Can you patent your AI? New guidance in Canada and the US for software and AI inventors

Corporate investment in AI continues to accelerate globally and patent filings for AI-related inventions are rising across every major jurisdiction. The threshold question, “Is this invention eligible for patent protection?” has become a central battleground for AI and software innovators. 

The Canadian Intellectual Property Office (CIPO) released a new Practice Notice on March 24, 2026, updating its framework for assessing patentable subject matter for computer-implemented inventions and including, for the first time, a worked example addressing machine learning. Meanwhile, the United States Patent and Trademark Office (USPTO) under its current Director has moved to lower barriers for AI patent applications through new examiner guidance and policy signals.  

This article examines recent developments to patenting software and AI inventions across the two regimes and offers practical guidance for AI and software inventors. 

CANADA: FURTHER CLARITY 

CIPO’s latest updates build on years of evolving case law. Subsection 27(8) of Canada’s Patent Act prohibits patents on mere scientific principles or abstract theorems. For computer-implemented inventions, eligibility has turned on a "physicality requirement": the claimed subject matter must have physical existence or manifest a discernible physical effect or change. 

How we got here 

The difficulties with subject matter eligibility in Canada are well documented. On the heels of the Supreme Court of Canada's hearing in Pharmascience v Janssen¹ regarding eligibility of methods of medical treatment, the Federal Court issued a subject matter eligibility decision in Dusome v Canada (Attorney General)². Dusome was the latest in a line of Federal Court decisions, including Amazon³, Choueifaty ⁴ and Benjamin Moore⁵, each of which overturned the Commissioner's subject matter eligibility determinations. 

In Dusome, the Commissioner had considered the "actual invention" to be the rules of a wagering poker game rather than the physical elements of the claims. The Federal Court held that this approach was wrong: the focus must be on the claims as purposively construed, and the Commissioner must then consider whether the subject matter "add[s] new knowledge to affect a desired result which has commercial value" as indicated by the Supreme Court of Canada in Shell Oil.⁶ In other words, the focus of the eligibility analysis must be the claims as purposively construed, and not the “actual invention”.

The NEW March 2026 Practice Notice 

CIPO’s new March 2026 Practice Notice, which takes into account recent court decisions, supersedes the previous CIPO practice notice issued in November 2020. Claims to software per se remain non-patentable; the preamble must refer to tangible media embodying instructions executable by a computer. Following Dusome, MOPOP section 17.03.09 on games no longer applies, since there is no per se prohibition on the patenting of games. 

Much of the analytical framework from 2020 is preserved. The 2026 notice uses purposive construction as the first step, grounds that construction in Supreme Court of Canada jurisprudence (Free World Trust ⁷, Whirlpool ⁸), distinguishes claim construction from subject matter assessment, and applies the same physicality requirement: physical existence or discernible physical effect or change, as the central gatekeeper. The 2026 notice reiterates the point from the 2020 notice that the problem-solution approach – under which examiners previously identified a problem and treated only the elements necessary to its solution as essential – should not be relied on.  

In purposive construction of a claim, the elements set out in a claim should be considered essential, unless the inventor establishes otherwise or if essentiality is contrary to the language used in the claim.  

The 2026 notice does away with the 2020 concept of identifying the "actual invention” as a subsequent step after claim construction, which featured heavily in the 2020 guidance as a distinct analytical construct, reflecting the Federal Court’s comments in Dusome.  

The Practice Notice also addresses medical diagnostic methods and medical uses. Notably, CIPO has acknowledged that it may need to adjust its guidance once the Supreme Court of Canada issues its anticipated decision on methods of medical treatment in Pharmascience (see our earlier article here). 

CIPO on claim construction 

Under the new guidance, CIPO first purposively construes the claims: examining the whole specification to ascertain the "nature of the invention," identifying common general knowledge, and presuming all claim elements essential unless otherwise established.  

