IP Scholars

September 26, 2021

Sent by e-mail

Minister of Innovation, Science and Industry
Minister of Canadian Heritage
copyright-consultation-droitdauteur@canada.ca

Consultation on a Modern Copyright Framework for Artificial Intelligence and the Internet of Things

Submission by IP Scholars

Copyright and Artificial Intelligence

Dear Ministers / Chers Ministres,

In response to the consultation process on the modernization of the copyright framework launched in the summer 2021, we hereby present our analysis and recommendations concerning some of the issues involved on the interaction between copyright and artificial intelligence (AI), and copyright and the Internet of Things (IoT). Although there are similar structural issues and overlaps involved in AI and the IoT, for practical purposes we are submitting two concurrent separate briefs addressing the interaction between copyright and AI on the one hand, and the interaction between copyright and IoT on the other hand.

This submission concerns the interaction between copyright and AI. The recommendations herein reflect the shared opinion of the intellectual property scholars who are signatories to this brief. They are informed by many combined decades of study, teaching, and practice in Canadian and international intellectual property law. Should there be hearings set up pursuant to the written submission phase, the signatories would welcome the opportunity to appear separately or together before the consultation committee to explain and expand upon specific aspects of this brief and/or other copyright reform proposals not addressed herein.

In what follows, we explain:

• The importance of approaching the questions raised in the consultation with a firm commitment to maintaining the appropriate balance of rights and interests in Canada's copyright system, consistent with a robust principle of technological neutrality.
• The importance of ensuring that text and data mining (TDM) activity can be undertaken in Canada without the threat of potential copyright liability. We therefore propose both an opening up of Canada's fair dealing doctrine to better accommodate TDM activities, and the enactment of a specific statutory provision to confirm that uses of copyright works and other subject matter for TDM (whether commercial or non-commercial) do not infringe copyright.
• The importance of resisting calls to extend copyright protection to AI-generated outputs. We therefore propose maintaining and confirming the existing principled requirements of human authorship and original expression as preconditions of copyright protection, and we caution against any move to establish new neighbouring or sui generis rights in respect of AI outputs. Works generated by AI should remain in the public domain.

As such, we recommend:

• Enacting a broad statutory provision confirming that use of a work or other subject matter for TDM does not infringe copyright. This specific exception should be available to all users, apply to commercial and noncommercial uses, permit the retention and sharing of copies, and be protected from contractual override.
• Amending section 29 of the Copyright Act to make the list of purposes an illustrative list ("for purposes such as") and adding TDM or data/informational analysis as an enumerated purpose therein.
• Confirming in section 2 of the Copyright Act that "author" means a human being/natural person; and confirming in section 5 of the Copyright Act that copyright shall not subsist in a work created without a human author.

Our comments and recommendations will occasionally make reference to the consultation paper "A Consultation on a Modern Copyright Framework for Artificial Intelligence and the Internet of Things" https://www.ic.gc.ca/eic/site/693.nsf/eng/00316.html [the Consultation Paper].

1. Introduction

As the Consultation Paper notes, copyright law has, from its inception, been shaped by major technological developments, from the arrival of the printing press to peer-to-peer file sharing. With the recent hype around developments in AI capabilities, it may be tempting to rush into action, assuming that copyright reforms are necessary to respond to demands for greater protection given the unprecedented capabilities arising from the new and emerging technology. The fact is that the existing body of copyright law already has answers to many, if not all, of the doctrinal questions presented by AI. The legislature and the courts have historically tried to protect the legitimate rights of copyright holders without extending protection in ways that unduly hamper follow-on innovation. But even assuming that emerging AI technology is sufficiently novel to necessitate a reevaluation of the existing copyright framework, the question of how copyright law ought to respond to technological change is not at all new. Copyright law should not create barriers to entry in the development and advancement of new innovations.^{Footnote 1}

(a) Technological Neutrality

A common response to the question of how the law should treat new technologies is simply that the law should be technologically neutral.^{Footnote 2} As the Supreme Court of Canada has repeatedly confirmed, copyright law should develop independently of any specific technology and should continue to apply equally across technologies as they emerge, without favoring or discriminating between new and old.^{Footnote 3}

There is an obvious appeal to this approach for policy makers in the digital age. It presents the promise of sustainable laws in a time of rapid technological change, "future-proofing" the copyright system to some degree by permitting old laws to apply to new technologies. Happily, it may also excuse lawmakers from following the twists and turns of each technological development as it occurs, and likely produces more comprehensible legislation for non-experts and the public who are expected to abide by it. For this reason, it is important that any definition of AI is, in itself, as technologically neutral as possible, referring to objectives and functional effects rather than specific techniques and methods that could be impacted or developed in the future.^{Footnote 4}

Most importantly, however, a robust version of technological neutrality means more than simply extending copyright to protect—or to protect against—new forms of expression.^{Footnote 5} It requires an analysis of the function and effects of new technologies and their impact on the copyright balance, with reference to the policy objectives of the copyright system as a whole; it is a matter of substance rather than form.

