In brief
- The UK Supreme Court has handed down its much anticipated decision in Emotional Perception AI Limited v Comptroller General of Patents, Designs and Trade Marks. The decision changes how patent applications and granted patents relating to artificial neural networks (ANNs), AI, and computer implemented inventions (as well as other categories) are to be assessed.
- The Aerotel test for assessing patent eligibility is gone, and is replaced by the lower hurdle of the "any hardware" approach applied by the European Patent Office for many years.
- An immediate and practical benefit of this decision is that UK patent applications should now always be substantively searched and examined, when previously they would likely not have been as a result of the Aerotel test. This is a boon for UK first filers who wish to rely on the UKIPO's search and examination to inform future patent and filing strategy.
- Whether the decision significantly impacts what ultimately gets patented in the UK is more nuanced and largely remains to be seen. In particular, the court left open a number of important questions, pushing the detailed assessment back to the UKIPO Hearing Officer for consideration.
Background
The specific system of the patent application (published as GB2583455) uses an ANN to replicate the similarities and differences which a human being might subjectively perceive between different files of information by reference entirely to objectively measurable physical properties of those files. The commercial purpose is to enable a person who likes a particular media file, such as a music track, to receive a recommendation of a file determined to elicit a similar emotional response.
The UKIPO's Hearing Officer initially rejected the application on the basis that it fell foul of the exclusion from patentability of "programs for computers... as such" under section 1(2) of the Patents Act 1977 (which reflects Article 52(2) of the European Patent Convention (EPC)). The Patents Court allowed the applicant's appeal, holding that an ANN is not a program for a computer at all. The Court of Appeal then allowed the Comptroller General's appeal and restored the Hearing Officer's decision. The matter reached the Supreme Court, with the decision handed down on 11 February 2026.
Legal framework
At the heart of the case lies section 1 of the Patents Act and Article 52 of the EPC. These provide that patents shall be granted for any inventions, provided they are new, involve an inventive step and are capable of industrial application. They also list matters not to be regarded as inventions, including discoveries, scientific theories, and mathematical methods; aesthetic creations; schemes, rules and methods for performing mental acts, playing games or doing business; programs for computers; and presentations of information. However, these exclusions apply only to the extent that the application relates to such subject matter "as such". The case before the court concerned the computer program exclusion, but the court's analysis of how the relevant provisions should be interpreted applies generally to all the excluded categories.
Three issues were before the Supreme Court: (1), whether the established Aerotel test for excluded subject matter remained good law; (2), whether an ANN is a "program for a computer"; and (3), whether the claimed invention as a whole was excluded from patentability.
Decision
Issue 1 — The end of Aerotel
This first issue has the widest implications. It determines how the excluded subject matter provisions in the legislation are to be applied, not only in relation to computer programs but to all the listed excluded categories, and it affects any applicant whose invention lies in or around those categories.
The Aerotel four step test required an examiner to:
- properly construe the claim;
- identify the actual contribution;
- ask whether it falls solely within the excluded subject matter; and
- check whether the actual or alleged contribution is actually technical in nature.
The Supreme Court decided that this test is no longer good law. It adopted the position that had been decided by various EPO Boards of Appeal (most notably in G1/19), that it improperly conflates two separate issues (that of eligibility and inventive step).
In particular, after construing the claim, the Aerotel test goes straight to identifying the actual contribution made by the claim, and then asks whether that contribution, rather than the invention, falls solely within the excluded subject matter. The Court of Appeal in Aerotel had defined "contribution" as what, as a matter of substance, the inventor "has really added to human knowledge." However, that plainly raises considerations of novelty and inventive step, both of which are separate and independent requirements of patentability, a point which the Supreme Court expressly acknowledged in its critique of Aerotel (mirroring similar criticism by the EPO's Boards of Appeal).
By contrast, the EPO's approach in G1/19 squarely addresses whether the claim amounts to an "invention" (i.e. something capable of being protected by a patent) uncluttered by any assessment of contribution, novelty or inventive step. It construes the term "invention" in light of the basic requirement that patents are to be extended only to that which is technical, treating Article 52(2) EPC as providing a non-exhaustive list of examples of matters which are non-technical.
Under this approach, a claim directed to a computer implemented invention, or indeed any invention engaging the exclusions, can avoid exclusion on eligibility grounds merely by referring to the use of a computer, a computer readable storage medium or other technical means. This "any hardware" test does impose a low hurdle, but that is justified by the addition of "as such" to each excluded item by Article 52(3).
The Supreme Court therefore confirmed that steps 2 to 4 of Aerotel could not be maintained. Many had anticipated that the Aerotel test would be abandoned. However, it was also thought that the Supreme Court would further find it necessary to overturn the UK's long-established approach to assessment of inventive step, namely the Windsurfing/Pozzoli approach, given the seeming incongruence between it and the "any hardware approach". That perceived incongruence is borne out of the fact that Windsurfing/Pozzoli again invites a holistic consideration of the inventive contribution, and without reference to the technicality/patent eligibility of that contribution. The concern was that inventions which trivially pass the "any hardware" assessment may go on to also pass the Windsurfing/Pozzoli assessment without actually making any inventive technical contribution.
The Supreme Court however was clear that replacing the Aerotel test with the "any hardware" approach did not also require wholesale adoption of the EPO's problem-solution (in its COMVIK form) approach to inventive step. For now, therefore, the Pozzoli approach remains the method for assessing whether a claim involves an inventive step.
