Artificial intelligence is transforming industries across Canada, and in 2026 the federal government responded with one of its most ambitious technology strategies to date. Canada’s national AI strategy sets out a sweeping vision for economic growth, job creation, and sovereign infrastructure, yet it says remarkably little about IP laws. This article surveys the new AI strategy’s key commitments and examines recent developments in patent and copyright law that apply to AI.

Canada’s National Artificial Intelligence Strategy: AI for All

In June 2026, the Minister of Innovation, Science and Industry published Canada’s National Artificial Intelligence Strategy: AI for All (the “Report”).

According to the Report, Canada is home to more than 3,500 firms actively developing advanced AI models, tools, and applications. Collectively, these firms have raised over CAD37 billion in venture capital funding. The Report also states that by 2030, generative AI alone could contribute as much as CAD187 billion a year to Canada’s economy, according to some estimates.

Among other things, Canada’s AI strategy aims to protect citizens from AI risks while expanding AI literacy training and access to trusted AI agents. The AI strategy aims to create up to 250,000 new jobs through AI adoption by 2031, increase business AI adoption from 12% in 2026 to 60% by 2034, and unlock a 3% increase in GDP through the commercialisation and application of AI in key sectors.

While the AI strategy applies across Canada’s AI ecosystem, the Report identified five priority sectors for strategic investment:

• health and life sciences;
• energy and natural resources;
• transportation;
• agriculture; and
• manufacturing and robotics.

The AI Strategy Report also mentions investments in sovereign AI infrastructure, including building a world-leading supercomputer by 2031, and a multilateral alliance to strengthen Canada’s technological resilience and autonomy.

Some of the proposed legislative and regulatory actions noted in the Report include:

• modernising consumer privacy legislation;
• introducing online safety laws to protect Canadians in the digital age;
• creating a “Canada Trusted AI Certification” program to help identify trustworthy AI products in the marketplace; and
• renewing funding for the Standards Council of Canada’s AI Program to enable standards-based AI testing, certification, interoperability, and global market access.

The Report includes commitments to AI investments including the following:

• launching an “AI Missions Program” to advance targeted, high-impact projects, starting with a CAD200 million commitment to improving health outcomes for Canadians;
• establishing a CAD500 million Canadian Tech Growth Fund to help Canada’s most promising AI companies;
• establishing a CAD50 million Creative Technology Program to support Canadian creators in using AI;
• investing  CAD130 million for commercialisation programs across the National AI Institutes to cultivate a new generation of AI entrepreneurs;
• providing Canadian SMEs with an additional CAD700 million in affordable sovereign compute through an expansion of the Compute Access Fund; and
• leveraging CAD159 million invested in Budget 2025 through the Elevate IP and IP Assist programs to protect Canadian intellectual property and support SMEs to commercialise their intangible assets in the global marketplace.

Notably, though the Report emphasises the importance of sufficient investment capital and a competitive investment landscape to retain critical IP and talent, it provides no guidance or direction on the future of IP protection and enforcement for AI inventions and AI-generated or AI-assisted works. The Report does not propose changes or updates to existing IP law, or the introduction of new AI-specific legislation, to address IP issues.

The practical consequence is that AI and IP in Canada remain governed by statutes enacted decades before the technology existed, including:

• the Patent Act, RSC 1985, c P-4;
• the Copyright Act, RSC 1985, c C-42; and
• the Trademarks Act, RSC 1985, c T-13.

Canadian Patent Law and AI

The AI Strategy is expected to increase the volume of Canadian-owned inventions and patents through new research and commercialisation funding and support for SMEs.

Two patent issues that are particularly relevant to developers and inventors of AI products and systems are inventorship and patentability. Recent developments in respect of these issues are discussed below.

Inventorship

The Canadian Patent Appeal Board (“PAB”), in Thaler, Stephen L. (Re), 2025 CACP 8, considered whether a non-human entity can be recognised as an inventor. Dr Stephen Thaler filed a Canadian patent application designating DABUS (Device for the Autonomous Bootstrapping of Unified Sentience), an AI system, as the sole inventor of a food and beverage container and signalling device.

The PAB recommended refusal of the patent application, concluding that the term “inventor” in the Patent Act refers to natural persons and does not apply to AI systems. This result is consistent with how courts and patent offices in the United States, United Kingdom, Australia and elsewhere have handled parallel DABUS patent filings.

