Generative AI tools are evolving quickly, and their impact on teaching, learning and research is significant, posing both opportunities and challenges. The following resource addresses copyright considerations for faculty, librarians, staff, and students at the University of Toronto (U of T) who are using these tools in their university activities. If you have questions about copyright or need further support, please contact us.

The content of this page is evolving and subject to change. Last updated: August 24, 2026

Quick Tip: Many of the copyright and licensing issues described below are resolved with the appropriate use of one of U of T's vetted and supported generative AI tools.

Note: The information and resources below are made available in support of faculty, librarians, staff, and students who are performing their university activities in Canada. If faculty, librarians, staff or students are using AI tools that implicate copyright law while residing outside of Canada, they should familiarise themselves with the legal and policy frameworks that govern AI use in the jurisdiction in which they are currently based.

On this page:

What is Generative AI?: Generative AI is predictive technology that can create or revise content, including text, images, video, and code. It is generally developed by training a model on many example texts, images, audio, and/or other content (“training data”). The model gleans underlying patterns and structures in the training data. In use, when prompted by a user (with a “prompt” or an “input”), the technology uses that understanding to produce new data (an “output”). As such, the process for developing these tools is very copy-intensive.
Generative AI tools include ChatGPT, Gemini, Microsoft Copilot, Claude and Cursor. Many other tools incorporate generative AI features: chat and text suggestions are common functionalities that are enabled by the technology. You may find generative AI in email clients, writing, and drafting tools, or within integrated development environments (IDEs), to name a few. Other models might include text-to-image or text-to-video functionality. To understand if a tool or application you use incorporates generative AI, review the relevant user documentation, privacy policy, and terms of use or service.

What practical steps can I take to respect and protect copyright when I use generative AI tools?

Choosing a Tool

Choose a tool that respects the copyright of others (and your own). Review its terms of service to ensure that you agree with how it will treat your input (including storage, training, licensing, and re-use). Ensure you review terms of use regularly, as these may change.
Consider using a pre-vetted tool (See Quick Tip above).

Providing an Input to Generative AI Tools; Prompting

Consider whether you have the right to access, download, ‘scrape’, reproduce, or otherwise use copyright- or Technological Protection Measure (TPM)-protected material in your prompts or other input. Remember that your use of some material may be governed by contract law (e.g., by the terms of library content licenses) as well as copyright law.
Remember that any content you enter (your own or that produced by others) may become part of the tool's training dataset, may inadvertently resurface in response to other prompts, or may be stored as "machine learning solution data" that has many secondary uses and potential misuses.
Ensure you review terms of use regularly, as these may change.

Using Output

Consider that while you can use these tools to revise, enhance, or create content, you may or may not own or hold copyright in the generated works that result from your use.
Record how you use generated output in your work.
If you are submitting your work (e.g. in class or for publication), be explicit in how you have used generative AI tools in your process.
Review attribution guidelines according to the style guide you are using (refer to the University of Toronto Libraries (UTL) Citing Sources Guide).
Before submitting your work for publication, consider publisher policies and guidance on generative AI use in the research and authorship process, as these may vary.

Training a Model

Consider whether you have the right to access, download, ‘scrape’, reproduce, or otherwise use copyright- or TPM-protected material before you use it to train a model. Remember that your use of some material is governed by contract law (e.g., by the terms of library content licenses) as well as copyright law.
Where appropriate, seek appropriate licenses (including open licenses) or use public domain data.

Deploying a Model

Be aware that output may infringe third-party rights if the model reproduces a substantial part of copyright-protected material either in its training or in its outputs.
If your model was trained using library-licensed resources, contact copyright@library.utoronto.ca before deployment, as many library licenses restrict the making available of tools trained on data derived from licensed content, or tools that may compete with or duplicate the function of vendor-provided content and discovery tools.

What copyright considerations are implicated by the development and use of generative AI tools?

From a copyright perspective, generative AI tools present many new and interesting issues. This is an evolving area of law and policy and our understanding and guidance will develop as new policies, regulations, and case law become settled.

The lawfulness of using copyrighted material for AI model training is a question currently litigated and discussed by courts and policy-makers worldwide. A critical question has emerged: does the non-infringing nature of an AI's output absolve the tool developer of liability for the data used during the training process?

A more detailed list of unresolved issues includes:

Training Data

Is the use of copyright-protected works in training generative AI models or in other machine learning contexts (for example, in text and data-mining activities) ‘use’ under the Copyright Act? If so, does that use fall under fair dealing or an exception to copyright infringement?
What is the effect of contracts and technological protection measures (TPMs) on the analysis?

Prompts or Inputs

What license does a user grant or what representations do they make with respect to their input in a generative AI tool? What copyrights or other rights do users hold in their prompts?
To what further use, including training, can these inputs be put by AI tool developers?

