Consultation paper: Consultation on Copyright in the Age of Generative Artificial Intelligence

Table of contents

• 1. Introduction: A Shifting Marketplace
  • 1.1 Goal of this consultation
• 2. Discussion
  • 2.1 Text and data mining (TDM)
    • 2.1.1 Previous stakeholder engagement
    • 2.1.2 Approaches in other jurisdictions
    • 2.1.3 Call for evidence and possible way forward for Canada
  • 2.2 Authorship and ownership of works generated by AI
    • 2.2.1 Previous stakeholder engagement
    • 2.2.2 Approaches in other jurisdictions
    • 2.2.3 Call for evidence and possible approaches in Canada
  • 2.3 Infringement and liability regarding AI
    • 2.3.1 Previous stakeholder engagement
    • 2.3.2 Call for evidence regarding liability
• 3. Conclusion

Consultation on Copyright in the Age of Generative Artificial Intelligence PDF - 421  KB

1. Introduction: A Shifting Marketplace

Recently, a major shift occurred in the digital marketplace with the proliferation and massive public use of generative artificial intelligence (AI) systems. These systems can generate novel creative content in a wide variety of forms and even of quality comparable to works created by talented human artists. This includes the generation of text (e.g., ChatGPT), images (e.g., DALL-E), and music (e.g., MusicGen), often prompted by short instructions from users.^{Footnote 1} Generative AI systems operate using models trained on vast datasets of text, images, or other data. These models build a representation of patterns identified in the training data using various machine-learning techniques.^{Footnote 2}

In the wake of the widespread use of generative AI tools, an increasing number of stakeholders in the creative industries have expressed concerns about the impact of this technology on the copyright framework. For a number of stakeholders, they see AI undermining rights holders' ability to consent to the use of their creative works and to receive due credit and compensation.^{Footnote 3} Particularly, they are concerned about the uncompensated use of copyright-protected works in the development of AI systems, the potential for AI-generated outputs to infringe existing copyright-protected works, and the lack of practical enforcement remedies for rights holders.^{Footnote 4} The evolution of AI technologies arises in a context where the digital shift was already challenging the creative industries, a sector that contributed around $60 billion to the Canadian economy in 2022.^{Footnote 5}

For its part, much of the AI industry continues to express concerns about the uncertainty surrounding the application of the copyright framework in the context of AI. According to some stakeholders in the AI industry and some scholars, this uncertainty may chill domestic investment and impact opportunities for Canada, notably as a destination for AI model development and training.^{Footnote 6} With the Pan-Canadian Artificial Intelligence Strategy and a number of successive investments, the Government has positioned Canada as a world leader in AI. The strategy was launched in 2017 with an initial investment of $125 million, to which the Government is adding more investments over the years, including $443.8 million announced in 2021.^{Footnote 7} These investments support organizations such as Scale AI, a business-led consortium working on AI-powered supply chains in Canada. Scale AI has the potential to contribute more than $16.5 billion to the Canadian economy by 2028, while creating 16,000 new high-skill jobs.^{Footnote 8}

As creative industries begin to be directly confronted with the impacts of generative AI, driven by the release of ever more sophisticated AI systems in the marketplace by developers and deployers, the Government is looking to engage further with stakeholders and Canadians on copyright policy issues related to AI. In the past years, the intersection of AI technologies and copyright has been the focus of consultations,^{Footnote 9} and legislative developments,^{Footnote 10} as well as a number of lawsuits in several jurisdictions,^{Footnote 11} and continues to fuel lively discussions globally. In 2021, the Government of Canada conducted A Consultation on a Modern Copyright Framework for Artificial Intelligence and the Internet of Things, in which it asked for stakeholder feedback on three copyright policy areas related to AI:^{Footnote 12}

• Text and data mining – i.e., whether any clarification is needed on how the copyright framework applies to the use of copyright-protected works and other subject matter (e.g., a performance or sound recording) in the training of AI systems;
• Authorship and ownership of works generated by AI – i.e., how the copyright framework should apply to AI-assisted and AI-generated works; and
• Infringement and liability regarding AI – e.g., who are the persons liable when AI-generated works infringe copyright-protected works.