Dr Thaler filed a Notice of Appeal with the Federal Court of Canada challenging the PAB decision. As of this writing, a hearing on the appeal has not yet taken place, and no court decision has been issued.

Patentable subject matter

In March 2026, CIPO released a new practice notice on patentable subject matter under the Patent Act (“2026 Practice Notice”) to account for three recent court decisions:

• Canada (Attorney General) v Benjamin Moore & Co. (2023 FCA 168);
• Dusome v Canada (Attorney General) (2025 FC 1809); and
• Pharmascience Inc. v Janssen Inc. (2024 FCA 23).

In Canada, the subject matter of a patent claim is determined through purposive construction, which examines the entire patent specification to ascertain the nature of the invention. After a claim has been construed, the next question is whether the subject matter is a patentable invention.

Section 2 of the Patent Act defines an invention as “any new and useful art, process, machine, manufacture or composition of matter, or any new and useful improvement in any art, process, machine, manufacture or composition of matter.” Section 27(8) excludes “any mere scientific principle or abstract theorem.” The definition of “invention” includes a physicality requirement.

The 2026 Practice Notice includes guidance on applying purposive construction and assessing patentable subject matter for computer-implemented inventions, as outlined below.

Purposive construction

• If a claim recites a computer, the computer is an essential element in most cases. However, merely identifying a computer as an essential element does not automatically satisfy the physicality requirement.
• The physicality requirement is generally met in computer-implemented inventions with additional physical elements, such as sensors, or output means, such as a robotic actuator.
• It must be considered whether the algorithm executed by the computer results in any improvement to the operation of the computer – for example, improvement to the computer’s operation for tasks in general (eg, reduced memory usage) or to the task at hand (eg, manipulations that speed up a mathematical simulation).

Assessment of patentable subject matter

If the computer-implemented invention has no physicality outside of the computer system elements, one must consider the following (the “Schlumberger question”):.

• If the computer runs an abstract algorithm in a conventional way without improving the functioning of the computer, then only the abstract algorithm was discovered, not the physical computer. Therefore, the physicality requirement is not met, making the subject matter unpatentable. “Something more” (eg, improvement to computer operation) is required.
• If the algorithm actually improves the computer’s functioning, the invention involves more than a “bare practical application”. The improved computer and its improved functionality satisfy the physicality requirement, making the subject matter patentable.

Applying the 2026 Practice Notice

The 2026 Practice Notice also provides guidance on applying purposive construction and assessing patentable subject matter for medical diagnostic methods, which often include an element correlating a specific analyte or test result to a disease. For the subject matter to be patentable, there also has to be a physical element (eg, physical steps for detecting the presence or quantity of an analyte in a sample).

Further, the 2026 Practice Notice noted that methods of medical treatment are not patentable subject matter under Canadian law. This was reaffirmed by a recent Supreme Court of Canada decision in Pharmascience Inc. v Janssen Inc., 2026 SCC 26. The Court confirmed that when determining whether a subject matter is unpatentable as a method of medical treatment, the ultimate question is whether the claimed subject matter amounts to professional medical skill and judgment (ie, whether it seeks to fence in an area of medical treatment).

For AI developers building diagnostic or treatment-recommendation tools, this decision suggests that innovations can remain patentable as long as they are framed as discrete technical outputs or products rather than as systems that direct or replace a clinician’s exercise of professional judgment at the point of care. This distinction will likely be a key drafting consideration for AI patent claims in the health and life sciences sector, since claims that read as substituting for physician skill and judgment risk being struck down as an unpatentable method of medical treatment.

Taken together, AI inventors should carefully consider whether their invention’s subject matter is patentable. Some development considerations could include the following:

• naming a human inventor;
• incorporating hardware elements, if possible;
• determining real-world applications;
• if the AI invention has no hardware or other physical element, ascertain whether the algorithm improves the computer’s functioning; and
• how to measure and record such improvements.

Canadian Copyright Law and AI

The Copyright Act does not expressly address whether works created with or by AI systems qualify for copyright protection, and Canada’s new AI strategy offers no legislative proposals on the subject. Canadian courts are now confronting a growing wave of copyright infringement claims against major AI developers, raising foundational questions about the legality of using copyrighted works to train AI systems, the jurisdictional reach of Canadian copyright law, and the scope of available defences.

