1.  Why Software Patents Are Legally Contested

Both UK and New Zealand patent law exclude "programs for computers … as such" from patentability. The critical phrase is as such: not every software-based invention is excluded — only those that are nothing more than a computer program. The challenge is drawing that line between a genuine technical invention implemented in software and a bare computer program seeking a monopoly on abstract logic.

The core tension is between protecting genuine technical innovation and preventing broad monopolies over mathematical methods or business logic that would stifle the wider software industry. Getting the balance right has produced decades of litigation in the UK, Europe, and New Zealand.

2.  The Aerotel Four-Step Test  (UK, 2006)

In Aerotel Ltd v Telco Holdings; Re Macrossan's Application [2006] EWCA Civ 1371, the UK Court of Appeal laid down a four-step methodology for assessing software patent claims. The core idea was to look past the form of the claim and identify what the inventor had actually contributed to human knowledge.

1. Construe the claim — determine what the claim actually covers on its true interpretation, not what the drafter says it covers.
2. Identify the actual contribution — what has the inventor really added to human knowledge, as a matter of substance? The contribution must be real and specific.
3. Does it fall solely within excluded matter? — is the actual contribution nothing more than a computer program, mathematical method, or business method? If yes, the claim is unpatentable.
4. Is the contribution technical in nature? — a final consistency check ensuring the earlier steps have not led to an incorrect result.

"The contribution is to provide a computer program which can be used to carry out the method. The hardware used is standard and is not part of the contribution." — Court of Appeal, applying Aerotel to reject an automated company-formation system.

3.  How IPONZ Applied the Test Under Section 11  (2023)

New Zealand's Patents Act 2013 (s11) codified a similar framework. Where a claim involves a computer program, the Commissioner must identify the actual contribution and decide whether it lies solely in being a computer program. Section 11(4) directs attention to: the substance of the claim; the problem addressed; how it is solved; and the advantages gained. IPONZ applied this framework in two recent decisions.

Diane Mould — iDose System  [2023] NZIPOPAT 3

Invention: A computer-implemented method for determining patient-specific medication dosing regimens using Bayesian modelling and pharmacokinetic simulation.

Applicant's argument: The software outputs concrete recommendations that improve real-world patient outcomes.

IPONZ finding: The recommendations are informational only — a clinician may follow or disregard them. No real-world dosing effect is an inevitable consequence of running the program. Unlike a washing machine controller (which inevitably controls the machine), iDose merely advises. The actual contribution lies solely in the computer program.

Outcome: Application refused.

Yodlee Inc. — Financial Web Scraping  [2023] NZIPOPAT 4

Invention: A computer-implemented method using a trained machine-learning classifier to automatically generate scripts for extracting financial data from bank websites.

Applicant's argument: The system interacts with external websites and creates new products — site maps and scripts — beyond the computer itself.

IPONZ finding: Site maps are XML files produced by the program. Scripts are themselves computer programs. Downloading a webpage does not 'affect' that website. The skills needed are entirely within computer programming — no expertise outside the digital domain is required.

Outcome: Application refused.

A key principle from both decisions: the actual contribution must result in a real-world effect that is not simply the output of the program running, and must be an inevitable consequence of working the invention — not something contingent on a third party's discretion.

4.  Emotional Perception AI: The Test Has Changed  (UK, 2026)

In Emotional Perception AI Ltd v Comptroller General of Patents [2026] UKSC 3, the UK Supreme Court handed down a landmark ruling overturning the Aerotel framework entirely. The case concerned an artificial neural network (ANN) trained to recommend media files by replicating how humans subjectively perceive similarity between files, using only objectively measurable physical properties. The UKIPO had refused the application as a computer program 'as such'. The Supreme Court allowed the appeal and replaced Aerotel with the approach endorsed by the EPO Enlarged Board in G1/19 (Bentley Systems, 2021).

The three steps that were set out:

Stage 1 — Is there an invention? (the "any hardware" threshold) Does the claim involve any hardware at all — a computer, server, database, network, user device? If yes, the subject matter has technical character and qualifies as an invention under Article 52 EPC. This threshold is intentionally low. Almost all software and AI claims will clear it (if drafted properly).

Stage 2 — The intermediate step (filtering features) Before assessing novelty and inventive step, identify which features of the invention actually contribute to its technical character, viewed as a whole. Non-technical features that do not interact with the technical subject matter are filtered out and cannot support inventive step. Non-technical features that do interact with the technical subject matter stay in. This is a feature-by-feature analysis — quite different from Aerotel's holistic contribution approach.

Stage 3 — Novelty and inventive step Assess novelty and inventive step in the usual way, but only by reference to the features that survived the intermediate step filter.