The Government received 38 submissions addressing at least one of these areas.^{Footnote 13} Stakeholders widely commented on the use of copyright-protected works in the training of AI systems. Stakeholders from the technology sector, scholars, and user groups generally argued for an exception making clear that the use of works in text and data mining activities (TDM) does not require additional authorization from rights holders.^{Footnote 14} Creative industries were of the view that a new exception is not desirable, as it would preclude rights holders from receiving fair compensation for the use of their works in TDM activities.^{Footnote 15} As for the two other policy areas, many stakeholders expressed the view it was premature to take a position on authorship and ownership of AI content, and very few submissions addressed questions regarding copyright infringement and liability raised by AI.

Since the 2021 consultation, the Government has remained active and engaged on AI policy issues. In June 2022, it introduced in Parliament the Artificial Intelligence and Data Act (AIDA) through Bill C-27, which includes a number of obligations for the responsible development and deployment of AI in Canada.^{Footnote 16} In response to rapid advancements in generative AI, Canada launched a consultation in August 2023 and recently released a code of conduct for the development and management of generative AI systems.^{Footnote 17} While AIDA would establish general cross-sectoral standards for the development, deployment, and use of AI in Canada, its scope is designed to complement existing legal frameworks, such as copyright, and is thus not the vehicle to address many concerns about the impact of AI on creative content.

1.1 Goal of this consultation

The purpose of this consultation is to continue the important fact-finding work to inform copyright policy in an era where content, including content that seems creative and original, can be routinely generated by an AI system. Canadians are invited to review the copyright policy considerations in light of their more recent experience with generative AI. This consultation paper revisits Section 2 of the 2021 consultation,^{Footnote 18} with some revisions to reflect events in the intervening period. Although the marketplace and technologies have evolved since then, the copyright policy issues essentially remain the same. While this consultation focuses on these copyright issues, the Government acknowledges that some stakeholders have raised concerns about the impacts of AI on creators and creative industries going beyond copyright per se as well.

The Government invites stakeholders to share views on potential copyright policy directions described in this paper, as well as evidence of a technical nature. The types of technical evidence sought in this consultation include information about how copyright-protected content is accessed, collected and encoded in training datasets; how training datasets are used in the development of AI systems; whether training datasets form part of AI systems after they are trained; the involvement of humans in the development and deployment of AI systems; and how businesses and consumers use AI systems as well as AI-assisted and AI-generated works. The Government also invites comments on legislative and jurisprudential developments related to AI and copyright in other jurisdictions, as Canada remains mindful of approaches taken by its international partners that could serve the needs of a functioning copyright marketplace.

In considering possible copyright policy options relating to AI, the Government will aim to balance two main objectives:

• To support innovation and investment in AI and other digital and emerging technologies in all sectors in Canada. AI has tremendous potential for society if used ethically and responsibly, and could also drive productivity growth across the economy.
• To support Canada's creative industries and preserve the incentive to create and invest provided by the rights set out in the Canadian Copyright Act (the Act), including to be adequately remunerated for the use of their works or other copyright subject matter.

2. Discussion

Developments in AI are raising a broad range of copyright policy questions that have a bearing on innovation, investment, and remuneration for copyright owners in Canada. Given that the rise of generative AI is the driver behind this consultation, this paper focuses on AI systems that use machine learning models.^{Footnote 19}

The major questions for copyright policy can be grouped into three categories. The first category relates to text and data mining (TDM) and the training of machine learning models, which involves the reproduction of large quantities of data, including those extracted from copyright-protected content (section 2.1). This includes questions about the impact on rights holders' rights under the Act, including when and how rights holders could or should be compensated for the use of copyright-protected content as inputs in the development of AI. The second category relates to the increasing capacity of AI systems to generate or assist in the production of creative outputs, including text, images, and music (section 2.2). This includes questions about how to attribute and determine copyright protection on AI-generated content and AI-assisted content. The third category relates to the use and commercialisation of AI systems and the liability for any infringement that occurs (section 2.3). This includes questions regarding the determination of the persons liable when AI-generated outputs infringe copyright-protected works.

