Les informations de ce site Web à été fournie par des sources externes. Le gouvernement du Canada n'assume aucune responsabilité concernant la précision, l'actualité ou la fiabilité des informations fournies par les sources externes. Les utilisateurs qui désirent employer cette information devraient consulter directement la source des informations. Le contenu fournit par les sources externes n'est pas assujetti aux exigences sur les langues officielles et la protection des renseignements personnels.
Geneva, September 17, 2021
Ref: Canada - A Consultation on a Modern Copyright Framework for Artificial Intelligence and the Internet of Things
The International Publishers Association (IPA) is the world’s largest federation of national, regional and specialist book publishers’ associations. Established in 1896, our membership comprises 86 organisations from 71 countries around the world, including the Association of Canadian Publishers (ACP), the Association nationale des éditeurs de livres (ANEL) and the Canadian Publishers’ Council (CPC). The IPA is based in Geneva and is an accredited observer at the World Intellectual Property Organization as well as an accredited non- governmental organisation (NGO) enjoying consultative relations with the United Nations.
We welcome this opportunity to offer the views of the international publishing industry represented by the IPA. Our comments are based on the importance of appropriate copyright protection as a key condition for the sustainable development of creative industries. We believe it is important to recall the essential role of copyright to support and reward creativity as a driver for copyright policy and law in Canada. Economic studies by the World Intellectual Property Organization (WIPO) provide evidence that a strong copyright protection and enforcement framework is of key importance to sustain creative industries’ contributions to local culture, communities, education and economies and is a necessary condition of investment in those industries.
AI & Publishing
The publishing industry has been an active contributor to the deployment and development of AI based products, services and platforms and of machine learning technologies. Technologies that might be included under a concept of AI cover multiple areas in the publishing industry, such as machine generated content (e.g. books, snippets), support to scouting and manuscript selection, editing services (e.g. data checking), automated peer-review services to facilitate human peer-reviewing, optimized search and delivery services (e.g. recommend content to readers, enhanced search features, cross-references for research reliability), market analysis and marketing strategies.
In the publishing industry, AI has been a topical issue, not only in the value chain (e.g. multi-layer integration of AI in production and distribution of published works), but also as the subject of industry-based research. Research contributes to assessing the impacts of AI on the industry’s development trends with a forward- looking view and to identifying the main questions that require consideration and additional research.
Research done so far evidences pressing questions, not exclusively related to copyright policy, but that instead arise from high-level issues and principles that can be impacted by the integration of AI technologies in producing and delivering published content to consumers. Some of the main topical questions are related to ethics of AI development and use, security, data privacy and transparency. In addition, priority issues arise from industry-driven concerns related with fostering development of new business models, allowing publishers to reap the benefits of their investments in deploying and integrating said technologies, through licensing in the marketplace. Below we quote examples of industry-based research and some of the relevant findings on the impacts of AI.
In 2018, Elsevier published the report “Artificial Intelligence: How knowledge is created, transferred, and used”Footnote 1, providing insight into “research output, collaboration and mobility for China, Europe and the United States”. The conclusions of the report seem to indicate that pressing questions are related with ethics: “The seeming underrepresentation of ethics in AI research, despite the urgent imperative for ethical AI, remains one of the most pressing questions posed by the report.”
In October 2019, Gould Finch and Frankfurter Buchmesse published the report “The future impact of artificial intelligence on the publishing industry”Footnote 2, the outcome of 6 months of research based on more than 300 interviews with industry professionals. The paper focuses on assessing operational impacts of AI: “Given the varying stages of development of different AI technologies, it is too early to definitively state how they will change the publishing industry–but without question the impact will be immense.” This report also recommends steps to implement AI into business, with an important caveat that “Implementing AI should not be an end in itself. The best way to assess the use of AI is by basing it on a challenge you may be currently facing. Analyze organizational areas, tasks, processes and services, which from your point of view can potentially be optimized by data-driven or automated steps. Identify problems from the user’s point of view and define clear, measurable goals. In order to take ROI into account in the following steps, also determine what costs are associated with this problem today and what value the planned solution could create.”
