not they themselves are technical). This is a new question under UK law. Guidance exists from EPO practice and case law, e.g. EPO Guidance G.II.3 at 3.3. and 3.6. for which case law appears to be highly context dependent. The recently overruled practice could be reinterpreted and applied, but this risks divergence from EPO practice. It will be interesting to see how the UKIPO and the UK courts develop assessment of the intermediate step, including the requirements for an invention to have technical character, and of inventive step applying Pozzoli (the inventive concept is determined without a view to the prior art). The extent to which EPO practice is adopted without the problem-solution approach will be keenly watched.

Although the UK practice of computer implemented inventions has taken a large step towards harmonisation with EPO practice, without legislation the road ahead to harmonisation will be long, with many more decisions, including even some from the Supreme Court.

European harmonisation also accounts for the UPC. The UPC Court of Appeal has yet to consider the patentability of a computer implemented invention and its standard for inventive step differs from the EPO (Meril v Edwards UPC_CoA_464/2024; Amgen v Sanofi UPC_CoA_528/2024). There could be parallels in how the SC and the UPC approach the assessment of patentability of computer implemented inventions". 

Further reading

• On the threshold of a dream? Patents and computer-implemented business methods in Europe
• High Court of Australia refuses special leave in Aristocrat: what this means for computer-implemented inventions