For computer-implemented inventions, a recited computer is generally essential, but this alone does not satisfy physicality. Where additional physical elements exist (e.g., sensors, actuators, and measurement steps), physicality is generally met. Where they do not, the guidance asks what CIPO calls "the Schlumberger question"⁹: can the invention be distinguished from a mathematical calculation merely programmed into a computer? If the computer merely processes an abstract algorithm without improving the computer's functioning, the invention fails.  

If the algorithm improves the computer's functioning, the improved computer imparts physicality and the invention is patentable. AI algorithms are treated like any other algorithm under this test. Indicators of a qualifying computer improvement include faster processing, reduced power consumption, lower memory usage, enhanced security, reduced latency, or error reduction.  

CIPO's new worked examples 

The March 2026 Practice Notice includes new worked examples. The first two relating to seismic data analysis and computational efficiency reach similar outcomes as their 2020 analogs. The latter example confirms that reducing arithmetic operations can establish eligibility even in an abstract domain like investment portfolio optimization. 

CIPO's first machine learning example  

The third example is entirely new and involves a neural network trained on historical weather, irrigation, and crop yield data that outputs recommended irrigation parameters. The specification discloses the neural network's layered structure and training process "in extensive detail" and teaches improved crop yields and reduced water consumption.  

CIPO finds during purposive construction that the neural network structure and how to program it can be taken as common general knowledge (CGK). The critical finding: "It is not evident that any aspect of the neural network prediction algorithm … addresses any issue related to machine learning or computing. No issues in implementing the machine learning algorithm on a computer are disclosed." The Schlumberger question is asked, and the claim cannot be distinguished: "nothing other than the algorithm has been discovered. The input is data and the output is irrigation 'advice' and has no physicality in and of itself." Adding physical "irrigation means" in a dependent claim restores patentability. 

CIPO is signalling that model sophistication does not matter; a detailed layered neural network is treated the same as a simple formula if neither addresses a computing problem. Domain-level improvements (better crop yields, reduced water consumption) are not enough; they are abstract "advice" until acted upon by physical means. Had the specification disclosed that the architecture addresses some issue related to the computer implementation, the result could differ, just as the reduction in arithmetic operations saved the investment portfolio optimization example. For AI systems that interact with the physical world (robots, sensors, manufacturing equipment), including those physical elements in the claims provides physicality.  

The recent Patent Appeal Board decision in Stanley Victor Campbell (Re)¹⁰ reinforces the point. Claims to AI-based medical claims adjudication with neural computational logic were rejected; the Board found generic computer elements were used only to implement an abstract algorithm, akin to Schlumberger. 

USA: A MORE AGGRESSIVE SHIFT 

In the United States, eligibility under 35 U.S.C. § 101 is governed by the Alice/Mayo¹¹ two-step test, which asks whether a claim recites a judicial exception and, if so, whether it integrates that exception into a practical application or includes an inventive concept amounting to significantly more. Recent USPTO memoranda and decisions signal an intention to reduce difficulties for getting AI innovations through examination. 

The August 2025 guidance: key changes for AI applicants 

In August 2025, the USPTO issued a memorandum on assessing eligibility for AI and machine learning inventions. The memo distinguishes between claims that "set forth or describe" a judicial exception and claims where AI is merely involved without the claim describing the exception – the latter of which are eligible without further analysis. Examiners are reminded not to expand the mental process exception to limitations that cannot practically be performed in the human mind. Claims reflecting an improvement to computer or technology functioning may qualify as a practical application, and the improvement need not be explicitly stated in the specification if apparent to one of ordinary skill. The memo cautions against oversimplifying the "apply it" analysis; examiners should not dismiss claims merely because they involve a computer implementing an algorithm.  

Examiners should reject only if it is "more likely than not" (>50%) that the claim is ineligible. Examiners should not reject simply because they are uncertain. The burden rests on the USPTO to establish ineligibility by a preponderance of evidence. For AI inventions presenting close questions, the benefit of the doubt now goes to the applicant. 

