Our Blog
Proposed Changes to MOPOP: What They Mean for AI and Software Patents in Canada
June 18, 2025
The Canadian Intellectual Property Office (CIPO) has released significant proposed changes to its Manual of Patent Office Practice (MOPOP) regarding artificial intelligence inventions, marking a crucial development for patent practitioners and their clients in the AI space. These updates are especially important for those seeking AI and software patents in Canada.
Section 22.03.03c: Patentable Subject Matter for AI Inventions
The changes introduce a new section 22.03.03c specifically addressing AI-related inventions. The section outlines multiple examples establishing that CIPO will treat AI inventions, particularly machine-learning algorithms, like any other computerized algorithms. CIPO emphasizes that the “actual invention” must be determined based on claim construction, focusing on elements that contribute to a physical or technical effect.
The updated guidance draws a sharp line between patentable and non-patentable AI inventions based on the physicality requirements. Under the proposed framework, AI systems that merely output recommendations or advice without physical implementation are considered abstract and, therefore, not patentable.
Patentability Examples: Abstract vs. Physical AI Systems
Examining Example 1, a neural network system that only processes weather data and outputs irrigation recommendations is deemed non-patentable because the “actual invention” is an abstract algorithm that produces information without a tangible effect. However, when the same system includes a component for physically irrigating crops according to the recommendations, it becomes patentable due to the added physicality in the system.
Example 2 describes an AI system utilizing a specialized back-propagation algorithm that reduces the number of neural network layers and computational requirements, making it patentable as it enhances the physical operation of the computer system itself. This distinction emphasizes that improvements to computer functioning can render AI inventions patentable.
Best Practices for Drafting AI and Software Patent Claims in Canada
The examples in section 22.03.03c make it clear that to enhance the likelihood of patentability in Canada, claims should do more than present abstract outputs. For example, the claims should include concrete physical steps that demonstrate real-world implementation or effect.
Applications claiming abstract AI outputs without physical implementation may face subject-matter objections. Additionally, applicants should emphasize how their AI system delivers tangible computational improvements or increased processing efficiency by providing comprehensive, detailed descriptions of the technical structures, neural network architectures, and training methodologies employed.
What This Means for Innovators in Canada
If you are developing AI, machine learning, or software solutions in Canada, now is the time to reassess your IP strategy. These MOPOP revisions highlight the importance of demonstrating either physical implementation or concrete computational improvements in your applications.
Need help navigating the revised CIPO guidelines? Please contact us today to get expert support tailored to your innovation.