Importantly, this three step test aligns UK patent law with European patent law under the European Patent Convention (which the UK is a part of).

Three Key Holdings

First: Aerotel is overruled. Steps 2–4 of the Aerotel methodology cannot be maintained. The contribution approach impermissibly conflates the threshold 'invention' test with novelty and inventive step, which are separate requirements under the EPC.

Second: ANNs are computer programs. An artificial neural network is an abstract model — a set of instructions to manipulate data — and therefore constitutes a "program for a computer". How those instructions were generated (human coding vs machine learning) is irrelevant.

Third: But these claims are not excluded. Because the claimed system involved hardware (a computer, database, communications network, and user device), it was not a computer program "as such". The claims qualified as an invention and should proceed to novelty and inventive step assessment.

Under the new approach, a claim passes the patentability threshold simply by involving any hardware. Since virtually every software or AI system runs on some form of computer hardware, this threshold is now very easy to clear. The serious analytical work shifts to an 'intermediate step' — identifying which features contribute to the invention's technical character — before assessing novelty and inventive step.

Old Test vs New Test — At a Glance

First question asked: Under Aerotel, the first question was what the inventor actually contributed to human knowledge. Under the new approach, the first question is simply whether the claim involves any hardware.

Role of novelty: Aerotel placed novelty at the centre of identifying the 'contribution'. Under the new approach, novelty is assessed separately, after the threshold test.

Software and AI claims: Under Aerotel, many such claims were excluded because the contribution was seen as 'just the program'. Under the new approach, claims are generally not excluded if hardware is involved.

Technical analysis: Aerotel required a holistic check of whether the whole contribution was technical. The new approach requires a feature-by-feature intermediate step to filter non-contributing features before the inventive step assessment.

Practical effect: Aerotel set a high threshold, resulting in many AI and software claims being refused at the eligibility stage. The new threshold is low — the focus moves to inventive step.

5.  What This Means for New Zealand

The Current Position: Uncertain but Significant

New Zealand is not bound by UK Supreme Court decisions. Section 11 of the NZ Patents Act 2013 is a domestic statutory provision, and its interpretation ultimately rests with IPONZ and the New Zealand courts. That said, the Supreme Court's decision is highly persuasive authority, and New Zealand patent law has historically drawn heavily on UK jurisprudence — including Aerotel itself, which formed the basis of both the 2023 IPONZ decisions, and other decisions over the last two decades.

Whether and how s11 will be reinterpreted in light of Emotional Perception AI remains to be seen. The question of whether the 'any hardware' threshold should replace the contribution analysis under s11(3) and (4) is genuinely open. Section 11's statutory language — which explicitly directs the Commissioner to identify the 'actual contribution' — may be harder to sidestep than the UK's non-statutory approach. On the other hand, IPONZ has never treated itself as entirely independent of the policy underpinnings shared with the UK and EPO, and the new framework may well prove persuasive.

In the meantime, applicants who received s11 refusals — including in cases analogous to Diane Mould or Yodlee — may wish to consider their options carefully. It would be unwise to assume that the New Zealand position will automatically mirror the UK change, but equally unwise to assume it will not shift at all.

Practical Takeaways for Applicants

• In the UK, the 'any hardware' threshold is now the test — the presence of a computer, server, database, or user device in the claim is sufficient to clear the Article 52 hurdle. Strategic focus should now be on novelty and inventive step.
• In New Zealand, the position under s11 is currently unsettled. Applicants facing s11 objections should preserve their appeal rights while monitoring how IPONZ and the courts respond to the new UK approach.
• For AI and machine learning claims, the Supreme Court confirmed that ANNs are computer programs — but because they always require hardware to run, they will not be excluded at the threshold stage under the new UK test.
• Claim drafting should ensure non-technical features are shown to interact with technical subject matter if they are to support inventive step under the new intermediate step analysis.

This guide is a summary for informational purposes only and does not constitute legal advice. Key cases: Aerotel/Macrossan [2006] EWCA Civ 1371; Diane Mould [2023] NZIPOPAT 3; Yodlee Inc. [2023] NZIPOPAT 4; G1/19 Bentley Systems [2021] EPOR 30; Emotional Perception AI Ltd v Comptroller General of Patents [2026] UKSC 3.

Mike Biagio

Mike Biagio has had previous experience in research and development and project engineering before becoming a patent attorney. Mike has been involved in specialist intellectual property law since 2001 in a variety of countries including South Africa, New Zealand and Australia, and has more than 21 years of experience as a patent attorney in obtaining patents and designs around the world for his clients.

Mike has been a lecturer and regular mentor at UNSW, University of Wollongong, and University of Sydney for a number of years, and has advised start-ups at incubators/accelerators on intellectual property and ideation.