In October 2020, the UK Publishers Association published a report from Frontier Economics “People Plus Machines: The role of Artificial Intelligence in Publishing”Footnote 3, which “highlights the huge importance that AI will have to the UK’s publishing industry, showing that while use of AI is still in its relative infancy in publishing that the industry is at a watershed moment”. With regards to key policy issues, the report “revealed three areas of concern or opportunity for publishers’ AI investment activities:
- Providing certainty for investment through a stable intellectual property framework.
- Promoting R&D collaboration between publishers, AI-tech SMEs and academia.
- Helping publishing SMEs access AI investment finance and skills.” The report also explains the current main trends in terms of applications of AI in the publishing industry: “The range of current AI uses cases by the industry is broad. Some of the most common industry-wide applications include identifying market trends to inform content acquisition, carrying out copyright infringement checks, running language and grammar checks, recommendation engines and demand forecasting to inform marketing strategy and manage stock levels. Some academic publishers also use AI to help researchers organise and share their content and identify relevant research. A few education publishers are creating interactive adaptive learning solutions and there is evidence of ML being used to automate consumer book pricing.”
In April 2021, STM published the white paper “AI Ethics in Scholarly Communication – STM Best Practice Principles for Ethical, Trustworthy and Human-centric AI”Footnote 4. The paper underlines the importance of transparency obligations as a key issue in the remit of how AI impacts copyright protection: “Clarity and transparency are required in the use of IP and copyright, and as part of any liability regime. AI systems can use huge volumes of copyright materials in the training process and as part of any commercial deployment, therefore transparency obligations may be necessary to enable rights holders to trace copyright infringements in content ingested by AI systems.” The white paper also highlights that copyright owners must be able to license their works for AI purposes, as a key principle to sustain creative industries’ continued production and distribution of works: “The owners of content, including datasets protected as copyright works or as protected subject-matter under a related right, should be rewarded in a manner consistent with the aim of copyright to encourage creativity and innovation, and in the case of related rights, incentivizing investment. It is critical that AI systems function within the intellectual property systems that incentivize the development of high-quality input, regardless of whether the quality is a measure of vetting, structuring, or curation of the information used as input.”
The study “L’intelligence artificielle dans l’art et les industries culturelles et creatives”, by Octavio Kulesz and Thierry Dutoit, published by the Organisation Internationale de la FrancophonieFootnote 5, mentions de debates at UNESCO and the implications on issues such as cultural diversity, pointing out that the role of the public sector at this stage should be to promote much needed debates and informative work, rather than focusing on immediate legislative action.
In Canada, the white paper “Artificial Intelligence and the Book Industry”Footnote 6, published in September 2020, by authors Tom Lebrun and René Audet, funded by the Conseil de recherches en sciences humaines du Canada (CRSH), within the framework «Littérature québécoise mobile», underlines the important role of licensing, by saying that “standard contracts authorizing use of works protected by copyright for the purposes of private AI research must be put in place. The added value of such contracts is that they would also be of great importance in establishing a coherent strategy for use of AI in the book industry.”
A New TDM Exception Should Not be a Priority for Canadian Copyright Policy and Law: Licensing is the First and Most Efficient Response to AI Needs.
While we appreciate the Government of Canada’s consulting with stakeholders at this very early stage, industry-based research over the past 3 years evidences a constant and fast-paced evolution of what AI means for the industry, how it is deployed and integrated in the value chain, and what challenges it brings on areas such as copyright licensing and infringement. As many of the above-mentioned studies conclude, it is still too early to draw final conclusions about the full impacts of AI in copyright-based industries.
As a first step, we recommend that consideration is given to enhancing the understanding of AI for each creative industry, prior to considering immediate legislative action. A complete understanding of the opportunities and challenges for creative industries is a key condition before steps are taken towards amending Canadian copyright law in this regard. Appropriate assessment of the impacts of policy options on creative industries is crucial to achieve one of the objectives outlined in the consultation paper: “Support Canada’s cultural industries and preserve the incentive to create and invest provided by the economic rights set out in the Act. Creators, innovators and rights holders should be adequately remunerated for their works or other copyright subject matter.”
Along these lines, IPA holds the view that introducing a TDM exception in Canadian copyright law is not necessary at this stage and would be detrimental to preserving the incentives necessary for publishers to publish and distribute copyrighted works. The first and most important step is to encourage the establishment and development of licensing markets for AI purposes, based on exclusive rights as set out in existing legal frameworks. As AI based licensing markets are nascent, imposing an exception would not only undermine those markets, but also impair the industry’s ability to continue investing in its own AI integration processes.
