In October of 2023, the Government of Canada announced a public consultation on "Copyright in the Age of Generative Artificial Intelligence."1 The submissions received in the context of this consultation were recently made public.2
Below, we highlight trends observed across the submissions and share a few notable solutions put forth by various stakeholders. To help us "mine" through this volume of text and data, we enlisted the help of a generative AI assistant. We asked the assistant to summarize each one of the 98 submissions and identify whether each submission 1) was favourable to amending the Copyright Act, 2) was favourable to AI model training as a form of fair use and 3) contained any creative suggestions for how to treat generative AI from a legal perspective. We used the AI assistant's responses in drawing conclusions about the nature and scope of the submissions.
Context
Generative AI, or Gen AI, refers to platforms that recognize patterns in the data they are fed or trained on to produce "new" content. Content is generated when the platform applies what it "knows" to information provided by end users in the form of prompts.
Beyond the question of whether AI-generated content is protected by copyright, issues arise with the data used to train AI models. Given these uncertainties, the consultation was meant to generate a conversation on possible statutory avenues.
The Canadian government was not the first to examine these issues. Notably, the U.S. Copyright Office launched a similar initiative in early 2023 and received over 10,000 comments from stakeholders.3 A more modest number of submissions was received in the course of the Canadian consultation, with 98 submissions in total.
Conflicting Stances
We note that a diverse set of stakeholders responded to the call for submissions. Many of them are associations representing the creative industries, such as cinema, television, radio and publishing. Companies and other private sector stakeholders operating in the field of AI are also represented, as are universities. Various actors from the field of intellectual property, such as the Intellectual Property Institute of Canada (IPIC), the Canadian Copyright Institute (CCI) and FICPI Canada (Federation Internationale des Conseils en Propriété Intellectuelle) also shared their views.
The submissions range from alarming to optimistic. Such sentiments are reflected on issues like whether it would be appropriate to amend the Copyright Act to address generative AI uses or expand the fair use exception for AI model training. On one end of the spectrum, certain stakeholders believe that infringement is inevitable and should therefore be legitimized under an exception. On the other end, others believe that the current protections should be leveraged to put the ongoing infringement to a halt.
More than half of the submissions are in favour of amending the Copyright Act.4 The proposed amendments vary from the addition or removal of specific provisions to more general clarifications, such as clarifying the definition of "author." A small number of stakeholders are of the view that the current framework is adequate as is and that courts are best placed to analyze fact-specific circumstances and interpret the law to adapt it to technological development. Yet others chose to stay away from explicitly taking a position on this issue.
Interestingly, more than a third of the submissions are favourable to accepting AI model training as a form of fair dealing.5 The majority of those are submissions from AI industry stakeholders and the academic sector, in particular institutions conducting research in the field. Certain submissions in favour of the fair dealing exception nonetheless advocate for more transparency, responsible use, and creator compensation and consent. Among the remaining two thirds, some are silent on this issue, while others vocally oppose the inclusion of an exception for AI model training and text and data mining (TDM) activities. For example, certain collective societies consider that TDM activities constitute substantial reproductions of copyrighted works and therefore require licenses from rights holders.
Creative Solutions
Many submissions suggest statutory avenues to address concerns relating to generative AI. Common suggestions include:
- Limiting authorship to works created by humans;
- Limiting copyright protection only to AI-generated works involving effective human control;
- Introducing a specific exception for TDM;
- Implementing an opt-out or opt-in system allowing authors to decide whether their copyrighted work can be used for AI training; and
- Implementing transparency requirements for developers and operators of AI systems about their use of copyrighted material.
The submissions were also rife with creative suggestions that may inspire innovative policy choices if the Canadian Government went ahead with a reform of the copyright regime. Here are a few examples:
- Distinguishing between "Art" and "Algorithmic Result";
- Allowing collective licensing and direct licensing with users for TDM activities;
- Taxing large companies that use generative AI to redistribute revenues to authors;
- Creating a separate legal framework outside of copyright to reward creators;
- Distinguishing between TDM for non-commercial research purposes and commercial purposes; and
- Amending the statutory damages framework in the Copyright Act to preclude its application to end users.
What's Next?
Following the end of the consultation, the Canadian Government stated that the submissions received will help inform its policy efforts.6 Two main policy avenues are foreseeable: (1) a legislative approach through a reform of the Copyright Act, or (2) a more passive approach leaving the matter to the courts. Given Canada's ambition to become a leader in AI regulation, it may not be surprising for Parliament to seek legislative amendments.
Fasken's Intellectual Property and Technology, Media and Telecommunications groups will continue monitoring and providing updates on copyright policy. Please do not hesitate to contact us if you have any questions about the Government's consultation, generative AI, or copyright.
Footnotes
1. See the bulletin our team published in November on this announcement.
2. See all submissions received here.
3. U.S. Copyright Office, Copyright and Artificial Intelligence.
4. To be exact, 55 out of the 98 submissions propose amendments to the Copyright Act, which amounts to approximately 56%.
5. This view is expressed in 37 out of the 98 submissions.
6. Innovation, Science and Economic Development Canada, Copyright policy.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.