(b) Inclusivity and Equality Considerations

Restricting the use of copyright-protected works in machine-learning has significant drawbacks for AI researchers and their projects. Requiring copyright clearance of AI inputs generates huge transaction costs for anything but the most well-resourced projects and actors. Given the extended reach and duration of copyright protection, reliance on public domain or other low liability-risk inputs often means training AI on data that is obsolete, incomplete, exclusionary, and unrepresentative, and which fails to reflect contemporary information and social values. This surely exacerbates the now well-documented built-in biases and discriminatory effects of AI systems.^{Footnote 6} The quality and scope of a dataset has a direct bearing on the quality and operation of the resulting AI.^{Footnote 7} At the same time, the kinds of copies used in AI are not consumed for their expressive content and therefore lie outside the legitimate interests of copyright holders.^{Footnote 8}

(c) Copyright Balance

The copyright system is typically justified as a means by which to encourage authorship and the dissemination of original works. The Supreme Court of Canada has famously described copyright as "a balance between promoting the public interest in the encouragement and dissemination of works of the arts and intellect and obtaining a just reward for the creator."^{Footnote 9} Most recently, a unanimous Supreme Court explained:

"[J]ust rewards for copyright creators provide necessary incentives, ensuring that there is a steady flow of creative works injected into the public sphere.…A proper balance ensures that creators' rights are recognized, but authorial control is not privileged over the public interest.

Ultimately, owners' rights and the public interest should not conflict with one another.…[C]opyright law has long been an 'integrated system that encouraged creators to generate knowledge, industry to disseminate it and users to acquire it and, hopefully, reshape it into new knowledge.'"^{Footnote 10}

The task before today's lawmakers, then, is to ensure that, in the age of AI, copyright law continues to provide the necessary incentives for creators, to encourage the dissemination of works, and to advance the public interest.

(d) Protecting Users' Rights and the Public Domain

A key ingredient of this copyright balance is the protection of user rights. As the Supreme Court of Canada recently reiterated, user rights play "a vital role in…promoting the public interest. The ability to access and use 'works' within the meaning of the Copyright Act are 'central to developing a robustly cultured and intellectual public domain.'"^{Footnote 11} As such, user rights and copyright limits are understood in Canada to be a core component of the copyright system and essential to its purpose. The unanimous Supreme Court recently confirmed this view, noting: "[t]he limits to these private [copy]rights, defined by fair dealing and other exceptions — and circumscribed by the boundaries of the public domain — are therefore essential to ensure that the copyright system does not defeat its own ends."^{Footnote 12}

It is therefore essential to ensure that user rights continue to be protected as novel technologies emerge. This means that any reforms aimed at responding to developments in AI should seek to ensure that they are equally available in new technological contexts. By the same token, they should approach with caution any effort to expand copyright's reach beyond its current boundaries and into what is presently the public domain.

To summarize: Copyright law should indeed keep pace with technological developments in AI. In many respects, however, existing legal doctrine and the principles that inform it will be sufficient to ensure that copyright continues to respond and adapt in a manner consistent with its history and objectives. Where legislative changes are required, these should be aimed at clarifying the law to provide certainty in preserving the traditional balance between authors and users in the new technological environment to ensure that copyright continues to serve the public interest.

It is with these concerns in mind that we turn to the questions presented by the consultation paper.

2. Text and Data Mining

It is important at this stage to define the key concepts discussed in this document. Text and data mining (TDM) involves the processing of large amounts of information in digital form, such as text, sounds, images or data, with a view to gaining new knowledge and discovering new trends. Text mining identifies facts, relationships and assertions that would otherwise remain buried in the mass of textual big data. Once extracted, this information is converted into a structured form that can be further analyzed, or presented in tables, mind maps, charts or other visualisation tools. Machine learning (ML) is a subset of artificial intelligence (AI) which provides systems with the ability to learn from experience "by extracting patterns from raw data" using machine learning algorithms.^{Footnote 13} Hence, TDM is used on an existing dataset to find patterns, while ML is trained on a 'training' data set, which teaches the computer how to make sense of data, and then to make predictions about new data sets.

(a) TDM and the Public Interest

The importance of text mining cannot be overstated, particularly as it is applied to and invoked in the context of machine learning and data analytics. The Canadian government has made significant investments of public funds to help ensure that Canada remains a leader in Artificial Intelligence.^{Footnote 14} The growth and support of AI research and development has clearly been recognized as an important policy goal for Canada and an important ingredient of a strong innovative sector in the 21st century. An arising concern is whether any copyright that subsists in what may be characterized as "AI inputs" is infringed when used in the training of AI systems. Given the vast volume of inputs required as training data for machine learning, restricting the use of copyright-protected works places an enormous burden on AI research and development.^{Footnote 15} It is important to ensure that copyright does not serve as a drag on AI innovation in Canada by increasing the transaction costs and liability risks of doing research and development in AI and machine learning.