2.1 Text and data mining (TDM)

Text and data mining (TDM) consists of the reproduction and analysis of large quantities of data and information, including those extracted from copyright-protected content, to identify patterns and make predictions.^{Footnote 20} TDM is an essential step in the training of machine learning models.^{Footnote 21} This technique enables the model to 'learn' to recognize and reproduce patterns that will enable it to accomplish certain tasks, including generating poetry, music, or artwork. Beyond generative AI, TDM can also advance science, and help businesses solve problems, innovate, and create more value.

Following the rise of generative AI tools such as ChatGPT and DALL-E, some rights holders have expressed growing concerns about the use of copyright-protected content by developers in training machine learning models, without the rights holders consenting or receiving due credit and compensation.^{Footnote 22} Some rights holders have started offering licenses for TDM,^{Footnote 23} but it remains difficult for them to enforce their rights and seek remuneration for the use of copyright-protected content in TDM activities. On the other hand, the AI industry points to economic opportunities for Canada and argues that the copyright framework should enable the use of copyright-protected content for the training of machine learning models. Because of the large quantity of data often involved in training such models, in particular when sourced from the Internet, obtaining any necessary authorisation from rights holders to make reproductions of the works or other subject matter in the course of these activities could be a significant burden. ^{Footnote 24}

As the Government aims to balance innovation and incentives to creativity, the main copyright policy question arising in this section is whether amendments should be introduced in the Act to clarify how the copyright framework applies to TDM activities and, if so, what those amendments should be. Specifically, the Government seeks to better understand how the Act could or should address concerns raised by creators, while ensuring it does not represent a significant hurdle for AI innovation. The Act exists to promote the creation and distribution of content, to foster investment and job creation, promote just rewards for creators, and to create a thriving marketplace that offers consumers choice and access to diverse content.^{Footnote 25} The Act generally achieves this balance by providing exclusive rights to authors on their works, which prevents others from reproducing all or a substantial part of the works, while limiting these rights through certain exceptions.^{Footnote 26}

Two existing exceptions to copyright infringement could potentially apply in the context of TDM activities: 1) the fair dealing exception for research in section 29; and, 2) the exception for temporary reproductions for technological processes in section 30.71.^{Footnote 27} Fair dealing allows for the use of copyright-protected content that might otherwise amount to copyright infringement under certain conditions.^{Footnote 28} A Canadian court applied this framework in the context of a "web-crawler" TDM activity. In that case, the crawler gathered text and photos from websites to populate the defendant's own website.^{Footnote 29} The court found the activities infringing but uncertainty remains as to whether the same analysis would apply to other types of TDM activities conducted for different purposes.

As for the exception for technological processes, the Copyright Board of Canada interpreted this provision as "intended to capture copies that happen automatically, or without the direct control of the user", and that are automatically deleted once the technological process is completed.^{Footnote 30} Like in the case of fair dealing, there is uncertainty as to whether and to what extent this exception would apply to various TDM activities. For example, while some TDM activities may require making ephemeral copies, other TDM may require copies of works to be stored indefinitely, which would make this provision inapplicable.

In addition to economic rights, the Act also grants authors certain moral rights in their works, including, where reasonable in the circumstances, a right to be attributed as the author where one's work is reproduced or used in other enumerated ways. The use of copyright-protected works to train AI systems may thus raise questions about the need to provide an obligation to disclose what content is used in AI training processes. However, the task of distinguishing protected content from unprotected content may be challenging and raise questions about what would be reasonable in the circumstances.

