One “intermediate step” to rule them all? The technical character question for patentability of AI and computer-implemented inventions in the UPC, UK and EPO
November 16, 2026
COMVIK and G 1/19 “technical character” as a shared filter, and what happens when inventive step is not assessed using the EPO’s problem–solution approach.
Europe has long grappled with the exclusions from patentability based on Article 52(2) EPC. These include presentations of information, programs for computers, mental acts and methods of doing business, among others.
The interesting recent development is not a change to those categories, but a growing convergence in how their effect on patentability is assessed: through a structured inquiry into which features of a claim contribute to its technical character.
One illustration is the UPC Court of Appeal’s decision in Abbott v Sinocare.1 There, the Court expressly endorsed the logic of the EPO’s decisions in COMVIK and G 1/19: a feature should not be excluded from the inventive-step assessment merely because it is non-technical in isolation. It may still contribute to technical character through its interaction with other features, meaning that the interrelationship and combined functioning of the claim features must be considered.
At the same time, the UK has undergone a more fundamental change. In Emotional Perception AI Ltd v Comptroller-General of Patents [2026] UKSC 3, the Supreme Court held that the previous Aerotel framework should no longer be followed. Instead, the UK should adopt principles A to F set out in Duns,2 drawing on COMVIK and endorsed in G 1/19.3 The UKIPO subsequently published guidance on how the new approach should be applied during search and examination.
At a high level, this looks like European harmonisation in action. The EPO, UPC and UK are converging on a common core question: which features contribute to the technical character of the claimed invention as a whole?
That does not mean that the three systems have adopted identical methodologies. In particular, the UPC and UK do not generally apply the EPO’s problem–solution approach to inventive step. The significance of the emerging convergence is narrower but still important: the jurisdictions increasingly share a feature-filtering discipline, even though the technical features that survive that filter may then be assessed for inventive step in different ways.
The low first hurdle: “any hardware”
Under the approach now adopted in the UK, the first hurdle is a low one. Under the “any hardware” test, a claim qualifies as an invention if it is directed to, or uses, technical means. For a computer-implemented invention, a reference to a computer, computer-readable storage medium or other hardware will ordinarily be sufficient. This assessment is made without reference to the prior art.4
The EPO and UK now have clear guidance on this first hurdle. The UPC has not yet been required to address it expressly. Its endorsement of G 1/19 suggests that it may take a similar approach, although that remains to be tested.
This low threshold does not mean that including conventional hardware makes an otherwise non-technical advance patentable. It merely establishes that the claim is an invention. A further filtering exercise is needed because only features that contribute to the technical character of the invention may support inventive step. The low “any hardware” threshold and the intermediate step are therefore two parts of a single structured assessment: the breadth of the former makes the latter necessary.
The “intermediate step”
At the intermediate step, the decision-maker identifies which features contribute to the technical character of the invention as a whole. Features that do not make that contribution cannot support inventive step.
Importantly, the exercise does not simply involve classifying each feature as inherently technical or non-technical. The analysis must be performed feature by feature, but by reference to the technical character of the invention as a whole. A feature that appears non-technical in isolation may nevertheless contribute through its interaction with other features. The inquiry must therefore consider the interrelationship and combined functioning of the claim features.5
This is essentially the discipline established by COMVIK and elaborated in G 1/19. What is new is its express adoption in the UK and its emerging application outside the EPO’s problem–solution framework.
Two early UKIPO applications of the new approach
The first UKIPO decisions applying the new approach demonstrate that the intermediate step is likely to become an important battleground for computer-implemented inventions.
Emotional Perception on remittal
The application in Emotional Perception concerned an artificial neural network (‘ANN’) trained to recommend files, particularly music files, by identifying semantic similarity from their physical properties. Following the Supreme Court’s judgment, the application returned to the UKIPO for the remaining stages of the analysis.
In October 2026, the hearing officer refused the application. In the decision, the hearing officer characterised the technical character of the invention as “Computer hardware running a program (the ANN) that selects and outputs a file” having concluded that outputting a “semantically better” file was not a technical result. The details of the ANN’s learning objectives and its attempt to close the “semantic gap” therefore did not contribute to the technical character of the claimed invention. Once those features were filtered out, what remained was obvious over common general knowledge.6
The decision illustrates the potentially decisive effect of the intermediate step. A claim may meet the initial requirement for an invention because it includes computer hardware, yet still fail because the features said to embody the advance do not contribute to technical character and therefore cannot support inventive step. The critical question was whether the claimed computational features contribute to a technical purpose or effect, rather than merely producing an output with greater semantic or cognitive value to the user.
FLIR Belgium BV
The UKIPO’s decision in FLIR Belgium BV provides a useful contrast. The application concerned a dynamically zoomed virtual gauge that displayed the current value and a selected part of the gauge’s overall range.