TDM also represents an increasingly important research method across a range of scholarly disciplines, as well as in journalism, education, civil society, and a range of commercial research. Text analysis is used in the humanities and social sciences to examine corpi of books, newspapers, social media, transcripts, web sites, historical and government documents, and other data to analyze and document historical events, places, media coverage, topics or themes, and language.^{Footnote 16} Data journalism has become a new field of journalism, in which text mining is used to bring data sources together for reporting, such as on crime, healthcare, or Airbnb.^{Footnote 17} Civil society groups have used text mining to examine housing affordability.^{Footnote 18} Copyright can be a barrier to such initiatives and can have a chilling effect on research, journalism, and civil society projects; at least one of us (Bannerman) has abandoned a research project due to copyright concerns about text mining, having been denied permission by the web site owner to scrape a website. The reasons for ensuring the lawfulness of text and data mining processes are many and drawing increasing support for the public interest they serve.^{Footnote 19}

The consultation paper asks whether amendments should be introduced to clarify how the copyright framework applies to TDM activity, and if so, what those amendments should be. In our view, copies made for the purpose, and during the course of text and data mining do not implicate the rights of copyright owners and so do not infringe copyright. Even if TDM is considered prima facie infringement, it will typically be fair dealing for the purpose of research or private study. While amendments to the Copyright Act may not, therefore, be necessary to bring TDM within the scope of user rights, amendments to the fair dealing provisions and the addition of a specific exception are recommended for the sake of clarity and certainty. Confirming the availability of fair dealing and the lawfulness of TDM (notwithstanding commercial purposes)^{Footnote 20} will serve Canada's strategic interests in innovation—and the public interest more generally—in the responsible development of new AI and machine learning applications.

(b) Current legal barriers to TDM

1. The Uncertain Reach of Section 3 Rights

Generally, the TDM process involves four stages: the crawling and scraping of information, the creation of a target dataset, the analysis of the content, and the publication.^{Footnote 21} In the first stage of TDM, identified sources such as books, newspaper or scientific articles, photos, images, music, or cinematographic works are often copied onto the miner's own storage facilities. Subsequently, selection is made from the collected subject-matter to copy into a new dataset. For the actual analysis to occur, the content analysis software will make a copy in a computer's RAM. In the last stage of TDM, the publication of the results will often take place in the form of an article or report that is either made available electronically or on paper. Whether any of these acts do or even should qualify as copyright-relevant acts pursuant to s.3(1) of the Copyright Act is unclear and subject to debate. It is clear, however, that the law does not treat all copies the same; nor are all copies infringing copies under the Copyright Act and related jurisprudence.^{Footnote 22}

In Society of Composers, Authors and Music Publishers of Canada v. Canadian Assn of Internet Providers (SOCAN v. CAIP),^{Footnote 23} the Supreme Court of Canada found, for example, at paragraph 115: "The creation of a "cache" copy, after all, is a serendipitous consequence of improvements in Internet technology, is content neutral, and in light of s. 2.4(1)(b) of the Act ought not to have any legal bearing on the communication between the content provider and the end user." And, at paragraph 116: ""Caching" is dictated by the need to deliver faster and more economic service, and should not, when undertaken only for such technical reasons, attract copyright liability." Without reference to any specific exception, then, it was understood that such copies did not implicate the legitimate interests of the copyright owner.

There are compelling arguments to be made that the reproduction right should not be implicated by "non-expressive" or "non-consumptive" copies, including the kind of digital copies involved in machine learning. If copyright is ordinarily concerned with copying that "relates to human appreciation of the expressive qualities of that work," then it need not apply to "any act of reproduction that is not intended to enable human enjoyment, appreciation, or comprehension of the copied expression as expression."^{Footnote 24} In other words, copies made purely for machine-learning processes, text, or data mining are not "material" to the copyright scheme, and should be beyond the scope of the copyright interest. This conclusion would be "entirely consistent with the fundamental structure of copyright law because, at its heart, copyright law is concerned with the communication of an author's original expression to the public."^{Footnote 25} It is not yet clear, however, whether copies made for TDM will be regarded by copyright owners or courts in Canada as prima facie non-infringing. The Supreme Court's majority ruling in Canadian Broadcasting Corp. v. SODRAC 2003 Inc., holding that "broadcast incidental copies" are relevant reproductions for copyright licensing purposes, presents some cause for concern.^{Footnote 26}

It should be noted that copyright law does not protect data as such. The Supreme Court has confirmed that "in Canada, as in the United States, copyright protection does not extend to facts or ideas but is limited to the expression of ideas." ^{Footnote 27} This limit is vital to the copyright balance and a robust public domain. It is suggested that TDM makes use of copyright protected texts not as original expressive works to communicate ideas but rather as information or data. Arguably, the copyright work is not used as such, in the sense that it is not experienced or enjoyed by an audience or addressee other that the machine.^{Footnote 28} Again, however, it is unclear whether Canadian courts and copyright owners will understand the limits of copyright protection in this way.

Finally, to regard TDM copies as non-infringing is also consistent with the principle de minimis non curat lex. The law does not concern itself with trifles; and the mere inclusion of a protected work amongst a vast collection of works only for the purpose of machine learning and data analysis is readily viewed as an insubstantial or immaterial use thereof. It is not clear, however, that this interpretation of the de minimis rule will be accepted by Canadian courts or copyright owners.

