A follow-up to CIPO’s New Subject-Matter Guidance — and the Structural Problem It Can’t Fix
In an earlier post, I, argued that the controversy over software patent eligibility in Canada and the United States flows from a structural mismatch: an inherently unstable distinction between the abstract and the concrete is pressed into service as a binary, ostensibly objective question of law, when in practice it is resolved through the intuition of the decision-maker rather than an evidence-based analysis. I argued that Canadian law already contains the tools to fix this — by confining subject-matter analysis to a bright-line exclusion of pure abstraction and relocating the substantive evaluation of computer-implemented inventions to the obviousness stage.
Turns out the UK Supreme Court seems to have done exactly what I argued should be done here to fix software patents. On February 11, 2026, the UK Supreme Court issued its judgment in Emotional Perception AI Ltd v Comptroller General of Patents (Emotional Perception AI Limited (Appellant) v Comptroller General of Patents, Designs and Trade Marks (Respondent) – UK Supreme Court and, in effect, did exactly that — for the UK.
Background to the Decision
The invention at issue was a deep learning system that recommended media files — music tracks, for example — by training an artificial neural network to align objectively measurable physical properties of files with the subjective emotional similarity that human beings perceive between them. The UK Intellectual Property Office had refused the application on the ground that it was excluded from patentability as a program for a computer. The High Court allowed the appeal. The Court of Appeal reversed. The Supreme Court then reversed the Court of Appeal.
The legal question was whether the long-standing UK framework for assessing software patent eligibility — the four-step Aerotel test — correctly implemented Article 52 of the European Patent Convention, in light of a 2021 decision of the EPO Enlarged Board of Appeal in G1/19.
The Structural Problem with Aerotel
The Aerotel framework required courts to identify the “actual contribution” of a claimed invention — what the inventor has really added to human knowledge — and then ask whether that contribution was technical in nature. If not, the claim was excluded.
This should sound familiar. It is the same structural problem I identified in the Canadian and US context: a threshold inquiry that is ostensibly about subject matter but is in practice resolved by asking whether the invention feels sufficiently technical, which is an intuition-driven exercise dressed up as a legal test. The EPO Enlarged Board said as much in G1/19, characterizing the Aerotel approach as incompatible with the EPC and not a good-faith implementation of it. The specific criticism: Aerotel conflated the question of whether something is an “invention” with the separate, independent questions of novelty and inventive step, by defining contribution as what the inventor has “really added to human knowledge.” That imports novelty analysis into a threshold question that the EPC requires to be answered independently.
The UK Supreme Court accepted that criticism and overruled Aerotel unanimously.
The Framework That Replaced It
The replacement framework has three stages, and its architecture will be familiar to readers of my earlier post.
First, a low threshold for statutory subject matter. If a claim involves technical means — any hardware at all — it clears the eligibility hurdle. An artificial neural network that can only be implemented on computer hardware clears it. The threshold question is answered quickly and cleanly, without any assessment of contribution, novelty, or inventive step.
Second, an intermediate filtering step. Once the threshold is cleared, the features of the claim are examined to identify which ones contribute to the technical character of the invention as a whole. Non-technical features that do not interact with the technical subject matter are filtered out. This is a feature-by-feature dissection, not a holistic characterization of the invention.
Third, an evidence-based assessment of novelty and inventive step, applied only to the features that survive the filter. The real analytical work — the question that actually determines patentability — is done here, anchored in the prior art, structured reasoning, and a record subject to appellate review.
The Court was careful to preserve the existing English approach to inventive step under Pozzoli rather than mandating the EPO’s problem-and-solution method. The point is that the EPC does not prescribe the vehicle for the inventive step analysis. What it prescribes is that the analysis happen there, not at the subject-matter threshold.
This Is the Argument I Made for Canada
The structural principle is identical to what I argued Canadian law already supports. Subject-matter analysis should perform a bright-line function at the extremes — excluding pure abstraction — without attempting to resolve the harder question of whether a computer-implemented invention deserves a patent. That harder question should be resolved under obviousness, in a structured, prior-art-anchored, evidence-based manner. Canada can do that work through Beloit/Sanofi, informed by Allergan’s adoption of Pozzoli: strip away unnecessary verbiage, focus on the claim’s material difference from the prior art, and ask whether that difference was obvious.
The specific vehicle differs by jurisdiction — problem-and-solution at the EPO, Pozzoli in the UK, and Beloit/Sanofi in Canada. That is a local implementation choice, not a structural divergence. All three are evidence-based. All three require engagement with the prior art. All three produce a record subject to structured review. The principle — relocate the analytical work from intuition-driven threshold to evidence-based merits — holds across all three.
Where This Leaves Canada
The UK was operating under a framework that the EPO Enlarged Board had explicitly condemned as a misinterpretation of the EPC. It took a leap-frog appeal to the Supreme Court to fix it. Canada is not bound by the EPC and faces no equivalent external constraint, but the structural diagnosis is the same and the doctrinal tools for the same fix are already present in Canadian law.
What the UK Supreme Court has now confirmed is that the intuition-driven contribution test — the software patent equivalent of the abstract/concrete binary — is not just impractical. It is analytically unsound. It asks the wrong question at the wrong stage using the wrong tools. The evidence-based approach is not merely a policy preference. It is what a properly structured patent system looks like.
CIPO’s recent subject-matter guidance reflects genuine effort. But as I argued in my earlier post, guidance cannot resolve a structural problem. The UK just resolved theirs. The path for Canada to do the same, within existing doctrine, remains open.