Ex parte Desjardins (September 2025) 

An Appeals Review Panel (ARP) decision led by the USPTO Director in Ex parte Guillaume Desjardins et al. (Sept. 2025)¹² held that claims directed to a method of training a ML model on multiple tasks were eligible under § 101. The invention, filed by Google’s DeepMind Technologies, was directed to the problem of “catastrophic forgetting” in ML models, where models lose prior task knowledge when learning new ones. The ARP decision vacated the PTAB’s § 101 rejection, finding that the claims sufficiently reflected improvements in how the ML model operated. It also found the PTAB’s rejection “troubling”, noting that excluding AI innovations from patenting jeopardizes America’s leadership in AI technology.  

AI inventorship (November 2025) 

The USPTO's November 2026 memo on "Revised Inventorship Guidance for AI-Assisted Inventions" addresses who qualifies as the inventor when AI contributes to the inventive process and requires a natural person to make a significant contribution to conception while recognizing AI tools may play a substantial role. Notably, the guidance treats AI systems as analogous to any other tool that assists in the inventive process and that inventors “may use the services, ideas, and aid of others” without those sources becoming co-inventors. 

Using declarations to overcome § 101 Rejections (December 4, 2025) 

On December 4, 2025, the USPTO Director issued a memo encouraging the use of Rule 132 declarations, ‘Subject Matter Eligibility Declarations’, or SMEDs, to overcome § 101 rejections. SMEDs are optional but allow applicants to submit objective factual evidence supporting eligibility. They must focus on facts (e.g., performance metrics, benchmarks, design differences), not legal conclusions. Effective SMEDs address the state of the art at filing, technical improvement evidence, implementation practicality (e.g., that claimed steps cannot be performed mentally or involve specific data structures), and claim nexus (connecting evidence to specific claim language). Examiners must assess SMEDs under the preponderance-of-the-evidence standard and document their impact in the next Office Action. For AI inventors facing § 101 rejections, SMEDs are a practical tool to build an evidentiary record of real technical improvement. 

MPEP updates following Ex parte Desjardins (December 5, 2025) 

The following day, the USPTO issued updated MPEP guidance in light of Ex parte Desjardins, broadening the improvements analysis to cover "an improvement in technology or a technical field", not only computer functionality. New AI-specific examples of qualifying improvements include training a machine learning model in a way that protects knowledge of previous tasks while learning new ones and improving system performance through adjustments to ML model parameters. This signals that training methodology and model optimization may be recognised bases for eligibility. 

Canada vs. the US: where does each regime leave AI inventors? 

While both regimes seek to avoid granting patents on abstract ideas, they diverge in how that principle is applied in practice. In Canada, the Schlumberger question asks whether the algorithm improves the computer's functioning; if not, and the only physical element is a generic computer, the claim fails. In the United States, the August 2025 "close calls" standard resolves uncertainty in the applicant's favour. Critically, the USPTO improvements analysis considers improvements to technology generally, not only to the computer's own functioning which is a broader standard than Canada's Schlumberger question.

The United States currently offers a more permissive environment for AI patents. The SCC's pending Pharmascience decision and US legislative reform of § 101 may further reshape the landscape.

Takeaways for AI inventors 

Across jurisdictions, patent eligibility of software and AI continues to hinge on whether the invention is characterized as more than an abstract algorithm, but inventors should be aware that different jurisdictions apply this threshold in different ways. Inventors considering patent protection should make note of any technological improvements or physicality provided for in their AI inventions.

---

Stay tuned 

Stay up to date on IP law developments in Canada in 2026 through our IP Insights and Rx IP Updates. Subscribe to receive these in your inbox throughout the year. 

The preceding is intended as a timely update on Canadian intellectual property and technology law. The content is informational only and does not constitute legal or professional advice. To obtain such advice, please communicate with our offices directly.