Recent empirical research shows that strict or unclear copyright rules have a negative impact on the use of TDM techniques for research purposes.^{Footnote 29} As a result it has become common practice to deliberately locate TDM activities in territories with weaker de facto copyright protection and to seek out suitable partners from such territories in international DM collaborations. Uncertainties around the reach of copyright owners' rights in Canada will inevitably have a negative impact on TDM activities. In countries like Canada, where no express exception applies, researchers risk becoming less competitive because of greater copyright restrictions for this novel type of research.

1. The Limits of Licensing Solutions

A system resting primarily on licensing agreements would be insufficient to enable TDM to take place in all instances where it would be socially desirable. First, only a portion of the databases that are interesting for TDM research would be offered as part of publishers' subscription agreements and an even smaller portion would be available under a Creative Commons licence. Without a statutory exception permitting TDM to take place, transaction costs would be too high for parties to negotiate a licence. Second, without a statutory exception recognizing the possibility of TDM, publishers might have little incentive to offer licenses under reasonable conditions. In both cases, too many databases would remain out of reach of researchers. And third, transaction costs would rise if researchers had to reconcile the terms and conditions of non-standard or non-interoperable licences. Requiring researchers to obtain licences from owners of copyrighted works may create burdensome thickets to navigate and contribute to inferior data sets with the potential of copyright owner holdout and underutilization of works functioning as data.

In practice, publishers' licenses vary dramatically from one another, ranging from an outright prohibition on all forms of TDM, to allowing TDM for non-commercial purposes, permitting TDM only through the use of an API, or allowing TDM without any conditions. Some databases are made available only via an API or other (graphical) interface. Consequently, the user may be restricted to the available functionalities in the API as regards the TDM possibilities, but the actual restrictions depend on the available interfaces in the specific case.^{Footnote 30}

Even more problematic are the licensing and conditions of use that some Canadian university libraries put forward.^{Footnote 31} Some of them take an even stricter position towards TDM than the publishers to whose databases they give access. McGill University's licensing conditions state, for example, that "[d]ownloading text to create a corpus of text for analysis, also known as Text and Data Mining (TDM) is not permitted. Platforms that allow Text and Data Mining are listed here". The list is very limited.

Finally, against this background, any collective licensing model that may be proposed is likely to reproduce and potentially exacerbate problems already associated with a licensing approach: restricting available TDM resources to a confined repertoire of works; limiting TDM uses to specific, now-known, and pre-approved TDM techniques; placing onerous administrative conditions on licensed TDM activities; and imposing costs that inevitably disadvantage smaller and non-commercial actors while disincentivizing TDM in Canada. Given the scale of works, text, and data used in TDM, a fair and accurate means of identifying copyright-protected inputs and their rights-holders and distributing payments would be all but impossible to operationalize, even if Canada's collective societies and Copyright Board were better equipped for such a task. Most importantly, however, establishing new collective licensing arrangements is not an appropriate means by which to bring non-infringing activities within the scope of the copyright framework, effectively monetizing user rights. To require a license for fair and non-infringing uses "would extend the scope of the owner's monopoly over the use of his or her work in a manner that would not be consistent with the Copyright Act's balance between owner's rights and user's interests."^{Footnote 32} The consultation paper expresses concern that a broad TDM exception "could decrease economic returns on copyright assets by discouraging licensing activity"; but licensing activities to maximize economic returns should not be encouraged in respect of uses outside the proper scope of the copyright owners' rights.

2. The Limits of Fair Dealing and Existing Statutory Exceptions

The consultation paper notes two exceptions that have been raised in relation to TDM activity: s. 29 fair dealing exception for research and the exception for temporary reproductions for technological processes in s. 30.71.^{Footnote 33} Turning first to s. 29, research and private study are central components of a balanced copyright regime and key user rights in Canada.^{Footnote 34} Unlike the provision of fair use in the United States, fair dealing in Canada must first be identified as for an allowable purpose under s. 29 of the Act before further determination is made as to whether the dealing is in fact fair. The framework for conducting a fair dealing analysis under s. 29 was provided for by the Supreme Court of Canada in Law Society of Upper Canada v. CCH Canadian Limited wherein the court identified 6 non-exhaustive factors for conducting a fairness analysis.^{Footnote 35}

Fair dealing in Canada is a user right, which, the Supreme Court has repeatedly explained, "must not be interpreted restrictively."^{Footnote 36} Given the large and liberal interpretation to be accorded to fair dealing purposes, most TDM is likely to qualify as research, private study, or review, thereby meeting the first step in the fair dealing analysis.^{Footnote 37} Many uses made for machine-learning purposes are likely to be "fair" under the second step, not least because such copies do not compromise the core interests of the copyright owner or substitute for the work of the author in the market.^{Footnote 38} But the need for such a context-specific assessment in relation to each work fed into the AI dataset in order to determine the lawfulness of its use is clearly unmanageable at the scale required by AI. While most TDM could likely qualify as fair dealing upon careful, contextual examination, the context-specific nature of the fair dealing inquiry and the common assumption that the list of enumerated purposes is exhaustive leave room for doubt.^{Footnote 39} Such uncertainty is an obstacle to TDM in Canada.