With regard to due process on setting out exceptions & limitations, we respectfully submit that the Government of Canada should consider conducting economic impact analysis regarding current and nascent market developments, to ensure that no new exception is enacted unless it is the only way to addresses a genuine market failure, serves a legitimate purpose, is suited to achieving this purpose, and is objectively justified and proportionate according to a fair impact assessment.
Impact assessments, including economic analysis, are crucial to establish an evidence-based approach and achieve a balanced copyright ecosystem. Notably, copyright laws must prevent exceptions & limitations from being used as a basis for new digital or any other businesses to operate as exploiters of protected content in which they have not invested. If this is not safeguarded, exceptions to copyright protection will inevitably violate the Berne Convention’s three-step test and undermine sustainability of local publishing industries, in addition to limiting the offer of legal and sound licensing models.
The first response to access is provided by a vast array of licensing solutions made available by publishers for all types of users. Publishers understand the needs of their different audiences, including corporate users and cultural institutions that might require licenses for AI purposes, and are always available to work closely with users to address those needs. The IPA calls on the Government of Canada to preserve the integrity of copyright protection as an enabler of a license-based markets for AI purposes, instead of disincentivizing publishers’ investments, which would be the ultimate result of introducing unnecessary exceptions in this domain.
Licensing remains the most flexible mechanism through which the use of data subsisting in copyrighted works may be facilitated – to the advantage of both copyright owners and users. Publishers have been developing sophisticated licensing mechanisms to respond to the needs of AI research and machine learning development, which are critical to ensuring that consumers, and in fact AI development, continue to benefit from quality published content delivered through innovative technical solutions. In Canada, these licensing mechanisms include both direct and voluntary collective licensing.
In order to maintain high-level investments required for content production and delivery in the digital age, publishers must be able to recoup their investments through reaping the income generated by licensing. Any policy that would replace licensing with exceptions & limitations as a means to develop AI and machine learning would undermine the development of new markets and prevent future investments, in addition to distorting market competition, in particular if the growth of AI is achieved at the expense of degrading the protection and value of copyrighted content. A premature venture into this matter could undermine creative industries’ integration of AI, damaging the interests of publishers and creators and preventing them from exploiting their economic rights. Instead, national copyright laws, consistent with the existing international legal framework, should continue to foster licensing mechanisms to promote AI development.
We understand from our Canadian members that Canada’s copyright policy still requires repair of the damage caused by the over-broad use of the fair dealing exception by educational institutions, which seems can only be achieved through legislative action. IPA has also recently participated in the “Consultation on a Modern Copyright Framework for Online Intermediaries”, calling on the Government of Canada to take urgent legislative action to update copyright law to effectively tackle online infringement, for example through replacing the obsolete notice-and-notice system with a notice-and-stay down obligation.
IPA respectfully submits that those areas should be prioritized for urgent legislative action, instead of a potential TDM exception, which still requires additional preparatory steps. In addition, other TDM-related issues should be prioritized, in the context of fostering and stimulating licensing solutions, such as transparency and discoverability obligations that enable copyright-owners to license TDM users and prevent infringement.
IPA holds the view that introducing a TDM exception could only be appropriate in a well-constructed, well- regulated copyright regime, equipped with efficient copyright enforcement mechanisms (such as notice and stay down obligations for online intermediaries). In addition, such exceptions should only be considered in jurisdictions where licensing markets are already established based on exclusive rights and where economic and legal impact assessments have been conducted to determine the effects of evidence-based policy options.
Moreover, when these conditions are satisfied, IPA holds the view that any potential TDM exception must be narrowly construed and carefully calibrated to ensure copyright law continues to enable and protect the booming commercial licensing market developed by copyright owners to license TDM uses of published works, which is creating jobs at a rapid rate and sustaining increased investments on a diverse legal commercial offer.
Along those lines, the following cumulative requirements, taking into account the Berne Convention’s three- step test, should be considered:
- Precisely defined beneficiaries and purposes: a TDM exception should only be invoked for scientific research by not-for-profit public interest beneficiaries
- Exceptions to copyright protection must only apply in certain special cases, provided they don’t conflict with the normal exploitation of the work and don’t unreasonably prejudice the legitimate interests of copyright owners.
