Software Patents and AI Inventions in Canada
Software patents are available in Canada when purposively construed claims define more than an abstract idea. After Choueifaty (2020 FC 837) and Benjamin Moore (2023 FCA 168), CIPO examines computer-implemented inventions for a practical solution with physicality or a physical effect.

Key takeaways
- Section 27(8) of the Patent Act excludes mere scientific principles and abstract theorems.
- Choueifaty rejected the Patent Office's problem-and-solution approach to identifying essential elements.
- Benjamin Moore (2023 FCA 168) struck down a judge-made three-step test and left key questions open.
- Improvements to computer functioning are strong indicators of patentable subject matter.
- AI inventions need concrete applications, sufficient disclosure and documented human inventors.
Software patents in Canada: the short answer
Software patents are available in Canada, but only for claims that, properly construed, define more than an abstract idea, a mere scientific principle or a set of mental steps. Section 2 of the Patent Act defines an invention as a new and useful "art, process, machine, manufacture or composition of matter", and s. 27(8) excludes "mere scientific principle or abstract theorem".
The courts have rejected the Patent Office's older "problem and solution" approach and require the claims to be construed purposively; a computer-implemented invention is generally patentable where the computer, or a physical effect, forms part of an essential element that solves a practical problem. The Federal Court of Appeal's 2023 decision in Benjamin Moore confirmed that some core questions remain open.

For technology companies, fintech firms and AI developers, the practical result is that software patents can be obtained in Canada with careful drafting and prosecution, but examination outcomes depend heavily on how claims and descriptions present the technical contribution. This article explains the case law, the Canadian Intellectual Property Office (CIPO) practice that followed it, and how to draft and prosecute claims for fintech and AI inventions, including questions of inventorship.
The case law: from Amazon to Choueifaty and Benjamin Moore
Purposive construction
Canadian patent law requires claims to be construed purposively, identifying their essential elements from the perspective of the skilled person in light of the description and common general knowledge. In Canada (Attorney General) v. Amazon.com, Inc., 2011 FCA 328, the Federal Court of Appeal held that the Commissioner must apply purposive construction when assessing subject matter, and that the elements of the patentability analysis need not be considered in a particular order. The court also recognized that a business method is not excluded merely because it is a business method, although a practical application is required.
Choueifaty
In Choueifaty v. Canada (Attorney General), 2020 FC 837, the Federal Court held that the Commissioner erred by identifying essential elements based on the "problem and solution" described in the application, rather than through purposive construction of the claims as established by the Supreme Court of Canada. The case concerned a computer-implemented method of portfolio selection, and the court returned the application for redetermination.
Benjamin Moore
In Attorney General of Canada v. Benjamin Moore & Co., 2023 FCA 168 (26 July 2023), the Federal Court of Appeal considered refusals of two colour-selection applications. The Federal Court had set the refusals aside and proposed a binding three-step test for computer-implemented inventions. The Court of Appeal deleted that test, holding that it was not supported by Canadian case law and addressed issues not yet properly considered, including whether novelty and ingenuity can be weighed in the subject matter analysis.
The Court of Appeal agreed the Commissioner had erred by identifying essential elements "solely" from the problem and solution, and sent the applications back for expedited redetermination under the current Manual of Patent Office Practice (MOPOP).
CIPO practice on software patents
After Choueifaty, CIPO issued practice notice PN2020-04, Patentable subject-matter under the Patent Act, and updated the MOPOP. In summary, examiners:
- identify the essential elements of each claim through purposive construction;
- assess whether the "actual invention" is limited to an abstract idea, such as an algorithm, mathematical method, scheme or rules, or mental steps;
- ask whether the abstract idea, together with the computer, forms a single actual invention that solves a problem and exhibits physicality or a discernible physical effect or change; and
- treat improvements to computer functioning, or the use of a computer to control physical processes, as strong indicators of patentable subject matter.
The approach is more favourable to applicants than the pre-2020 practice, but it still produces objections for claims where the computer only performs well-known functions on data and the contribution lies in a business rule or financial model.
| Type of claim | Typical subject matter outcome (indicative) | Drafting focus |
|---|---|---|
| Improvement to computer operation (memory use, processing speed, security) | Generally favourable | Explain the technical problem and how the claimed steps change computer functioning |
| Software controlling hardware or physical processes | Generally favourable | Claim the physical outputs and interactions |
| Data processing producing information for a human | Fact-specific; frequent objections | Show integration of the computer with the solution, not mere automation |
| Pure business or financial methods on generic computers | Often refused | Look for technical implementation details that form part of the solution |
| Machine learning models and training methods | Fact-specific | Tie the model to a concrete application or technical improvement |
Fintech claims
Fintech applications face the most subject matter objections. Claims to trading strategies, credit scoring, insurance pricing or payment rules implemented on conventional computers are often characterized as abstract schemes. Applicants have better prospects where the claimed solution addresses a technical problem, such as reducing latency in order matching, securing transactions against specific attacks, reconciling distributed ledgers more efficiently or reducing data storage, and where the description explains how the computer elements interact to achieve that effect.
Draft the description so that a skilled reader can see the technical problem and the technical solution, not just the commercial benefit. Include system architecture, data flows and performance considerations, and claim the implementation at more than one level of generality. These details are hard to add after filing.
Drafting AI inventions
AI inventions raise the same subject matter questions as other software patents, plus disclosure and inventorship issues. Good practice for machine learning applications includes:
- Concrete application: claim the model in the context of a specific technical task, such as image analysis for defect detection, signal processing or control of equipment.
- Technical contribution: where the novelty lies in model architecture, training efficiency or hardware use, explain the technical effect, such as reduced memory or computation.
