Artificial Intelligence (AI) is reshaping industries across Canada, driving efficiency in innovation and enforcement. However, its rapid adoption has significant legal implications for intellectual property (IP) and data privacy. This article explores three key areas: how AI tools are used to enhance the detection of patent infringement, the IP implications of AI’s inputs and outputs, and the privacy risks associated with generative AI.
Beyond manual reviews: Using AI-powered tools for patent analysis and enforcement
AI-powered tools use techniques such as natural language processing (NLP), machine learning (ML) algorithms, and computer vision to automate and improve the detection of potential infringement, thereby helping businesses safeguard their IP rights. Through these tools, AI provides a marked improvement over traditional patent enforcement that has relied heavily on manual reviews and keyword searches. These older methods tend to be slow, costly, and prone to error.
AI tools can identify and analyze similarities between patents to detect potential violations faster and more efficiently. These include, for example: (a) NLP that enables models to interpret complex technical language in patents and identify conceptually similar claims; (b) computer vision that examines diagrams and drawings in patents to identify visual similarities that may be missed by text-based searches; and (c) ML that adds predictive analysis by not only identifying similarities between patents but recognizing patterns in historical infringement cases and assessing which patents are most at risk. Together, these capabilities allow AI to understand text and drawings, track trends, and forecast infringement—functions that manual reviews and keyword searches cannot match.
Businesses can leverage AI-powered tools to automate and enhance patent infringement detection. AI enables real-time monitoring of global patent databases and product portfolios, alerting businesses to new filings or emerging products that pose infringement risks. It can also map patent claims to corresponding patents or products to automate the generation of claim charts, and conduct comprehensive prior art searches to assess validity, streamlining litigation preparation and reducing human error.
AI offers a powerful solution for monitoring and enforcement, given the challenges associated with the volume of global filings, complex patent language, and the cross-border nature of IP. The result is significant savings in terms of overall cost and time.
AI and IP: Navigating IP challenges with AI training and generated outputs
The prominence of the adoption of AI in the global marketplace raises complex IP questions around both the data used to train AI models and the content generated.
Training AI models can create significant legal uncertainty under Canadian copyright law, as it often involves large-scale text and data mining (TDM) of copyright-protected works—such as articles, images, and audio—without consent from rights holders. Currently, Canada has no clear exemption for TDM, and while fair dealing allows limited uses like research or private study, its application to mass AI training remains unsettled. This exposes businesses to compliance and litigation risks, as illustrated by recent legal filings.
For example, various Canadian news publishers have commenced a claim in Ontario against OpenAI alleging unauthorized scraping of copyrighted articles to train ChatGPT (filed November 2024). The news publishers claim copyright infringement, breach of website terms, and circumvention of technological protection measures. Similarly, a proposed class action against two AI service providers commenced in British Columbia claims unauthorized use of datasets containing thousands of pirated books and removal of copyright management information (filed July 2025). These cases signal that Canadian courts will be asked to scrutinize AI training practices, making robust compliance strategies essential for businesses that intend to deploy AI in a commercial context.
Canadian courts have also been asked to address whether AI-generated outputs that mimic protected works constitute infringement and whether such works can be copyright protected. The ownership of AI-generated works remains a live question—while the Copyright Act appears to require human authorship, AI-assisted works may qualify for protection if a natural person was involved through the exercise of meaningful skill or judgment. This issue is central to a Federal Court application brought by the CIPPIC, which seeks to expunge a registration listing an AI system as a co-author (CIPPIC v Sahni). In its application, CIPPIC argues that AI cannot be an “author” and that minimal human input fails the originality requirement.
While legislative reform may be forthcoming, there is uncertainty around infringement risks and ownership of AI-generated works. Business can nonetheless implement practical strategies to reduce exposure and safeguard their IP, such as by auditing and documenting all training datasets, logging prompt/output events, using licensed or openly available data, and ensuring contracts with employees, vendors, and licensors include clear obligations, indemnities, and warranties that AI systems do not infringe third party rights.
Managing privacy and confidentiality risks in AI workflows
Generative AI tools—such as large language models and image generators—also pose unique privacy challenges. These systems are trained on massive data sets that often include personal information and can inadvertently reproduce sensitive data in outputs.
AI systems often process large volumes of publicly accessible information during training, including data scraped from public sources. While this data may appear “public”, it frequently contains personal information and proprietary content that may be subject to privacy laws. Although the legality of large-scale scraping remains unsettled, Canadian courts have indicated that privacy protections apply to public data. In Alberta, one court held that scraping publicly available images for facial recognition violated provincial privacy laws, confirming that public accessibility does not automatically remove privacy protections (Clearview AI v Alberta, 2025 ABKB 287). An order was also issued in British Columbia, where the Court required the company to stop collecting and delete scraped personal data, reinforcing that consent may still be necessary even for publicly available information (Clearview AI Inc v BC, 2024 BCSC 2311).
Beyond personal data, businesses face heightened risks of exposing proprietary or confidential information when using AI tools. Inputs from employees or vendors can appear in generated outputs or be stored in ways that compromise confidentiality, especially if AI service providers retain prompts for model improvement. Weak confidentiality terms, default cloud settings, caches, and backups that retain confidential data further increase the risk of leaks. Sharing trade secrets without safeguards may undermine protection, making it critical to understand confidentiality implications before inputting sensitive data into AI systems.
Although Canadian regulators and courts have not yet fully clarified the legality of large-scale scraping for AI, they emphasize that generative AI must comply with existing privacy laws. Businesses can mitigate risks by adopting a multi-layered approach. This includes ensuring lawful authority and consent for data collection, limiting sensitive information in prompts, implementing strong data governance and encryption for storage and transmission, and conducting regular audits and vulnerability assessments to address potential weaknesses.
Conclusion
AI is transforming IP enforcement from a reactive, manual process into a proactive, data-driven strategy. By combining real-time monitoring, automated claim charting, and predictive analytics, these tools reduce infringement risk and litigation costs for businesses. However, its rapid adoption also introduces potential legal challenges, for example in copyright and privacy.
Canadian courts are already scrutinizing AI training practices and outputs, while regulators emphasize compliance with existing privacy laws. Businesses should adopt rigorous compliance strategies to mitigate risk and protect both IP and sensitive information until such time as legislative reforms can provide some long-awaited clarity.
***
Geoffrey Mowatt co-leads the Blakes IP Litigation group. He is a certified specialist in patent law and a registered patent and trademark agent whose practice spans complex patent and trademark disputes. He regularly advises on IP issues across various industries, including pharmaceuticals, biotech and technology.
Diane Hwang is an Associate in the Blakes IP Litigation group. Her practice focuses on patent, trademark and copyright disputes. She also assists with regulatory and compliance matters, including those related to pharmaceuticals, medical devices and healthcare.