Authorship

Under the Copyright Act, copyright subsists in every original literary, dramatic, musical, and artistic work where the author meets certain conditions, including citizenship or residency in a treaty country. However, the Act does not define “author.” Canadian courts have not yet ruled definitively on whether a non-human entity, such as an AI system, can qualify as an author for copyright purposes.

A pending Federal Court application (T-1717-24) may provide early guidance on authorship for copyright protection in an AI context. In that application, Samuelson-Glushko Canadian Internet Policy and Public Interest Clinic (“CIPPIC”) challenges the copyright registration of “Suryast,” an image created by an AI system. The user allegedly uploaded a photograph of a sunset he took and a copy of “The Starry Night” by Vincent Van Gogh to an AI painting app, then prompted the app to apply Van Gogh’s style to the base image. CIPPIC argues that AI cannot be an author and that the image lacks originality to attract copyright protection. The outcome could set an important precedent for AI-related authorship claims in Canada.

Infringement actions and class actions

Canadian courts have seen a surge of copyright infringement claims targeting major AI developers. Examples of such cases are described below. As of this writing, no court decisions have been released on the copyright issues raised in any of these cases.

In Toronto Star Newspapers Limited v OpenAI Inc. et al. (Ontario, CV-24-00732231-00CL), seven of Canada’s leading news media companies alleged that OpenAI scraped their copyrighted journalistic works to train its GPT models without consent, circumvented technological protection measures such as paywalls and robots.txt protocols, and breached website terms of use. OpenAI challenged the Ontario Superior Court’s (“ONSC”) jurisdiction, arguing that the alleged wrongdoing underlying the plaintiff’s claims took place outside of Canada. In the alternative, they argued that the United States is the proper and convenient forum.

In the interlocutory decision, Toronto Star Newspapers Limited v OpenAI Inc. 2025 ONSC 6217, the ONSC confirmed that it has jurisdiction over the OpenAI defendants. The Court considered the following presumptive factors that allow a court to assume jurisdiction:

• the defendant is domiciled or resident in the province:
• the defendant carries on business in the province;
• the tort was committed in the province;
• a contract connected with the dispute was made in the province; and
• property located in the province relating to the asserted claims.

The ONSC found that the OpenAI defendants carried on business in Ontario. At least some of the user-generated training data comes from users in Ontario, and it advertises to customers, including Ontario users. OpenAI made the results of web crawling, scraping, and model training, including unauthorised reproduction of the plaintiffs’ content, available to Ontario users. The Ontario-based plaintiffs’ website terms of use created a contract between those who accessed the websites (including the OpenAI defendants) and the relevant plaintiff. Some OpenAI defendants were unjustly enriched by profits from wrongful acts committed in Ontario. Taken together, the presumptive factors were established and were not rebutted by OpenAI.

In Doan v Clearview AI Inc. (Federal, T-713-20), Ms Doan commenced a proposed class action proceeding in the Federal Court, alleging that Clearview infringed copyright and moral rights by collecting, copying and storing images of human faces taken in Canada without the consent of the individuals who took the images or who hold copyright and/or moral rights in those images. The Federal Court dismissed Ms Doan’s motion to certify in 2023 FC 1612 for failing to establish factually that there is an identifiable class of two or more persons as required for certification. The Federal Court of Appeal in 2025 FCA 133 set aside the Federal Court order and returned the appellant’s motion to the lower court.