- In order to avoid harmful effects, an exception enabling unlicensed TDM uses of copyrighted works should
- only be available to specified not-for-profit public interest beneficiaries (e.g. public research institutions) and
- only be invoked to serve purposes of scientific research.
- No commercial uses of data (and the works from which that data is derived)
- When users use copyright owners’ works for commercial purposes, they must seek licenses from copyright owners.
- Allowing commercial uses of copyrighted content under an exception is arguably incompatible with the Berne Convention’s three-step test as it conflicts with the normal exploitation of works, preventing copyright owners from exploiting their works through licensing.
- Commercial uses must not be permitted under an exception, to avoid undermining incentives for commercial users to seek licenses and impairing copyright owners’ ability to exploit their works.
- A balanced TDM exception must only apply for narrowly defined non-commercial purposes, such as non- commercial research.
- In the European Union, Article 3 of the Directive of the European Parliament and of the Council on copyright in the Digital Single Market (the “DSM Directive”) provides for an exception for “reproductions and extractions made by research organisations and cultural heritage institutions in order to carry out, for the purposes of scientific research, text and data mining of works or other subject matter to which they have lawful access.”
- In the United Kingdom, the Copyright, Designs and Patents Act s.29A(1), as amended in 2014, sets out an exception for: “Making of a copy of a work by a person who has lawful access to the work does not infringe copyright in the work provided that—(a) the copy is made in order that a person who has lawful access to the work may carry out a computational analysis of anything recorded in the work for the sole purpose of research for a non-commercial purpose.”
- This policy option, adopted by the EU and the UK, recognizes that an exception of this sort must be carefully crafted so as not to affect the exploitation of the works by copyright owners and their legitimate interests.
- Lawful access to copyrighted works must be a requirement for users to avail themselves of an exception, taking into account machine-readable reservations of rights
- A balanced TDM exception must apply only when users have lawful access to copyrighted works, taking into account machine-readable reservations of rights.
- Copyright owners have been investing to enable an ecosystem to license content for machine learning and other text and data mining applications.
- As a result of publishers’ investments, a vast array of licensing solutions is available for TDM users, especially in the STM and legal sectors.
- Absent a requirement of lawful access, an exception would operate as a permission to use infringing content and would undermine incentives to seek licenses made available by copyright owners, encouraging the creation of a parallel black market, where unlicensed users will be entitled by law to legally avoiding paying licensing fees.
- Machine-readable reservations of rights are technological protection measures (TPM) widely used by the publishing industry, to protect copyrighted works and facilitate users obtaining a license from the copyright owner. They must therefore be taken into account, notably by preventing a user from invoking an exception when machine-readable solutions indicate a reservation of rights through licensing.
- It should be noted that TPM are protected under Canadian law, as a requirement of Canada’s statute as Contracting Party to the WIPO Copyright Treaty. It is therefore a national and international obligation to maintain protection of machine-readable reservations of rights.
- Consequently, a user must not avail itself of an exception when (i) it has not obtained lawful access to the copyrighted works and/or (ii) the works are technologically protected through machine-readable reservations of rights.
- In the United Kingdom and in the European Union TDM exceptions require lawful access to copyrighted content. This is because those jurisdictions have carefully calibrated their laws to be in compliance with their international obligations and to preserve the conditions for their creative industries to thrive and become leading players in a global market.
- In the UK, the Copyright, Designs and Patents Act s.29A(1), as amended in 2014 sets out an exception for: “Making of a copy of a work by a person who has lawful access to the work does not infringe copyright in the work provided that— (a) the copy is made in order that a person who has lawful access to the work may carry out a computational analysis of anything recorded in the work for the sole purpose of research for a non- commercial purpose, and (b) the copy is accompanied by a sufficient acknowledgement (unless this would be impossible for reasons of practicality or otherwise).”
- In the EU, Article 4 of the DSM Directive sets out that a text and data mining exception applies “on condition
- that the use of works and other subject matter has not been expressly reserved by their rightholders in an appropriate manner, such as machine-readable means in the case of content made publicly available online.” In the EU, copyright owners retain a contractual override to opt out of the application of this exception, which is a crucial safeguard to preserve copyright owners’ ability to exploit their works through licensing.
