Column

A Flurry of Filings: Canada’s AI Litigation Landscape Evolves in a Single Month

One of the earliest projects that was launched at the University of Victoria’s AI Risk and Regulation Lab was a mapping initiative that tracked both how artificial intelligence (AI) is regulated and litigated. To date, litigation tracking has primarily been focused on cases arising from the United States and internationally as until November 2024, there was virtually no domestic litigation to discuss. That changed recently when two lawsuits were filed in the month of November, signaling that Canada is now joining an international surge of AI-related legal disputes. In this column I will briefly review the two recently launched cases and discuss their possible implications for the regulation of AI in Canada.

AI and Data Harvesting

The growth of AI has been fueled largely by access to vast datasets that serve as the foundation for machine learning models. These data sets can include everything from news media—which offers a rich source of factual and contextual information—to publicly accessible case law, a longstanding cornerstone of the justice system that is relied upon by lawyers, judges, academics, and self-represented litigants. Historically, Canadian law has facilitated relatively broad avenues for accessing and using these sources within the confines of copyright and fair dealing. Yet, as AI-driven data extraction operations become more technologically sophisticated and economically meaningful, established legal doctrines are being tested.

CanLII v. Caseway

The Canadian Legal Information Institute (CanLII) is a non-profit organization that all readers are likely familiar with that is dedicated to providing free online access to Canadian legal information. Founded to enhance open access to justice, CanLII’s database is an important resource in the Canadian legal landscape. However, when Caseway—an AI legal research company —purportedly engaged in large-scale data extraction from CanLII’s website, allegations arose that it was doing so in violation of CanLII’s terms of use. On November 4, 2024, CanLII filed a Notice of Civil Claim against Caseway and related entities in the Supreme Court of British Columbia.

This dispute underscores a key tension: while the public has free access to CanLII, the database itself represents a significant investment of time, effort, and resources. The challenge posed by Caseway’s alleged conduct is whether commercial actors can appropriate these resources wholesale, bundling and selling them or incorporating them into AI-driven tools, without meaningful compensation. If courts find that Caseway’s activities fall outside acceptable boundaries, we might see the emergence of new judicial guidance that clarifies when large-scale data extraction from legal databases crosses the line into unfair or infringing conduct.

Canadian Media Companies v. OpenAI

Later in the same month, a consortium of leading Canadian media organizations, including the Globe and Mail and the Canadian Broadcasting Corporation, initiated legal action in the Ontario Superior Court of Justice against Open AI. The crux of their complaint revolves around OpenAI’s alleged use of copyrighted journalistic content to train its AI models. While OpenAI has maintained that its practices are consistent with applicable laws, the media companies argue that using their proprietary content to develop a commercial AI tool amounts to unauthorized reproduction and distribution of copyrighted works.

What distinguishes this case from more traditional copyright disputes is the sheer scale and nature of the alleged infringement. Unlike a straightforward scenario in which a publisher reprints an article without permission, AI systems copy, process, and internalize large swaths of text to “learn” patterns. The process is more akin to data ingestion than traditional publication. This raises a novel question for Canadian courts: does using copyrighted text as training data constitute a form of infringement that falls outside established exceptions such as fair dealing?

The outcome of this litigation has the potential to reshape how Canadian law views the training of AI models. A ruling that data extraction for training purposes is inherently infringing could force AI developers to seek licenses, implement more robust filtering measures, or drastically reduce the scope of data they use. Conversely, a ruling that recognizes some form of implied license or fair dealing exception for AI training would give developers latitude to innovate, potentially at the cost of reducing content owners’ ability to control the use of their works.

Copyright, Data Extraction, and Fair Dealing

At the heart of these cases is a re evaluation of long-standing legal concepts. Canadian copyright law and the fair dealing exceptions have traditionally provided a somewhat flexible framework. For example, fair dealing for the purposes of research, private study, or education has historically permitted a variety of limited uses that, while technically “copying,” serve the public interest by improving access to knowledge.

But do AI models operate as part of “research” or “education”? Should courts interpret fair dealing in a way that accommodates the machine-driven ingestion of data, given that no human directly reads every line of the text involved? The evolution of these doctrines may hinge on how judges perceive the purpose and effect of AI training. If courts view AI training as an indirect but vital form of research or transformative use, they might carve out new doctrinal space for such practices. If they see it as a commercial shortcut that exploits creators’ investments, the decisions could swing the other way.

Foundational Jurisprudence

With these high-profile cases pending, Canadian courts stand at a critical juncture. The jurisprudence that emerges may set the baseline for how we treat AI-driven data extraction for years to come. Given the novelty and complexity of the issues, it is unlikely that a single decision will provide a definitive answer, however, the initial judgments will offer signposts and guiding principles. Yet, one thing is clear: the legal battles unfolding around AI, copyright, data extraction, and fair dealing are more than a passing trend. They represent just the beginning of what I predict will be a significant wave of litigation moving forward. They also represent an opportunity for Canadian courts to issue foundational jurisprudence that will shape the parameters of innovation, information sharing, and intellectual property rights for years to come.

Note: Generative AI was used in the preparation of this article.

Start the discussion!

Leave a Reply

(Your email address will not be published or distributed)