- Sufficiency: Canadian law requires the specification to enable the skilled person to practise the invention. For AI, that may mean describing training data characteristics, model structure and training steps in enough detail.
- Claim strategy: consider separate claims to training methods, inference methods, systems and computer-readable media, as each may be infringed by different actors.
Inventorship and ownership
The Patent Act and CIPO practice are built around human inventors, and an application must identify the inventor. Where AI tools assist research, the question is which individuals contributed to the conception of the claimed invention. Companies should document human contributions and ensure that employees and contractors assign their rights, since Canadian law has no general rule giving employers ownership of employee inventions without an agreement. See our insight on IP ownership between employees and contractors.
An illustration
Illustration: a payments start-up has developed a fraud-detection system. A first draft claims "a method of approving a transaction comprising receiving transaction data, computing a risk score using a trained model, and approving the transaction if the score is below a threshold". An examiner is likely to treat the actual invention as a scoring rule run on a generic computer.
A stronger application would describe and claim, for example, how the system processes streaming data at the point of sale to reduce authorization latency, how the model is deployed across distributed nodes with a particular synchronization method, or how the system reduces false positives by using device-level signals collected in a specific way. Each of those features ties the computer to the solution of a practical problem, which is what the Canadian approach looks for.
Software patents, trade secrets and copyright
Software patents are not the only way to protect software. Source code is protected by copyright automatically, and confidential algorithms, model weights and training data can be protected as trade secrets if kept confidential. Patents have the advantage of protecting functionality against independent development, but they require public disclosure and cost more to obtain and enforce. Many companies combine the tools: patents for features that competitors can observe or reverse engineer, trade secrets for back-end methods that remain hidden, and contracts and security measures to protect both. See our page on copyright and trade secrets.
Prosecution strategy for software patents
Most subject matter disputes are resolved in prosecution, not litigation. Useful tactics include responding to objections with arguments grounded in purposive construction, amending claims to bring technical features into the essential elements, conducting examiner interviews, and appealing to the Patent Appeal Board where necessary. Applicants can request accelerated examination in some circumstances, including through Patent Prosecution Highway arrangements based on allowed claims abroad, but should consider the effect on patent term adjustment; see our insight on patent term adjustment.
Because the law is still evolving, applicants should keep continuation options open through divisional applications where available, and monitor further court decisions on whether novelty can be considered in subject matter analysis, because a change on that point could affect both pending applications and the validity of granted patents.
Software patents for foreign applicants
Many software patents in Canada are filed by foreign companies entering through the PCT or the Paris Convention. Claims allowed in the United States or Europe are a useful starting point, but they do not transfer automatically, because each office applies its own subject matter test. European practice focuses on technical character and technical effect; the United States applies its own eligibility framework; Canada asks whether the actual invention, identified through purposive construction, has physicality or a discernible physical effect and solves a practical problem.
In practice, claims drafted to satisfy European technical-effect standards often fare reasonably well in Canada, but applicants should review Canadian claims rather than simply copying them. Where a Patent Prosecution Highway request is available, the Canadian claims must correspond sufficiently to the claims allowed abroad, which can limit the scope for Canadian-specific amendments. Applicants should therefore decide early whether speed or claim flexibility matters more for each Canadian filing.
Enforcement and value of Canadian software patents
Software patents in Canada are enforced in the Federal Court or provincial superior courts. Validity attacks frequently combine subject matter, obviousness and claim breadth arguments, so patents drafted with clear technical contributions are more robust in litigation. For investors and acquirers, a portfolio of Canadian software patents is most valuable when it covers core technical features rather than business rules, when inventor assignments are complete, and when the claims read on competitors' implementations. See our pages on patents, technology and AI and fintech, payments and digital assets.
How TRW & Co helps
We advise technology, fintech and AI companies on software patents in Canada. We help develop filing strategies, review claims and descriptions for subject matter risk, coordinate responses to examiner objections with the client's registered patent agents, assess freedom to operate, and support enforcement and transactions. We also advise on inventorship, assignments and ownership of AI-assisted inventions. To discuss a portfolio or a specific application, contact us. This article is general information, not legal advice.
Frequently asked questions
Are software patents allowed in Canada?
Yes, but not for abstract ideas, mere scientific principles or mental steps. A computer-implemented invention is generally patentable where, properly construed, the computer or a physical effect forms part of a solution to a practical problem, such as improving computer functioning or controlling a physical process.
What did Choueifaty decide?
In Choueifaty v. Canada (Attorney General), 2020 FC 837, the Federal Court held that the Commissioner of Patents erred by identifying essential elements from the problem and solution rather than through purposive construction. CIPO then issued practice notice PN2020-04 and revised its approach to computer-implemented inventions.
What is the significance of Benjamin Moore?
In 2023 FCA 168, the Federal Court of Appeal deleted a three-step test that the Federal Court had imposed for computer-implemented inventions, saying it was not supported by case law. It left open questions such as whether novelty can be considered in the subject matter analysis and remitted the applications for redetermination.
Can fintech methods be patented in Canada?
Sometimes. Business and financial methods implemented on generic computers often attract subject matter objections. Applications fare better where the claims address a technical problem, such as transaction security, latency or data efficiency, and the description explains how computer elements achieve the solution.
Can an AI system be named as an inventor in Canada?
The Patent Act and CIPO practice are built around human inventors, and applications must identify the inventor. Companies using AI in research should identify and document the individuals who contributed to conceiving the claimed invention and secure assignments from them.
Sources
General information only, not legal advice. References to statutes, courts and authorities describe published law and practice at the date shown.


