Two Steps Forward: CIPO Resets on Non-Patentable Subject Matter

On March 24, 2026, the Canadian Intellectual Property Office (CIPO) released a new practice notice on patentable subject matter under the Patent Act (the March 2026 Practice Notice), together with updated examples illustrating its application. The March 2026 Practice Notice replaces the 2020 Practice Notice, which we have previously commented on at length.

As a top line, the March 2026 Practice Notice is a positive development. It follows three significant decisions:

• The Federal Court of Appeal’s decision in Canada (Attorney General) v Benjamin Moore & Co (which we commented on here)
• The Federal Court’s decision in Dusome v Canada (Attorney General) (which we commented on here)
• The Federal Court of Appeal’s decision in Pharmascience Inc v Janssen Inc, the appeal of which is currently under reserve at the Supreme Court of Canada.

Against this backdrop, CIPO has updated its guidance in ways that are largely encouraging. But the devil, as always, is in the detail.

The Good: “Actual Invention” Is Out

The most significant improvement in the March 2026 Practice Notice is also the most straightforward: the term “actual invention” is gone. As we cautioned after Dusome, the 2020 Practice Notice’s reliance on identifying an “actual invention” risked creating a free-standing inquiry, separate from and subsequent to purposive construction, that could morph into its own test. This is precisely the kind of approach the Supreme Court of Canada rejected in Free World Trust v Électro Santé Inc and Whirlpool Corp v Camco Inc. In Dusome, the Federal Court agreed, finding that the Commissioner had erred by focusing on what the inventor had “actually invented” rather than properly construing the claims.

The March 2026 Practice Notice now consistently refers to the “construed claim” rather than the “actual invention.” It reiterates that subject matter is determined through purposive construction, and that discerning “the nature of the invention” is part of that exercise, not a separate step. It adds that “what the inventor has actually invented or what the inventor claims to have invented is relevant and necessary in a number of contexts,” citing the Federal Court of Appeal’s decision in Canada (Attorney General) v Amazon Inc. This is a clear and welcome improvement.

The Defensible (With Limits): The “Schlumberger Question”

For computer-implemented inventions where there is no physicality outside of the computer system itself, the March 2026 Practice Notice introduces what it calls “the Schlumberger question.” The question asks whether the situation under consideration can be distinguished from Schlumberger Canada Ltd v Commissioner of Patents, a 1981 Federal Court of Appeal decision in which a claimed invention that amounted to programming a mathematical formula into a computer was held not to be patentable subject matter.

The Schlumberger question is arguably defensible as a threshold check, but only if its application remains narrow and circumscribed to Schlumberger’s facts. In Schlumberger, the invention at issue was, in the Federal Court of Appeal’s characterization, limited to “the discovery of the various calculations to be made and of the mathematical formulae to be used in making those calculations.” The claims recited a “machine operated method of processing well logging data” with a data input step, processing steps, and an output step. The court found that, in substance, nothing had been discovered beyond the mathematical method itself.

Used as a narrow question (does this claim, like Schlumberger, amount to nothing more than programming a known formula into a known computer?), the inquiry is grounded in the case law. Expanded beyond that, it risks becoming the very kind of unbounded, post-construction analysis the courts have repeatedly told CIPO to avoid.

The March 2026 Practice Notice signals that where the processing of the algorithm improves the functioning of the computer, whether in general performance or in a specific task, what has been discovered includes the use of a computer as more than a bare practical application, and the invention pertains to the art of computing. That framing is reasonable. The concern lies in how it is applied.

The Concerning: Old Habits in New Examples

If the body of the March 2026 Practice Notice represents progress, the accompanying examples raise serious concerns about how examiners will apply that progress. Two issues immediately stand out.

1) The Schlumberger Question as “Actual Invention” by Another Name

The examples suggest that, notwithstanding the Practice Notice’s emphasis on purposive construction, the Schlumberger question may function as the old “actual invention” analysis under a new label. Consider the analysis of claim 3 in Computer-implemented Example 1 (a method of analyzing seismic data using a new algorithm). After purposive construction identifies all elements as essential, the assessment states:

The exercise at work in that analysis, asking what has “really been discovered” and stripping away elements that do not independently provide physicality, is strikingly similar to the “actual invention” inquiry the Practice Notice itself purports to abandon. The conclusion that “nothing other than an algorithm has been discovered” is effectively an inventive-concept determination in different clothes. The label has changed. The analysis, as applied in the examples, has not.