2.1.1 Previous stakeholder engagement

In response to the 2021 consultation, stakeholders widely commented on the use of copyright-protected works in the training of AI systems. Stakeholders from the technology sector, scholars, and user groups generally argued for clarifying that the use of works in TDM should not require additional authorization from rights holders, either as a result of expansion of the fair dealing exception or through the creation of a new exception. They asserted that TDM falls outside the scope of copyright protection, as the process does not rely on the creative expression underlying the copyright-protected content, nor exploit it. They also argue that access to a greater variety of more recent copyright-protected content in the development of AI technology will help foster innovation and reduce bias in AI algorithms that would occur if the training was limited to public domain content that may not reflect evolving cultural awareness and diversity of voices.^{Footnote 31}

On the other hand, creative industries argued that an explicit exception to copyright infringement for TDM is not desirable, as it would preclude rights holders from receiving fair compensation for the use of works. In any event, they were of the view that any amendment to the Act in this regard would be premature since it could undermine the development of market licensing solutions in the nascent AI industry.^{Footnote 32}

The views stakeholders shared in the 2021 consultation were very similar to those they expressed before the Standing Committee on Industry and Technology (INDU) during the 2018-19 parliamentary review of the Copyright Act. Technology sector businesses and researchers perceived existing exceptions in the Act as not well suited to the needs of TDM.^{Footnote 33} For their part, the rights holder community made few remarks about TDM during the review, but expressed the general view that there are too many exceptions in the Act and that additional exceptions could deprive them of revenues or harm their ability to exploit their works.^{Footnote 34}

2.1.2 Approaches in other jurisdictions

In recent years, some of Canada's international partners have increased clarity in their jurisdiction that the use of copyright-protected content for certain TDM activities does not infringe copyright. Some elected to introduce specific TDM exceptions to copyright infringement in their legislation. For instance, Japan and Singapore both implemented broad exceptions covering TDM activities, while Switzerland and the United Kingdom (UK) have implemented narrower exceptions limited to TDM activities conducted for scientific research and non-commercial research respectively.^{Footnote 35} As of June 2023, the UK is also working with users and rights holders to produce a code of practice on copyright and AI, with the aim of making licences for data mining more available.^{Footnote 36}

Moreover, the European Union (EU) adopted a directive in 2019 requiring its member states to provide two mandatory TDM exceptions: one applying to research organizations and cultural heritage institutions for the purposes of scientific research (Article 3), and another applying to anyone and for any purposes (Article 4). Rights holders can exclude their content from the application of the broader exception set out in Article 4, including through contractual terms and the use of machine-readable means for online content.^{Footnote 37}

The United States (US) does not have an explicit exception for TDM in its copyright legislation. However, certain appellate-level courts have provided guidance about the extent to which copies made as part of TDM activities would constitute fair use.^{Footnote 38} Lawsuits recently initiated in the US by creators for the use of their works in the training of AI may provide more guidance on the application of fair use in this context.^{Footnote 39} Additional clarity could also arise following the consultation on copyright and AI launched by the US Copyright Office in August 2023, which notably seeks stakeholder feedback on TDM.^{Footnote 40}

Finally, it is worth noting that in December 2022, Israel's Ministry of Justice issued an opinion concluding that apart from certain circumstances (e.g. training a system on the works of a single author), the use of copyright-protected materials for TDM is permitted in most circumstances under its existing copyright doctrines and will typically be covered by its fair use doctrine.^{Footnote 41}

2.1.3 Call for evidence and possible way forward for Canada

In light of the recent marketplace shift in the use of AI systems, the Government is seeking more information on the nature of TDM activities in Canada and the views of stakeholders on whether and how to clarify TDM activities under the copyright framework. For that purpose, this consultation paper asks the following questions that arise in the development of possible copyright policy measures regarding TDM:

• What would more clarity around copyright and TDM in Canada mean for the AI industry and the creative industry?
• Are TDM activities being conducted in Canada? Why is it the case or not?
• Are rights holders facing challenges in licensing their works for TDM activities? If so, what is the nature and extent of those challenges?
• What kind of copyright licenses for TDM activities are available, and do these licenses meet the needs of those conducting TDM activities?
• If the Government were to amend the Act to clarify the scope of permissible TDM activities, what should be its scope and safeguards? What would be the expected impact of such an exception on your industry and activities?
• Should there be any obligations on AI developers to keep records of or disclose what copyright-protected content was used in the training of AI systems?
• What level of remuneration would be appropriate for the use of a given work in TDM activities?
• Are there TDM approaches in other jurisdictions that could inform a Canadian consideration of this issue?