Output

Is copyright protection extended to AI-generated content? If so, who is the owner of the copyright in generated works?
If no copyright protection is extended, are related rights affected? For example, does a performer have rights in their performances of an AI-generated song?
How are moral rights implicated?

Liability

If there is copyright infringement in respect of any generative AI use, who is liable (e.g. the developer, the owner, or the user of the AI system)?

Evidence and Provenance

How might a copyright-holder know or prove the use of copyright-protected works in generative AI training?
If copyright subsistence or ownership is sought by a purported author, how might it be proved that generative AI was used (or not) in the creation of the work?

Remedies

Are the current remedies for copyright infringement sufficient, and against whom may they be pursued?
Are the legal presumptions about liability and copyright subsistence appropriate?

Jurisdiction

Which laws apply to activities related to generative AI use? Does it matter in which jurisdiction the training, prompting, or generation occurs?

The novelty and scope of the technologies involved with generative AI mean that there is significant legal uncertainty on many of these issues. It is unclear if these issues will be settled through the courts or if legislative changes will be required. Cases before the court and policy work being done as of the date of writing include:

An application has been filed with the Federal Court of Canada (Samuelson-Glushko Canadian Internet Policy and Public Interest Clinic (CIPPIC) v Sahni, Court File No. T-1717-24) asking the Court to expunge or amend a copyright registration naming an artificial intelligence as a co-author. (Courts have previously found that an author must be a human being or a group of human beings.)

In Toronto Star Newspapers Limited v. OpenAI Inc., 2025 ONSC 4685 (CanLII), https://canlii.ca/t/kf0w9, the plaintiffs are suing the defendants for copyright infringement, breach of Terms of Use, and circumvention of technological protection measures.

In Canadian Legal Information Institute v 1345750 BC Ltd. et al. SCBC Action No. S-247574 (also known as CanLII v Caseway), CanLII alleges that an AI research company engaged in large-scale data extraction from the non-profit organization that hosts Canadian legal information (including court decisions, legislation, and secondary sources).

As of late 2025 multiple Canadian proposed class-actions have been initiated with respect to AI Tool developers use of “shadow libraries” to train generative AI models. See Kerry Clare et al. V Meta, MacKinnon v. Meta. And Mackinnon v Anthropic PBC.

The Government of Canada published the final report generated by the Consultation on Copyright in the Age of Generative Artificial Intelligence conducted between October 12, 2023, and January 15, 2024 respecting the effects of generative AI on copyright and the marketplace.

The proposed Artificial Intelligence and Data Act, tabled in 2022, did not proceed through the legislative process in the last session of parliament.

Professional associations, courts, and unions have developed guidelines and policies for the use of generative AI.

Our understanding will change as new legislation, regulations, case law, and policies develop.

What material can I use to prompt, train, or tune a generative AI tool?

Consider using content that is openly licensed or in the public domain. In all cases, ensure you are attuned to the individual copyright status, license conditions, and/or contractual terms relevant to the resource you intend to use.
The Creative Commons FAQs on AI and CC Licenses indicates that “if someone uses a CC-licensed work with any new or developing technology, and if copyright permission is required, then the CC license allows that use without the need to seek permission from the copyright owner so long as the license conditions are respected.” There are many repositories of CC-licensed material, including the Wikimedia Commons.
You may find the Internet Archive, the Library of Congress, Europeana, and TSpace to be helpful for sourcing content that is CC-licensed or in the public domain.
If you would like assistance determining whether you can use content in generative AI tools or would like access to larger quantities of textual data, please contact your liaison librarian for support with your request.

Can I use full-text library- licensed e-resources (such as journal articles or eBooks) in generative AI tools?

Generally, no. UTL signs contracts with various vendors and publishers, which set out specific terms and conditions that users are responsible for complying with when accessing content. While what is permitted is not uniform across all resources, most agreements do not allow for the uploading of copyright protected works to third-party platforms, including Generative AI tools.
Note, however, that U of T's vetted and supported tools, such as the protected enterprise edition of Microsoft Copilot Chat, are exceptions to this in most cases, as input and output are not saved, used to train models, or made available to Microsoft or other tool providers; and the University of Toronto Information Security Team has determined that vetted tools are approved for use with data up to Level 3 on its Data Classification Standard.
Consult UTL's Guidelines for using library-licensed content for more information.

Are there other U of T resources I can consult to learn more?

Many other considerations - both opportunities and risks - are relevant to generative AI use and development, including those related to:

Environmental issues
Pedagogy
Ethical labour practices
Economic ramifications
Privacy rights
Security and data protection
Indigenous rights
Accessibility rights

The following University of Toronto resources address these issues and more in the context of research, teaching and learning:

All content on this page is licensed under a CC BY 4.0 license.

Generative AI Tools and Copyright Considerations

Related to

Scholarly Communications and Copyright Office (SCCO) Services