- within Intellectual Property topic(s)
- within Intellectual Property, Transport, Media, Telecoms, IT and Entertainment topic(s)
- with Inhouse Counsel
- with readers working within the Law Firm industries
In a landmark judgment on AI and software patents, the Supreme Court has reset the UK approach to excluded subject matter by overruling the long‑standing Aerotel test and aligning UK law with the EPO's interpretation of Article 52 EPC (Emotional Perception AI Ltd v Comptroller General of Patents [2026] UKSC 3, 11 February 2026).
This is the first time the Supreme Court has considered the patentability of artificial neural network (ANN)‑implemented inventions. It follows contrasting decisions in the High Court and the Court of Appeal, which had taken fundamentally different views on whether (and how) an ANN fits within the established framework for excluded subject matter. We covered the High Court decision here and the Court of Appeal's decision here.
It's fair to say that most observers were expecting the decision to focus on the narrower questions of specific application to this case, or more generally whether ANN's were "computer programs" as meant by Article 52. However the Supreme Court has gone much further in overturning the approach to Article 52 which has been followed by UK Courts since 2006.
In summary, the Supreme Court has held that:
- the prevailing Aerotel approach to the test for excluded matter should no longer be followed in the UK in light of the decision of the EPO Enlarged Board of Appeal (EBA) in G1/19 which should instead be followed (the so called "any hardware" test);
- ANNs of the relevant kind are "programs for computers" so the computer program exclusion is engaged; but
- under the correct EPC approach (in G 1/19), the claims in this case are not excluded - they do not relate to a computer program "as such" (a pre-requisite of the exclusion) because they involve "hardware" and technical means.
- Thus the subject‑matter qualifies as an "invention" (ie is not excluded) but must now be assessed for novelty and inventive step before the UKIPO.
In doing so, the Court confirmed that eligibility should be assessed using the EBA's "any hardware" approach, with a separate intermediate step to identify which claim features contribute to technical character before novelty and inventive step are considered. The Court stressed that this did not change the UK's approach to inventive step (as set out by the Supreme Court in Pozzoli) but commented that "it may require some modification".
HSF Kramer IP & Tech partner, Peter Dalton, was quoted today in Juve Patent on the implications of this decision:
"Whether this results in an aligning of the UK approach to inventive step with that of the European Patent Office remains to be seen. Should the patent be allowed, the impact on the AI industry could be major, allowing AI operators to seek monopoly protection for up to 20 years over elements of AI functionality and implementation. Indeed, the claims in this case reference certain fundamental concepts of LLM implementation."
Andrew Moir, one of HSF Kramer's tech patent litigators, comments:
"Given that computer-implemented inventions almost by definition will involve hardware (i.e. the computer), the effect of the Supreme Court's decision is effectively to squash the previous excluded subject matter test into inventive step: only those features that contribute to the technical character of the invention viewed as a whole can form part of the inventive step."
IMPLICATIONS OF THE SUPREME COURT'S DECISION
An immediate impact of this decision is that software (including AI software such as ANNs) that incorporates hardware in its claimed implementation will now not be automatically rejected as unpatentable subject matter. However, this does not mean that such inventions will necessarily proceed to grant more easily as applications will still need to clear the usual novelty and inventive step tests.
The Supreme Court's decision is significant for two reasons:
- First, it brings UK practice into closer alignment with the EPO by adopting the EPO Enlarged Board's approach in G1/19. The findings of the Supreme Court have wider implication for patent law in general in terms of the direct application of EBA decisions in any field of technology.
- Second, while the Court held that ANNs are "programs for computers", it also makes clear that this does not end the analysis: on the EPO's "any hardware" eligibility threshold, computer‑implemented claims will generally clear the "invention" hurdle, and the key work shifts to identifying which features contribute to technical character and then to novelty/inventive step.
In practice, this is likely to change how AI/ANN patentability arguments are run. The eligibility hurdle is low, but the Supreme Court highlights the need for an "intermediate step" to filter which features contribute to technical character before assessing novelty and inventive step. For applicants and their advisers, that reinforces the importance of drafting specifications and claims so that the technical character of the invention (viewed as a whole) and the interaction of the relevant features with the technical means are clearly articulated.
THE INVENTION AND HOW THE DISPUTE AROSE
The patent application concerned a system for recommending media files, typically illustrated by music. Unlike conventional recommendation systems that rely on human‑defined categories such as genre, the invention used a trained ANN to identify files that are "semantically" similar based on how media is perceived and described.
Put simply, the system trains the neural network by comparing pairs of files and learning how their measurable features relate to the way people describe and perceive them. Once training is complete, the trained network can identify similar files and recommend them to users.
The UK Intellectual Property Office refused the application as excluded subject matter (being a "program for a computer...as such") under section 1(2)(c) of the Patents Act 1977. That refusal led to appeals to the High Court, the Court of Appeal and, ultimately, the Supreme Court.
LEGAL FRAMEWORK
Under section 1(2)(c) of the Patents Act 1977, a "program for a computer ... as such" is excluded from patentability. The Supreme Court confirms that these provisions are to be approached by working "directly from the underlying source" i.e.: Article 52 EPC (specifically Art 52(2)), consistent with section 130(7) of the 1977 Act. In doing so it expanded on the approach UK courts should take to EPO decisions:
"The starting point in any analysis of the question whether the Aerotel approach should now be treated as overtaken by G1/19 lies in the principles which courts in the UK have adopted in deciding how closely to follow the case law of the EPO on the meaning and effect of the EPC. On this, the authoritative dicta speak with one voice. These courts are not strictly bound by decisions of the Boards of Appeal of the EPO, even of the Enlarged Board, but they should respect and follow the Enlarged Board's decisions and any uniform jurisprudence of the Boards of Appeal unless convinced that they are wrong or beyond the ambit of reasonable difference of opinion. This is essentially for good policy reasons. The national courts of all the contracting states should seek a uniform interpretation of the EPC, and of domestic legislation passed to implement it." (emphasis added)
Prior to this Supreme Court judgment, the UKIPO and the UK courts had applied the four-step approach set out in Aerotel Ltd v Telco Holdings Ltd [2006] EWCA Civ 1371 ("Aerotel"), which required the court to consider both whether the claimed invention falls within the excluded subject matter "as such", and if the exclusion is engaged, whether the invention nevertheless makes a technical contribution. A technical contribution could save the invention from being excluded, but purely excluded matter, even if novel or inventive, did not count as a technical contribution.