The examiner had treated a number of features as not contributing to technical character. The hearing officer disagreed in relation to several of them, finding that features concerning the operation and display of the virtual gauge contributed to technical character when considered in the context of the invention as a whole. The hearing officer did not, however, accept every display-related feature: the claimed panning operation was held not to contribute and remained filtered out.7 Because inventive step had not been assessed by reference to the correct set of contributing features and a full Pozzoli analysis had not been carried out, the application was remitted for further examination.
Read together, the two decisions show why the intermediate step cannot be reduced to categorisation. In Emotional Perception, the semantic quality of a recommendation was not a technical result capable of bringing the ANN’s training features into the technical contribution. In FLIR Belgium BV, by contrast, some display-related features contributed because of their function and interaction within the claimed system, although others did not.
Where does Abbott v Sinocare fit?
Against that background, Abbott v Sinocare is best treated as a useful but limited indication of the same direction of travel.
The disputed features concerned the display of event icons alongside sensor-derived glucose data. The UPC Court of Appeal rejected the suggestion that they should simply be excluded as a presentation of information. The Court considered the functioning of the features together and concluded that they contributed to improved human-machine interaction, helping the user to manage diabetes and achieve glycaemic control.
That reasoning is consistent with the interaction-based inquiry now required in the UK: a feature is not excluded merely because it appears informational or non-technical when viewed alone. Its contribution must be assessed in the context of the claimed combination.
The decision should not, however, be made to bear too much weight. It arose in preliminary injunction proceedings and concerned a relatively discrete challenge to a particular group of user-interface features. It does not provide a complete methodology for assessing technical character and contribution of features in more complex AI or computer-implemented claims.
Its narrower significance is that the UPC, alongside the UK, has expressly looked to COMVIK and G 1/19 when determining which features may contribute to technical character, while retaining its own downstream inventive-step methodology.
Harmonisation, but only up to a point
It would therefore be an overstatement to say that European approaches to excluded subject matter and inventive step have been harmonised.
The EPO continues to apply the technical-character filter within its problem–solution approach. The UK must now integrate the intermediate step with its own obviousness principles. The UPC has taken an initial step by endorsing the COMVIK/G 1/19 interaction analysis, but continues to develop its own approach to inventive step and must similarly integrate these principles.
Nevertheless, there is meaningful convergence at the filtering stage. Across the three systems, the inquiry is moving away from asking whether a feature can be labelled, in isolation, as a computer program, presentation of information or mental act. The more important question is whether, in the context of the invention as a whole, the feature interacts with the technical subject matter in a way that contributes to technical character.
Questions about the operation of the new UK approach also remain. In particular, the precise method for identifying the technical character of the invention as a whole is not prescribed. The Supreme Court left the UKIPO and lower courts free to adopt an appropriate method, suggesting that identifying the alleged inventive concept under Pozzoli might be suitable in some cases.8 What is required is a transparent, feature-by-feature analysis, followed by an assessment of inventive step under established UK principles.
The sequencing nevertheless matters. As the UKIPO guidance currently envisages, the intermediate step should identify which features are eligible to support inventive step before obviousness is determined. Otherwise, the conclusion that a feature does not contribute to technical character risks becoming blurred with the separate conclusion that it is obvious.
Conclusion
The recent decisions do not establish a single European test for excluded subject matter and inventive step. They do, however, reveal increasing convergence on a common analytical discipline: identifying which features contribute to the technical character of the invention as a whole, including through their interaction with other features, before inventive step is assessed. Emotional Perception and FLIR Belgium BV show how consequential and fact-sensitive that inquiry can be in UK practice. Abbott v Sinocare provides a useful UPC comparison, but should be understood as an early illustration of the shared principle rather than a definitive harmonising decision.
The emerging direction is nevertheless clear: the analysis turns not on the label attached to a feature, but on what it does, how it interacts with the rest of the claim and whether that interaction contributes to the technical character of the invention as a whole.
- 1Abbott Diabetes Care Inc v Sinocare Inc and A Menarini Diagnostics srl, UPC_CoA_901/2025, Order of 17 April 2026, Headnote 1 and para 112
- 2T 0154/04
- 3Emotional Perception AI Ltd, UKSC, paras 65–67
- 4Emotional Perception AI Ltd, UKSC, paras 98–100
- 5Emotional Perception AI Ltd, UKSC, paras 101–110, in particular paras 106–109.
- 6Emotional Perception AI Ltd, BL O/0894/26, UKIPO, 1 October 2026, paras 30–43, in particular paras 37–56
- 7FLIR Belgium BV, BL O/0673/26, UKIPO, 29 July 2026, paras 44–50
- 8Emotional Perception AI Ltd, UKSC 3, paras 106–107
You may also like