That said, the court established that an intermediate step is required between the "any hardware" threshold and assessment of inventive step. This intermediate step calls for a dissection of the subject matter of the invention into its component features and a review of the contributory role of each feature by reference to its contribution to the technical character of the invention viewed as a whole. The intermediate step is not limited to computer program claims. It applies whenever a claim contains a mix of technical and non-technical features, which will be the case for many inventions.
The intermediate step serves as a filter, identifying which features of the invention contribute to its technical character, excluding those that do not form part of the inventive step assessment. The court described this as a necessary implication of the EPC, not merely a matter of EPO practice, because the low threshold set by the "any hardware" test would otherwise leave the assessment of inventive step without any mechanism for addressing the exclusions in Article 52(2)..
Issue 2 — Is an ANN a computer program?
The court held that, whatever the specific form of the machine on which an ANN is implemented, it constitutes a set of instructions to manipulate data in a particular way so as to produce a desired result. It therefore rejected the applicant's attempt to distinguish a "hardware ANN" from a software emulated ANN. There is no principled justification for privileging one form of implementation over another, the choice between them being a practical one. Software and hardware implementations are the same in terms of architecture, weights and biases, and outputs produced. An ANN is an abstract entity, not a physical object, and the machine on which it is implemented functions as a computer regardless of the specific technology used. Accordingly, an ANN was a computer program.
Issue 3 — Is the claimed invention excluded as a whole?
However, 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. The claims further refer to a database for storing data files, a communications network and a user device, all of which constitute hardware. That was sufficient to show that the subject matter of the claims had technical character and the invention was not therefore to a computer program "as such".
Outcome
Given its decision on issues 2 and 3, the Supreme Court allowed the appeal and set aside the Hearing Officer's decision. The case has now been sent back to the UKIPO for substantive examination. In particular, the Hearing Officer will need to consider how the intermediate step applies to this particular invention, and whether the relevant features can be considered to involve an inventive step.
In the meantime, the UKIPO has confirmed that it has:
- Suspended the published guidance for "Examining patent applications relating to artificial intelligence (AI) inventions"
- Committed to updating the Manual of Patent Practice to reflect the Supreme Court decision
- Confirmed it will align examination practices with the European Patent Office approach
It has also stated that each AI patent application will "continue to be examined on its individual merits" and that it will consider transitional arrangements as it implements the new approach.
Comment
The decision is no doubt a positive development. However, it is worth being clear about what it does and does not achieve. Its significance extends well beyond AI. Any applicant whose invention touches on excluded subject matter, whether that is a mathematical method, a business method, a method for performing a mental act, a computer program, or a presentation of information, will be affected by the change in approach.
The most useful element of the judgment is the removal of the Aerotel test. The UKIPO is, in many respects, a good jurisdiction for first filing. It is relatively cost effective, offers good quality search and examination, and operates with reasonable speed. A first filed UK application can therefore provide valuable early intelligence about prospects of grant to inform a broader global prosecution strategy. However, as a result of Aerotel, applicants in certain technical fields and sectors (e.g. AI) were often denied that intelligence. Applying the Aerotel methodology, the UKIPO would routinely decline to search or examine applications on the basis that the contribution lay in excluded subject matter, leaving applicants with no search report, no prior art analysis and no substantive examination. That has now been addressed.
That said, the impact on overall patentability outcomes is potentially more muted. The battleground has only (very likely) shifted rather than disappeared. Inventions that previously fell at the Aerotel threshold will now instead face scrutiny at the inventive step stage, with the intermediate step (almost certainly) serving as its own filter that ultimately results in the same types of inventions being filtered out. For example, a business method implemented on a computer will now clear the first hurdle but will still need to demonstrate that its technical features involve an inventive step over the prior art. The likelihood of grant for the types of inventions that previously encountered difficulty under Aerotel is therefore not expected to change dramatically, but the debate will be had in a more structured way, with the benefit of a search report and identified prior art, giving both applicants and the UKIPO a clearer basis for discussion. That is a more productive position to be in.
It is also worth addressing the suggestion in some commentary that the Supreme Court has decided that "AI is now patentable" or that "ANNs are no longer excluded." This is an oversimplification. AI-related inventions have been patentable in the UK, and at the EPO, for a considerable time, and patents have been regularly granted for ANN-based inventions where a clear technical effect is demonstrated. The Supreme Court has merely corrected the approach by which such applications are assessed at the eligibility stage; it has not introduced patentability of AI or of any other previously excluded category. The question of what counts as sufficiently "technical" remains the real area of debate.
Where the decision arguably falls short is in its treatment of the consequential issues. The court left the specifics as to what the "intermediate step" entails and how the "any hardware" approach is to be married together with the Pozzoli approach to inventive step unresolved.
Whether the Pozzoli approach can, in practice, comfortably coexist with the intermediate step is an open question. The intermediate step inherently invites a feature-by-feature assessment of the claim, requiring a dissection of the claim into its components and an assessment of each feature's contribution to technical character. The Pozzoli approach is broader, being more concerned with the inventive concept of the claim as a whole. How the two are reconciled in the context of complex inventions where technical and non-technical elements are closely intertwined may well require further judicial consideration.
On the whole, the decision is to be welcomed. Aerotel is gone, the "any hardware" approach now applies in the UK, and applications that may engage the Article 52(2) exclusions will nonetheless receive substantive examination. For now, applicants and practitioners can take reasonable comfort from a ruling that brings UK patent practice into closer alignment with the EPO and that should ensure first filing in the UK delivers the practical value it always ought to have done.