All comments on whether and how to clarify TDM activities under the copyright framework are welcome, including additional ideas, legal analysis, and supporting evidence and data.

2.2 Authorship and ownership of works generated by AI

Aided by increasingly more sophisticated TDM, machine learning, and other technological advancements, AI can now create content that is difficult to distinguish from content created by human persons. The creation of a work or other subject matter by AI may involve some degree of human input, either by programmers or users instructing an AI application to perform its task. For instance, generative AI systems often produce creative content prompted by short instructions from users. Over the years, AI systems' capacity to independently generate works or other subject matter has increased and that is only expected to continue. There are now a number of AI applications that can write movie scripts, software, and music with little human input beyond the development of the AI itself.

Although the Act does not explicitly define the term "author", Canadian copyright jurisprudence suggests that 'authorship' must be attributed to a natural person who exercises skill and judgment in creating the work, reflective of the fact that the Act ties the term of protection to the life and death of an author.^{Footnote 42} A human may contribute sufficient skill and judgment in a work produced with the assistance of AI technologies to be considered the author of the work. However, it is far less probable that this criterion would be met for works produced by generative AI systems, such as ChatGPT and DALL-E, based solely on short instructions by human users.

Rapid developments in AI technology, combined with its burgeoning application across various sectors of the economy, lead the Government to consider whether the Act is suited to address questions of authorship and ownership of AI-generated works or AI-assisted works. Moreover, the Government is considering whether, even if the Act is suited to address these issues, additional clarity regarding the authorship and ownership of such works could be provided to create more certainty in the marketplace. In considering these questions, the Government aims to ensure the Act supports creators and the creative industries in Canada, while also fostering Canada's competitiveness in AI, innovation and access to creative content.

As discussed in more detail below, there could be ways of clarifying first ownership of AI-generated or AI-assisted works by reconsidering how to define an author, or even without relying on authorship.^{Footnote 43} For certain works (e.g. works by employees) and other copyright subject matter (i.e. a performance or sound recording), the Act already provides for a first owner of copyright other than an author.^{Footnote 44} However, considering the varying degrees of human contribution in the development and use of AI technologies, the persons who could be the appropriate authors and first owners of AI-generated or AI-assisted works (e.g., AI developer, deployer, or user) remains an open question.^{Footnote 45}

2.2.1 Previous stakeholder engagement

In response to the 2021 consultation, stakeholders from a wide spectrum argued that granting protection to AI-generated works would be a shift from promoting human creativity, which is central to the Canadian copyright framework.^{Footnote 46} In any case, many stakeholders found it was premature for Canada to take a position on authorship and ownership of AI-generated works. They were of the view that this question should be addressed once AI is further developed and implemented, as rapid intervention could result in unforeseen consequences.^{Footnote 47} Some of these stakeholders, including from the creatives industries, were, however, open to exploring the possibility of providing 'thin' copyright protection on AI-generated works (e.g., inspired by the protection regimes for other copyright subject matter) when the circumstances would warrant.^{Footnote 48} Only four submissions genuinely supported a form of copyright protection for AI-generated works. The range of views stakeholders shared in the 2021 consultation was similar to those expressed during the 2018-19 parliamentary review of the Copyright Act.^{Footnote 49}

It is also worth noting that since the 2021 consultation, some academics and commentators expressed concerns about the registration by the Canadian Intellectual Property Office (CIPO) of AI-generated works.^{Footnote 50} Canadian registration provides refutable evidence, which can be challenged in court, that copyright exists in a work and that the person registered is the owner of the copyright.^{Footnote 51} In administering the Copyright Registry, CIPO does not examine the claims made in applications for copyright registration.

