Copyright
Things to know
- Canadian copyright law does not provide specific rules for AI systems, but existing principles — especially relating to authorship, reproduction and fair dealing — are relevant when AI tools are developed, trained, and deployed.
- Authorship and ownership of AI-generated works is an open legal question in Canada. The default rule is that the author is the first owner of copyright (subject to certain exceptions). There is no definition of “author” in Canada’s Copyright Act, but copyright jurisprudence suggests that an author must be a natural person.
- Text and data mining (TDM) exemptions are not expressly identified within the Copyright Act. While TDM may arguably be justified under a fair dealing exception to infringement, the legal status of TDM activities, including the reproduction of works to create datasets for training models, is unsettled.
- Fair dealing provides a potential defense for certain uses of copyrighted material by AI developers. Canadian courts have not ruled, however, on whether large-scale AI training constitutes “research” within the meaning of the exception. The fair dealing exception applies only to certain enumerated purposes, including research, private study, criticism, review and news reporting. To qualify, the dealing must also be “fair” based on a multi-factor analysis.
Things to do
- If you are training a machine learning model, evaluate whether the source material is protected by copyright and whether your use could be covered under the fair dealing exception. When possible, seek licenses or use public domain/openly licensed data.
- If you are deploying an AI model, be aware that output may infringe third-party rights if the model reproduces substantial parts of training data.
- If you are fine-tuning an AI model, assess copyright compliance in light of the new datasets and outputs, particularly where fine-tuning may lead to memorization or output similarity.
- If you are deploying an AI system, anticipate requirements to make copyright-related representations and warranties in commercial contracts. Representations may include that your system does not infringe third-party rights and that appropriate permissions or licenses are obtained.
Useful resources
- “Consultation on Copyright in the Age of Generative Artificial Intelligence,” Innovation, Science and Economic Development Canada, 2023
- “Why AI could mean more work, not less, for copyright lawyers,” Law Times, March 12, 2025
- “Time to talk about ownership of AI-generated intellectual property assets,” Osler, December 13, 2021
Next