- Apply to the reproduction right only and not enable subsequent uses of the data.
- Text and data mining relies on the reproduction right, not on the communication to the public right. Therefore, an exception for TDM purposes must not enable any communication to the public or any act protected by copyright other than those required for TDM activities.
- Applying an exception to a non-required exclusive right conflicts with the Berne Convention’s three-step test.
- If there is a need to enable certain specific subsequent uses, those uses have to be carefully assessed in light of the Berne Convention’s three-step test to determine whether they are necessary, justified and do not affect copyright owners’ legitimate interests and ability to license their works.
- An example of a justified subsequent use could be exchanges between non-profit institutions for purposes of examination of non-commercial research.
- In those cases, legal provisions should be constructed narrowly to permit only those transmissions that are required for justified purposes, between specific beneficiaries, instead of applying to all acts of communication to the public.
- Users invoking a TDM exception must take reasonable measures to prevent the distribution or transmission of reproductions of works or other subject matter, except in defined circumstances.
- This can be achieved through an obligation requiring users to make use of the exception only through secure systems, including assuming liability for leaks arising from insufficient security requirements and enabling copyright owners to protect their servers when text and data mining takes place.
- In addition, users must delete the text corpus after completion of research, except when trusted and secure repositories can be set up in order to store the content in a way that prevents additional uses, such as distribution and transmission.
- Legal provisions should provide for regulations specifying rules or standards for safe storage.
- In the EU, Article 3 of the DSM Directive sets out the following rules: “2. Copies of works or other subject matter made in compliance with paragraph 1 shall be stored with an appropriate level of security and may be retained for the purposes of scientific research, including for the verification of research results. 3. Rightholders shall be allowed to apply measures to ensure the security and integrity of the networks and databases where the works or other subject matter are hosted. Such measures shall not go beyond what is necessary to achieve that objective. 4. Member States shall encourage rightholders, research organisations and cultural heritage institutions to define commonly agreed best practices concerning the application of the obligation and of the measures referred to in paragraphs 2 and 3 respectively.”
- A requirement that users must compensate copyright owners.
Authorship and Ownership of Works and Generated by AI
The consultation paper outlines the challenges in determining authorship of AI-generated works and suggests areas for additional consideration. IPA agrees that additional evidence of how AI is used to generate new works is needed before a legislative framework related to authorship and ownership can be developed. Hence, we will not provide comments on the three policy options outlined in the paper. However, we would like to note that any regime that is intended to recognize authorship and ownership of AI-generated works must ensure that appropriate exclusive and moral rights are granted to the authors and copyright owners of works used as inputs by the AI technology.
Infringement and Liability Regarding AI
The above-mentioned study published by the UK Publishers Association mentions that carrying out copyright infringement checks is one of the main uses of AI in the publishing industry. However, the study also underlines, based on industry surveys, “that SME publishers (particularly micro-enterprises) recognise the potential benefits of AI investment but lack the resources they need to invest. The costs of researching, acquiring and implementing AI solutions and AI skills were identified as significant investment barriers by SME publishers that responded.”
Along these lines, IPA respectfully submits that the Government of Canada should consider the following principles:
- Online platforms, which are heavy users of AI technologies, should have a role in using AI to carry out copyright infringement checks, and be subject to notice-and-stay down obligations, as they can easily use AI technologies to identify infringing content, as some publishers do.
- Machine-readable reservations of rights, as a means of protecting content, preventing infringement and facilitating identification of infringing acts, must be protected under TPM protection provisions, and not subject to exceptions & limitations.
- TDM users must be subject to obligations to prevent further distribution or transmission of mined copyrighted works.
Internet of Things
On the questions raised in the paper that concern the publishing industry, IPA respectfully submits that the Canadian TPM provisions should be maintained as they are, in order to enable creative industries’ digital business models and encourage the licensing marketplace for AI purposes. We re-iterate that a vast array of licensing solutions is available for all kinds of users to respond to their needs. User stakeholder concerns mentioned in the consultation paper, notably those that state that TPMs « undermine education and access to works, and impede the preservation mandate of libraries, archives and museums » are unfounded. Not only publishers already offer solutions for those users and purposes, but they are also willing to listen and respond to new licensing needs that may arise, including with regard to preservation roles.
We thank you for your time and consideration and remain available for any additional information that you may require.