As the consultation paper notes, the Copyright Act does contain a limited and specific exception for technological processes. This exception is insufficient to meet TDM needs given its limited application only to ephemeral or transitory copies, and so would have to be amended to respond effectively to the concerns raised here. Some "incidental uses: and other miscellaneous uses are expressly permitted under the Act,^{Footnote 40} but these exceptions are insufficient to address TDM in their current form.

It should be noted as well that technological protection measures (TPMs), which are the subject of a separate submission by the undersigned, can also pose a barrier to the exercise of any fair dealing or specific statutory exception that might be created for TDM activities.^{Footnote 41} By creating a technological barrier to accessing the underlying work and providing that circumvention of that barrier is an infringing act regardless of any underlying users' rights, anti-circumvention provisions can effectively preempt reliance on such exceptions.

The current Canadian statutory landscape cannot but have a chilling effect on anyone who would like to engage in TDM activities in Canada. This is compounded by the existence of statutory damages provisions that create an additional significant barrier when TDM consumes a very large volume of works, since the potential liability could be staggering.^{Footnote 42}

This situation runs contrary to both Canadian innovation policy and the public interest for copyright law to unduly restrict or distort the progress of AI research and development by obstructing TDM for machine learning and AI.

1. Looking abroad

Several countries have introduced a specific exception in their copyright legislation with a view to promoting lawful TDM activities. Those exceptions vary in their scope and conditions of application. Most of them have only been recently adopted; their practical impact can therefore hardly be measured. Among these, some provisions are also the result of great legislative compromise; they should therefore not all necessarily be taken as models to follow for Canada. A TDM exception must correspond to the Canadian values of technological neutrality, copyright balance and protection of users' rights expressed above.

Japan was the first country in the world to introduce, in 2009, an exception in Article 47septies of the Japanese Copyright Act to specifically permit TDM activities. The rather obscure wording of this provision left such ambiguity that the Japanese Copyright Act was amended in 2018^{Footnote 43}. Two provisions, to be read jointly, now state:

Article 30-4It is permissible to exploit a work, in any way and to the extent considered necessary, in any of the following cases, or in any other case in which it is not a person's purpose to personally enjoy or cause another person to enjoy the thoughts or sentiments expressed in that work; provided, however, that this does not apply if the action would unreasonably prejudice the interests of the copyright owner in light of the nature or purpose of the work or the circumstances of its exploitation:

(ii) if it is done for use in data analysis (meaning the extraction, comparison, classification, or other statistical analysis of the constituent language, sounds, im-ages, or other elemental data from a large number of works or a large volume of other such data; the same applies in Article 47-5, paragraph (1), item (ii));

47-5 (1)(ii) undertaking computerized data analysis and furnishing the results of that analysis;^{Footnote 44}

It is worth noting that this provision applies to any act of data analysis, e.g. whether commercial or non-commercial, and it allows the publication of results. According to Ueno,

The underlying theory behind this relates to the nature of copyright, or the justification for copyright protection that an exploitation not for 'enjoyment' purposes is beyond the inherent scope of copyright because it does not prejudice the opportunities of the copyright holders to receive compensation.^{Footnote 45}

In Europe, the UK was the first country to adopt an explicit TDM exception.^{Footnote 46} Whereas s. 29 of the Copyright Designs and Patents Act (CDPA) already provided for a fair dealing defence allowing limited copying for the purpose of non-commercial research, s. 29A was introduced in the Act in 2013 to expressly allow TDM activities for non-commercial research purposes. The specific exception permits researchers to make copies of any copyright material for the purpose of computational analysis, but it is limited to non-commercial research and to lawfully accessed works.^{Footnote 47} While the scope of both provisions is rather narrow in terms of the non-commercial nature of the research activities allowed, they do declare that a contractual term that purports to prevent or restrict the exercise of the fair dealing or TDM exception is deemed unenforceable. In 2016 France followed the UK example. Without waiting for the outcome of the then ongoing negotiations around the EU Directive 2019/790 on Copyright in the Digital Single Market (DSM Directive),^{Footnote 48} the French legislature enacted the LOI n° 2016-1321 du 7 octobre 2016 pour une République numérique.^{Footnote 49} Article 38, incorporated as Article L. 122-5, 10° of the Code de propriété intellectuelle, aimed at non-commercial public research, is narrower than the subsequently adopted EU exception in that its application is limited only to data 'included in or associated with scientific publications', rather than to any type of work.^{Footnote 50}

Article 3 of the EU DSM Directive expressly permits 'reproductions 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'. Copies of works may be retained for the purposes of scientific research, including for the verification of research results. Any contractual provision contrary to the TDM exception is deemed unenforceable. As a result of intense lobbying from potential TDM users that do not qualify as 'research organisations', the members of the European Parliament and the Council agreed to establish a separate regime for TDM activities that are conducted outside of the scientific research context. Article 4 of the DSM Directive permits reproductions of 'lawfully accessible works' for TDM for any purpose on condition that this activity has not been 'expressly reserved by rightsholders in an appropriate manner'. Reproductions made pursuant to paragraph 1 of Article 4 may be retained for as long as is necessary for the purposes of text and data mining.