The same pattern emerges in:

• Computer-implemented Example 2. The analysis dismisses claim 1 (applying an investment portfolio algorithm on a computer) because “nothing other than the abstract algorithm has been discovered.”
• Computer-implemented Example 3. The analysis dismisses claim 1 (a neural-network-based crop irrigation scheduling system) because “nothing other than the algorithm has been discovered,” even though the specification disclosed in “extensive detail the particulars of the layered neural network structure” and demonstrated improved crop yields and reduced water consumption.

If the Schlumberger question is to be a legitimate, limited inquiry, it should ask a narrow question: does this claim, as purposively construed, amount to nothing more than Schlumberger itself? It should not be a vehicle for collapsing the subject-matter assessment back into a freestanding determination of what the examiner considers the “real” invention to be. As applied in the examples, the Schlumberger question appears to be doing exactly that.

2) Presuming Common General Knowledge from Silence

The second significant concern with the examples relates to how common general knowledge (CGK). In the purposive construction sections, the examples repeatedly presume that something is CGK simply because the description provides limited detail about it. From Computer-implemented Example 1:

The same approach recurs across the examples:

• Computer-implemented Example 2. “The computer and how to program the simulation and algorithm A and transform B can be taken as CGK, given the lack of detail in the description of the programming aspects.”
• Computer-implemented Example 3. “The processor and how to program the disclosed neural network structure can be taken as CGK.”

This is problematic. The level of detail in a patent specification does not determine whether something constitutes CGK. A patent applicant may choose not to describe well-known steps in detail for perfectly legitimate reasons; patent specifications are not textbooks. But the reverse does not follow: the absence of detail does not prove that a given element is CGK. CGK is a question about what the skilled person would know, not an inference from drafting style. Using description length as a proxy for CGK creates a dangerous incentive: patent applicants who provide unnecessarily lengthy explanations for every possible aspect of their invention may be rewarded, while those who focus their specification (whether for cost or other legitimate reasons), risk having essential elements dismissed as known. The jurisprudence does not support that approach.

The Neural Network Problem

Computer-implemented Example 3 warrants scrutiny. The example describes a sophisticated machine-learning system that uses a neural network to generate crop irrigation schedules. The specification discloses “in extensive detail the particulars of the layered neural network structure of prediction algorithm V and how the neural network is trained using database X.” The system achieves “improved crop yields and reduced water consumption.” And yet, the analysis concludes that claim 1 (the system without physical irrigation equipment) is not patentable because “nothing other than the algorithm has been discovered.”

This conclusion treats a complex, multi-layered neural network trained on specific datasets as interchangeable with the simple mathematical formula at issue in Schlumberger, a case decided in 1981, well before machine learning existed in any practical form. The Practice Notice states that improvements to the computer’s functioning can provide the necessary physicality. Arguably, a neural network that trains a model to solve a specific real-world problem does more than process an “abstract algorithm in a well-known manner.” The example’s dismissal of this as equivalent to Schlumberger suggests a narrow and technical reading of what “improves the functioning of the computer” means, and one that may not survive judicial scrutiny.

Key Takeaways

1) Dropping “actual invention” is genuine progress. The March 2026 Practice Notice recentres the analysis on purposive construction and treats “the nature of the invention” as a component of that construction rather than a separate step. This is a meaningful improvement over the 2020 Practice Notice, responds to the Federal Court’s concerns in Dusome, and aligns with the Supreme Court of Canada’s framework in Free World Trust and Whirlpool.

2) The Schlumberger question is defensible — if it stays narrow. As a limited inquiry comparing the claim under consideration to the specific facts of Schlumberger, the question is grounded in the case law. But it must not become a vehicle for asking what the examiner thinks the inventor “really” invented. The examples, unfortunately, suggest it may already be headed in that direction.

3) The examples are the real test — and they raise real concerns. The examples accompanying the March 2026 Practice Notice will guide examiners on a day-to-day basis. The approach to the Schlumberger question in those examples is both problematic and inconsistent with the progress the Practice Notice itself represents. The approach looks remarkably similar to the old “actual invention” analysis, and the CGK is treated as something that can be inferred from specification length, neither of which is consistent with judicial guidance.

4) Important implications for computer-implemented inventions. The March 2026 Practice Notice opens the door for inventions where the only physical element is a computer to still qualify as patentable subject matter, provided the claimed invention improves the computer’s functionality. This is welcome, but practitioners should be attentive to the narrow way in which “improves the computer’s functionality” appears to be interpreted in the examples.

5) Courts are still shaping the landscape. We are still awaiting Supreme Court guidance in Pharmascience. In the meantime, CIPO cites the Federal Court of Appeal’s decision in its discussion of medical uses, signalling that it is closely monitoring ongoing court decisions. Given the tensions between the Practice Notice’s principles and the examples’ application, further judicial guidance will be essential.