This often led to diverging outcomes compared to the EPO's approach to excluded matter, which effectively allowed any invention to pass as long as it involved "any" use of hardware (which was often achieved by reference to software "running" on a form of hardware in the claim.
Under the EPO approach now endorsed by the Supreme Court, the threshold question of whether a claim qualifies as an "invention" under Article 52(1) EPC is addressed using the EBO's "any hardware" approach. On this approach, a computer-implemented claim will generally have technical character, and therefore clear the eligibility hurdle, if it involves technical means, such as hardware. The Supreme Court emphasised that this is a low threshold, and this is not the stage where one considers whether the invention is new or inventive. Those questions are addressed later, following a separate intermediate step that filters which claim features contribute to technical character before novelty and inventive step are assessed.
THE SUPREME COURT'S DECISION
The Supreme Court framed the issues before it in three parts.
- First, whether the UK courts should continue to follow the Aerotel approach in light of the EBA's decision in G1/19 ("Issue 1").
- Secondly, whether an ANN is (or contains) a "program for a computer" within article 52(2)(c) EPC ("Issue 2").
- Thirdly, if the computer program exclusion is engaged, whether the claims are nevertheless excluded "as such" ("Issue 3").
Issue 1: Aerotel is no longer the correct approach
The Court held that the Aerotel approach "should not be followed" in light of G1/19. In essence, it considered that the EPO's approach is based on the correct interpretation of Article 52 and that, as a matter of comity and consistency in EPC interpretation, the UK should follow the EBA unless convinced it is wrong beyond reasonable disagreement. The Court also accepted the EPO criticism that Aerotel risks mixing the "invention" inquiry with novelty and inventive step by focusing on "contribution" to the art, contrary to the EPC's structure.
The Court held that the UK should adopt the EBA's interpretation of Article 52 reflected in G1/19, replacing the Aerotel framework with a staged analysis. In practical terms this means:
- Eligibility / invention: A computer‑implemented claim will generally qualify as an "invention" if it involves technical means — the EBA's "any hardware" threshold — which the Court described as a low hurdle.
- Intermediate step: Before novelty and inventive step are assessed, decision‑makers must apply an "intermediate step" which was part of the G1/19 decision, identifying on a feature‑by‑feature basis which elements of the claim contribute to the technical character of the invention as a whole, and filtering out those that do not. The Court stressed that this is materially different from Aerotel's holistic "technical contribution" check, which asked whether the invention's contribution, viewed overall, was technical in nature.
- Novelty / inventive step: Only those technical features are then assessed for novelty and inventive step. The Court emphasised that this does not require wholesale adoption of the EPO's problem–solution method: the established UK approach to inventive step set out in Pozzoli remains applicable.
Issue 2: ANN is a "program"
On Issue 2, the Court rejected the applicant's primary case and concluded that ANNs of the relevant kind are "programs for computers" within article 52(2)(c). While it disagreed with the Court of Appeal's description of an ANN as itself a "computer", it held that an ANN is an abstract model which can be implemented on different types of hardware, and that the machine implementing it functions as a computer. The Court then held that, viewed as a whole, an ANN constitutes "in essence, a set of instructions to manipulate data in a particular way so as to produce a desired result" and therefore "an ANN is a program for a computer."
Issue 3: The claims are not excluded "as such"
On Issue 3, having decided that the ANN is a computer program, the Supreme Court needed to consider whether it was one "as such": "to consider whether the entire subject matter of the applicant's claims falls within the scope of the exclusion of "programs for computers ... as such" or whether it includes technical features and so qualifies as an "invention" within the meaning of article 52(1). Applying the EPO "any hardware" approach, the Court held that the claims are not excluded "as such" because they necessarily involve technical means (hardware implementation of the ANN) and also refer to hardware elements such as a database, a communications network and a user device.
The Supreme Court therefore allowed the appeal; the UKIPO was wrong to refuse the application on the basis that it was excluded subject matter, and the case was remitted for the UKIPO to consider novelty and inventive step applying the post‑G1/19 approach.
CONCLUDING THOUGHTS
In many quarters this may be seen to be a welcome simplification when it comes to cross border patenting of software between European states and the UK. The Supreme Court emphasised that while it was endorsing the "low hurdle" approach to Article 52(2) when considering the exclusion for software as such, such patents would still need to be considered at the novelty and inventive step stage, where excluded matter would be excluded when interpreting the technical character of the invention. The EPO has not traditionally been seen as substantially more permissive than the UK IPO when it comes to the granting of patents relating to software implemented inventions, and as such it may be that this decision does not substantially impact patent outcomes. However, it does introduce the potential for improved patentability outcomes given that the automatic exclusion is less likely to apply, which would have far reaching impacts for both AI and software more generally.
To date competition between AI developers has largely been commercial and technical rather than rooted in patent protection. However the patent in question in this case includes claims which refer to fundamental concepts and techniques used by most LLM models. If there were to be an uptick in patent rights being granted in respect of these techniques, this would have significant impacts on the AI race currently underway.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