On July 6, 2021, the government of the Republic of Singapore presented a new Copyright Bill in first reading. the proposed amendments were passed by Parliament on 13 September 2021 and are expected to be enacted into law in November 2021. If passed in their current state, ss. 243 and 244 would create an exception to allow for "computational data analysis", which includes text and data mining and machine learning. The Bill defines 'computational data analysis' on a work as "using a computer program to identify, extract and analyse information or data from the work" or "using the work as an example of a type of information or data to improve the functioning of a computer program", such as training algorithms. The reproduction and communication to the public of lawfully accessible works for purposes of computational data analysis would be permitted, irrespective of the commercial nature of the activity, if the reproductions of the works are not shared other than to verify the results or for the purposes of collaborative research.^{Footnote 51} Making a copy includes a reference to storing or retaining the copy. Moreover, section 187 of the Bill states that 'any contract term is void to the extent that it purports, directly or indirectly, to exclude or restrict any permitted use under any provision in Division 8 (computational data analysis)'. Note that this new exception would co-exist with a new 'fair use' defence, redesigned in the Bill from the current 'fair dealing' defence.

By contrast, the lawfulness of TDM activities in the United States rests only the possibility that they qualify as fair use pursuant to s. 107 of the US Copyright Act. The Act contains no express exception pertaining to TDM. Several court rulings have confirmed that text and data mining meets the four fair use factors of s. 107 and does not, therefore, amount to copyright infringement.^{Footnote 52}

As the consultation paper notes, such explicit exceptions are increasingly common. The absence of such an exception currently puts Canada at a disadvantage, and it runs counter to Canada's copyright framework and its policy objectives.

Recommendations

The consultation paper notes that there are four possible avenues to permitting TDM under the Copyright Act. These include:

1. "expand the purposes allowed under the fair dealing exception to include TDM;"
2. "amend the fair dealing exception to make it open-ended, like the fair use provision in the United States;"
3. "amend the exception for temporary reproductions for technological processes;" and
4. "create a new dedicated exception specifically for TDM."^{Footnote 53}

We support each of the above separately and in combination.

We agree that there is a pressing need to encourage and facilitate research and development in the field of AI and machine learning in order to advance Canada's innovation policy objectives and the public interest more generally.

It is our view that most TDM activities are already non-infringing in Canada's copyright framework. However, for greater certainty, we agree that explicit treatment under the Act as part of law reform efforts may more clearly and decisively address this issue, thereby encouraging and facilitating TDM and protecting user rights.

We recommend making the fair dealing exception open-ended by adding the words "such as" prior to the list of enumerated purposes, thereby making the list illustrative and not exhaustive. The benefit of a general exception is that it can accommodate unforeseen uses that are fair to the right holder, avoiding the need to add new purposes as new technological capabilities emerge. For further certainly, however, we would recommend specifying that TDM is an allowable purpose under fair dealing by adding "text and data mining", "data analysis", or "informational analysis" to the non-exhaustive list.

We, and the consultation paper, have noted that the exception for temporary reproduction is of limited utility for some forms of TDM due to uncertainty about the applicability of this provision and the need, at times, to store copies of works indefinitely for the purposes of TDM.^{Footnote 54} We therefore support the creation of a specific exception for TDM, much as many of Canada's key trading partners (Japan, the United Kingdom, France, Germany, and the EU), as the consultation paper notes, have done.^{Footnote 55} This could be a stand-alone provision or an addition to current exceptions for technological process or incidental uses.

We strongly recommend that such an exception is broad in scope, available to all users, and applicable without regard to the commercial or noncommercial nature of the TDM activity. Such an exception should permit the retention and sharing of the copies for verifying the results or for the purposes of collaborative research.

It is understood that, in the Canadian context, the explicit exception would complement the general fair dealing provision, providing added certainty by confirming that qualifying TDM uses do not infringe copyright.^{Footnote 56} We strongly recommend that to the extent that a term of a contract purports to prevent or restrict the exercise of the new exception, it be deemed unenforceable.

To ensure that such amendments do in fact facilitate TDM, we further recommend amending the existing anti-circumvention provisions to confirm the lawfulness of TPM circumvention when undertaken for the purpose of fair dealing and other permitted acts in respect of underlying works.

3. Authorship and Ownership of Works Generated by AI

The consultation paper asks whether and how to change Canada's copyright framework to address uncertainties surrounding the authorship and ownership of works generated by AI or created with the assistance of AI.

In our opinion, the copyright status of AI-generated works in Canada is clear: They are not copyrightable works of original expression and so they belong in the public domain. In the absence of a human author who exercises more than trivial or mechanical skill and judgment in the expression of an original work, there is no existing basis on which to claim copyright.