2.2.2 Approaches in other jurisdictions

Notable approaches in other jurisdictions to address authorship issues around AI-generated works are those of the UK, Ireland, and New Zealand, which attribute authorship of computer-generated works to the person who arranged for the created work. The UK Copyright Designs and Patents Act states that "in the case of a literary, dramatic, musical or artistic work which is computer-generated, the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken."^{Footnote 52} Recently, the UK Intellectual Property Office consulted on AI and intellectual property, notably asking for views on copyright protection for computer-generated works without a human author. In its response to the consultation, the UK government announced its decision to not change the existing law as there was no evidence at that time that protection for computer-generated works is harmful, and the use of AI to generate creative content was still in its early stages.^{Footnote 53} In the US, works made without any creative contribution from a human actor cannot be protected by copyright but works containing human-authored elements combined with AI-generated images can be, so long as the AI contributions are the result of an author's own original mental conception.^{Footnote 54}

2.2.3 Call for evidence and possible approaches in Canada

AI capabilities to generate creative content continue to evolve, with uncertain ramifications for creators, rights holders, innovators and content users. In light of the recent marketplace shift in the use of AI systems, the Government invites stakeholders to share their views and present evidence in response to the following questions:

• Is the uncertainty surrounding authorship or ownership of AI-assisted and AI-generated works and other subject matter impacting the development and adoption of AI technologies? If so, how?
• Should the Government propose any clarification or modification of the copyright ownership and authorship regimes in light of AI-assisted or AI-generated works? If so, how?
• Are there approaches in other jurisdictions that could inform a Canadian consideration of this issue?

The Government is seeking the views of stakeholders on three approaches that could be considered in Canada to address uncertainties surrounding the authorship and ownership of works generated by AI or created with the assistance of AI. Each of these approaches would have fundamentally different consequences for the marketplace, and introduce unique ancillary policy issues to resolve. These three approaches are presented for discussion purposes and do not foreclose other possible approaches that could come to light as a result of further analysis and stakeholder responses to the above questions.

• Clarify that copyright protection apply only to works created by humans. As a result of this approach, it would be clear that any works generated by AI without a minimum creative contribution from a human author would immediately fall into the public domain for others to use without permission.^{Footnote 55} However, when a human author uses an AI to generate a work and in the process contributes skill and judgment to the work, that human would be the author and first owner, and the work would not immediately fall into the public domain. While this approach could add some clarity for market participants and the courts, the difficulty in differentiating human from non-human contributions to AI-assisted works would remain a challenge.
• Attribute authorship on AI-generated works to the person who arranged for the work to be created. This approach would be similar to the UK's copyright framework. This could require establishing factors to distinguish between AI-assisted works that meet the human authorship threshold from those that do not. This approach could mean that AI-generated works receive similar copyright protection as works created by humans.
• Create a new and unique set of rights for AI-generated works. Such an approach would grant economic rights on AI-generated works to a person who did not provide any original contribution to such works (e.g., AI developer, deployer, or user), without deeming that person an author.^{Footnote 56} Additional evidence would be required to establish which new and different rights would subsist in such works,^{Footnote 57} and to guide a number of policy choices, including the appropriate term of protection and the remedies in case of infringement. There are precedents in the Act of granting rights to non-authors. For instance, the Act grants rights to the maker of a sound recording, who is the person by whom the arrangements necessary for the first fixation of the sounds are undertaken. ^{Footnote 58}

All comments on whether and how to clarify authorship and ownership of AI-assisted or AI-generated works under the copyright framework are welcome, including additional ideas and other possible approaches,^{Footnote 59} legal analysis, and supporting evidence and data.

2.3 Infringement and liability regarding AI

Infringement and liability surrounding AI-generated works is the third policy area that raises questions of copyright law. Given the novelty of AI technologies, Canadian courts have not yet rendered decisions regarding liability for infringement that may result from the use of AI, either through the inputs used to train an AI or through the outputs generated by an AI system in the form of works. This section focuses on the potential for AI to generate works that infringe the copyright in other works, or for an AI application itself to be found to infringe copyright.