Other notable pending cases

• Clare v Meta Platforms Inc. and Facebook Canada Ltd. (Federal, T-1204-25): Author Kerry Clare filed a proposed class action at the Federal Court on behalf of Canadian copyright holders (excluding Quebec residents) whose copyrighted works were allegedly downloaded and/or torrented by Meta for use in training its LLaMA language model. Meta challenged the Federal Court’s jurisdiction by filing a motion to strike for want of jurisdiction. The jurisdictional motion hearing took place in June 2026.
• MacKinnon v Meta Platforms Inc. and Facebook Canada Ltd. (BC, S-252936), Anthropic PBC (BC, S-253893), NVIDIA Corporation (BC, S-255561) and Databricks Inc. and MosaicML Inc. (BC, S-255562): Canadian author James Bernard MacKinnon filed proposed class actions against Meta, Anthropic, NVIDIA, and Databricks/MosaicML in the Supreme Court of British Columbia alleging that each defendant used pirated datasets comprising copyrighted books to train their large language models.
• Gagné v Stability AI Ltd., Midjourney, Inc., Google LLC, and Runway AI, Inc. (Federal, T-2676-25): Professional visual artist Mark Gagné filed a proposed class action in the Federal Court of Canada alleging the defendants infringed the copyrights of class members by obtaining datasets containing URLs to hundreds of millions of copyrighted images to train their image generators (input infringements) and by allowing their image generators to reproduce the copyrighted images and the distinctive styles of artists (output infringements). Stability AI and Midjourney challenged the Federal Court’s jurisdiction by filing motions to strike for want of jurisdiction.
• Thomas v Apple Inc. (Federal, T-3831-25): Canadian author Joan Thomas filed a proposed class action in the Federal Court alleging that Apple reproduced copyrighted works without authorisation (through downloading from shadow libraries and data scraping practices) and used the works to train its OpenELM Language Models and Apple Intelligence Foundation Language Models.
• Jackson v OpenAI, Inc. et al. (BC, S-246286): Artist and illustrator Michael Dean Jackson filed a proposed class action against OpenAI and Microsoft in the Supreme Court of British Columbia alleging that the defendants reproduced the plaintiff’s and class members’ copyrighted works in two key ways: through the training process of their AI models and through the models’ outputs. The plaintiff alleged that the new works created by the models represent a competitive product to works of the plaintiff and class members. The defendants brought a motion to strike on the ground that the court does not have jurisdiction.
• VerticalScope Inc. et al. v OpenAI et al. (Ontario, CL-26-00000243-0000): VerticalScope operates online forums, including RedFlagDeals.com and AutoGuide.com, and professional content properties that contain editorial, review, guide and vertical content. VerticalScope’s websites contain billions of articles, posts, individual comments, product reviews, and topic summaries, shared by users across threads and forums. VerticalScope alleged that OpenAI used copyrighted works it owns to develop its software programs and train its AI systems.

Several cross-cutting themes emerge from this wave of AI copyright litigation. These proceedings raise multiple issues in the AI context, including:

• copyright infringement (use of copyrighted material in AI training);
1. moral rights infringement (modification and distortion of copyrighted works and authors’ right to integrity and association);
• whether exceptions to infringement apply (eg, fair dealing)
• circumvention of technological protection measures (ie, measures that prevent unauthorised users from viewing, opening or copying the content);
• removal of rights management information (ie, metadata, labels and codes that identify the author, copyright owner and other information);
• breach of contract (ie, violation of plaintiff’s website terms of use);
• unjust enrichment; and
• class certification, in the case of class actions.

Jurisdictional reach is a threshold issue. Courts apply the “real and substantial connection” test established by the Supreme Court of Canada in Club Resorts Ltd. v Van Breda, 2012 SCC 17 to determine whether a court can assume jurisdiction over a certain claim. AI businesses can be found to conduct business in a Canadian jurisdiction when their business activities in the context of their business as a whole are sufficiently connected to that jurisdiction. As shown above, much AI-related litigation may involve more than one defendant, and in personam jurisdiction must generally be established for each defendant. If a Canadian court is found to have jurisdiction, it should exercise that jurisdiction except where another forum is clearly more appropriate, suitable and convenient.

Fair dealing and other statutory defences remain untested in the AI training context. No Canadian court has ruled on whether reproducing copyrighted works for AI model training qualifies as fair dealing under the Copyright Act. Resolving this issue will have significant implications for the legality of AI training practices in Canada.

Conclusion

Canada’s new AI strategy represents a significant commitment to building the country’s AI ecosystem through funding, infrastructure, and sector-specific investment. But the strategy’s lack of guidance on intellectual property matters means that the most consequential questions about AI and IP are being resolved not by legislators, but by courts and patent offices applying statutes that predate generative AI by decades.

In the patent space, CIPO’s 2026 Practice Notice and the pending DABUS appeal are key developments shaping the rules for the eligibility of subject matter and inventors. In the copyright space, a wave of infringement litigation is forcing Canadian courts to confront foundational questions about AI training, jurisdictional reach, and the protectability of AI-generated works.

The outcomes of these proceedings will have lasting implications for AI developers, rights holders, and policymakers. Stakeholders across the AI and creative industries should monitor these developments closely and assess the legal risks and opportunities presented by this rapidly shifting landscape.