1. A Human Author

Section 5(1) of the Copyright Act provides in part that copyright shall subsist if "the author was, at the date of the making of the work, a citizen or subject of, or a person ordinarily resident in, a treaty country". The clear implication is that an "author" of a "work" is a natural person.^{Footnote 57} This conclusion is further supported by Section 6, which establishes the statutory term of copyright in Canada: "the life of the author, the remainder of the calendar year in which the author dies, and a period of fifty years following the end of that calendar year." The author of a copyright work is thus expected to have a natural human lifespan (in contrast to the potentially infinite existence of AI), permitting the duration of rights to be determined by the date of their death. In addition to the economic rights that copyright protects, the author is also vested with moral rights—these unassignable personal rights would "make no sense other than in relation to human authors," reinforcing the conclusion that works of authorship in Canada's copyright system are the products of human authors.^{Footnote 58}

1. Originality

Originality is the threshold requirement of copyright protection. In much of the commentary around the current capabilities of AI, it is assumed that AI-generated works are "original," but this tends to be based on a misconception: originality is not determined objectively by assessing the work as an output/product; rather, originality is subjective and a matter of process. To be protected by copyright, an author's expression must not be copied and must involve the exercise of their "skill and judgment."^{Footnote 59} Canada's Supreme Court has defined "skill" as "the use of one's knowledge, developed aptitude or practised ability in producing the work", while "judgment" involves "the use of one's capacity for discernment or ability to form an opinion or evaluation by comparing different possible options in producing the work."^{Footnote 60} The amount of skill and judgment involved "must not be so trivial that it could be characterized as a purely mechanical exercise."^{Footnote 61} This Canadian CCH test (much like the US Feist test, which requires a minimum of creativity, and the Anglo-Australian "skill and labour" test) has rightly been taken to mean that "a human author is required to create an original work for copyright purposes."^{Footnote 62} The originality standard attempts to define the exercise of "intellectual effort" involved in the "expression of ideas"—of course, this is not what is involved in the technological processes by which AI generates outputs.^{Footnote 63}

While AI-generated outputs may come to facially resemble original works of authorship, they are categorically different things. Where a human programmer or user has intervened in technological processes such that the final work constitutes their original expression, copyright will attach as a matter of course. Copyright will also protect the original software code on which an AI system runs. But when AI generates outputs, however objectively novel or interesting these may be, they are not original works of expression within the meaning of copyright law.^{Footnote 64} "Although AI has, and will continue to advance rapidly…genuinely autonomous machines are at best decades away…. The current generation of AI machines are tools that assist and enhance human endeavors."^{Footnote 65}

1. Maintaining the Copyright Balance

The copyright balance is appropriately struck when AI-generated outputs remain in the public domain and outside of private control. Giving copyright to AI-generated outputs serves none of the purposes of copyright protection.

Original AI software will be eligible for source code protection under the Copyright Act while the inventor of a computer implementation of the AI software could be issued a patent if the requirements of the Patent Act are met. To extend an additional layer of protection to the software's creations would therefore over-reward software developers while posing real threats to legitimate uses of their products, as well as future innovations.

The creative user of the AI can claim copyright in any outputs that sufficiently involve their skill and judgment in the expression of ideas within the work. Where the work is truly AI-generated (i.e. in the absence of significant original expression by the programmer or user in the final work), no author is denied their "just reward" when that output is unprotected by copyright.

As for the public interest in encouraging and disseminating works of the arts and intellect, even if AI works qualify as such, there is no compelling reason to assume (far less evidence to prove) that they will be under-produced in the absence of copyright protection. The AI itself does not require a copyright incentive, of course, while AI developers are already incentivized by copyright in the software code. As such, no statutory intervention is required or justified. Rather, extending copyright law to protect a new category of unauthored AI-generated outputs would upset the essential balance between protection and access within the copyright scheme^{Footnote 66} and unduly limit the ability of the public domain to incorporate and embellish creative innovation in the long-term interests of society as a whole, or create practical obstacles to proper utilization.^{Footnote 67}

In the absence of a demonstrable need to extend protection to a new category of works, the government should resist pleas to grant private rights over that which is currently public domain. Indeed, this is an area in which maintaining the current boundaries of the public domain is "essential to ensure that the copyright system does not defeat its own ends."^{Footnote 68} The rapid production and proliferation of AI-generated outputs, if protected by copyright, could quickly throw up a thicket of privately-owned obstacles that human authors would be required to create around. Mass-produced AI-generated works, if protected by copyright, would effectively become "copyright landmines,"^{Footnote 69} creating liability risks, depleting the public domain, and impeding the creative activities of Canadians, all while burdening the Canadian taxpayer with funding the judicial process necessary to resolve private claims of this sort.