First, it could be difficult for a copyright owner who is alleging infringement in the AI application or AI-generated work to identify the person, or persons, responsible and to establish liability in a court. Determining liability and infringement may become increasingly complex as the level of human involvement in AI-generated works decreases and AI's capacity to independently create works increases. These uncertainties pertain both to primary infringement, as well as secondary infringement, which occurs when a person knows or should have known that a copy is infringing copyright and undertakes a secondary act in relation to that infringing copy that contravenes the Act.^{Footnote 60} In the context of AI, secondary infringement might arise when users distribute content they asked an AI system to generate and that content infringes copyright.

Another consideration in this regard relates to establishing infringement by reproduction. A plaintiff must establish that the infringing party had access to the original copyrighted work, that the original work was the source of the copy, and that all or a substantial portion of the work was reproduced.^{Footnote 61} AI presents many challenges to establishing these facts, as it may be difficult to determine whether a programmer, user, some other party, or the AI itself accessed the plaintiff's work in the process of generating or contributing to an infringing work, and whether access by one of those parties can be imputed to the others. It may be even more difficult to establish that a substantial part of a work was reproduced during this process.

In this context, the Government is calling for additional evidence from stakeholders to help guide consideration of whether and how to amend the Act to provide more clarity in the marketplace. Removing ambiguity in these policy areas could have a positive impact on the marketplace, facilitate more efficient enforcement of copyright, and further support innovation and investment in AI. While international discussions regarding liability for infringement by an AI-generated work are taking place in various countries and at WIPO,^{Footnote 62} no country appears to have introduced amendments to their copyright law to provide greater clarity with respect to infringement and liability regarding AI.

2.3.1 Previous stakeholder engagement

The Government particularly invites stakeholder views on copyright liability regarding AI since during the 2021 consultation, a mere 12 submissions discussed this question and did so only very briefly. Some stakeholders were of the view that existing liability rules in the Act are sufficient, while others noted that more evidence of how AI technology is used is needed before reviewing copyright infringement and liability regimes.^{Footnote 63} A few stakeholders, however, provided more detailed views. A stakeholder from the music industry suggested implementing record-keeping requirements for those involved in the development and deployment of AI.^{Footnote 64} Moreover, library associations recommended that in the event there is no protection for AI-generated works, control mechanisms should be implemented to ensure that AI-generated content infringing copyright-protected works is removed from circulation by the owner of the AI system that produced such content.^{Footnote 65}

2.3.2 Call for evidence regarding liability

The Government welcomes receiving both additional evidence and recommendations on possible measures before considering whether and how Canada's copyright framework should be clarified with respect to infringement and liability by AI applications, AI-generated works, and AI-assisted works. To advance policy discussions, the Government welcomes responses to the following questions:

• Are there concerns about existing legal tests for demonstrating that an AI-generated work infringes copyright (e.g. AI-generated works including complete reproductions or a substantial part of the works that were used in TDM, licensed or otherwise)?
• What are the barriers to determining whether an AI system accessed or copied a specific copyright-protected content when generating an infringing output?
• When commercialising AI applications, what measures are businesses taking to mitigate risks of liability for infringing AI-generated works?
• Should there be greater clarity on where liability lies when AI-generated works infringe existing copyright-protected works?
• Are there approaches in other jurisdictions that could inform a Canadian consideration of this issue?

All comments on whether and how to clarify infringement and liability regarding AI are welcome, including additional ideas, legal analysis, and supporting evidence and data.

3. Conclusion

The Government welcomes all comments providing additional perspectives or evidence concerning these issues and other copyright policy questions related to AI. Input could include reactions to possible approaches discussed, suggestions of other options and further supporting evidence and data. To participate, submit your feedback in our online consultation form no later than January 15, 2024. The use of this form is critical to facilitate the accessibility of content in a timely manner. Comments received through the online form will be made publicly available following the close of the consultation. If you have any questions, please contact us by email at copyrightconsultations-consultationsdroitdauteur@ised-isde.gc.ca.