The consultation paper expresses concern that "the difficulty in differentiating human from non-human contributions to AI-assisted works would remain a challenge." While this may indeed be a challenge, it is, in our view, something of a red herring in this context. Under our copyright framework, there is always some challenge involved in distinguishing between protected and unprotected works—and indeed between protected and unprotected elements of works—and in identifying when and whether works (or elements of them) are public domain (and if not, to whom they belong). The challenge that this may continue to present when distinguishing AI-generated works from human-authored works is simply not a sound basis for sweeping more works into copyright and out of the public domain.

1. Freedom of Expression

It should be recalled that copyright protection places significant limits on non-owners' expressive activities.^{Footnote 70} The Canadian Charter of Rights and Freedoms therefore requires that any new statutorily created copyright-like right is reasonable and demonstrably justified within the meaning of section 1. It would be practically untenable, on the basis of current evidence, to hold that maximizing the production of AI-generated works is a sufficiently pressing and important goal to justify limiting freedom of expression. If the objective is to advance the established purposes of copyright, then we doubt whether protecting AI-generated outputs is rationally connected to that goal at all. Either way, there are many less rights-impairing means^{Footnote 71} by which the government could encourage investment in generative AI without extending exclusive copyright control over a new category of unauthored AI outputs.

1. A Caution Against Sui Generis or Neighbouring Rights

In the absence of any traditional copyright interest, and without grounds to justify extending copyright, some might argue in favour of a sui generis right or a new "neighbouring right" in respect of computer- or AI-generated works. We caution against this approach. The proliferation of new rights in respect of new subject matters upsets the balance between owners' and users' rights in the copyright system, restricting access and use, increasing transaction costs, and limiting the benefits that might otherwise flow from new technologies. Fundamentally, the creation and layering of new rights in response to each new technological innovation is contrary to the principle of technological neutrality.^{Footnote 72} Moreover, as we have seen with respect to, e.g., sui generis database rights, once they are granted such rights will persist even if the evidence subsequently shows that they fail to achieve their intended purpose or have harmful unintended consequences.^{Footnote 73}

Nothing in the above should be taken to mean that AI-generated works have no value or ought not to be encouraged. The public domain is a tremendously valuable resource for future creativity and innovation. It is simply not the case that every intangible thing of value must be protected by intellectual property; nor must everything that involves a substantial investment of labour or expertise be rewarded with the grant of an exclusive right. Private exclusive rights impose social costs—they should be granted only when the benefits outweigh the costs. There is presently no compelling evidence to suggest that new rights in respect of AI-generated works are necessary to advance the public interest, incentivize new innovations, or expand the expressive universe.

1. Looking Abroad

In the United Kingdom, a legislative amendment appears to bring computer-generated works into the ambit of copyright by deeming the author to be "the person by whom the arrangements necessary for the creation of the work are undertaken"^{Footnote 74} (but limiting protection to a flat 50 years and withholding any moral rights' protections).^{Footnote 75} It remains unclear, however, what standard of originality (if any) is to determine the subsistence of copyright in such works. If copyright does not subsist in such works, of course, the deeming of an author is of limited legal significance. Meanwhile, in Europe, where copyright is philosophically regarded as a matter of le droit d'auteur, it is widely agreed that machine-generated works fail to satisfy the threshold requirement that the work expresses "the author's own intellectual creation"^{Footnote 76} (by which it is meant that "it reflects the author's personality" in the sense that "the author was able to express his creative abilities in the production of the work by making free and creative choices").^{Footnote 77} In the US, with its more utilitarian conception of copyright's purpose, originality also requires at least a minimal degree of creativity and entails a human authorship requirement.^{Footnote 78} Even in Australia, which employs a "skill and labour" standard closer to the traditional UK approach, courts have repeatedly insisted that an original work must demonstrably be "the product of human authorship."^{Footnote 79}

Recommendations

The consultation paper notes that one possible response to the emerging capacities of generative AI would be to "clarify that copyright and authorship apply only to works generated by humans or that no copyright may subsist in a work created without a human participating in some shape or form in the creation of the work."

We support this approach.

It is our view that AI-generated works are not original works within Canada's copyright framework and do not currently attract copyright protection. However, for greater certainty, we agree that explicit treatment under the Act as part of law reform efforts may more clearly and decisively address this issue, thereby confirming that AI-generated works without human authors are not protected and belong in public domain.

As such, we recommend confirming, in section 2 of the Copyright Act, that "author" means a human being/natural person.

We further recommend confirming, through an addition to section 5 of the Copyright Act, that copyright shall not subsist in a work created without a human author.

---

We thank the Ministers once again for the opportunity to submit this brief in the context of this important consultation on the modernization of the Copyright Act. Nous vous remercions à l'avance pour l'attention que vous porterez à ce mémoire.

Respectfully / Respectueusement,

Signatories and affiliations:

Carys Craig, Associate Professor, Osgoode Hall Law School, York University

Bita Amani, Associate Professor, Queen's University, Faculty of Law

Sara Bannerman, Canada Research Chair in Communication Policy and Governance (Tier II), McMaster University

Céline Castets-Renard, Chair of Accountable AI in a Global Context, Faculty of Law, University of Ottawa

Pascale Chapdelaine, Associate Professor, Faculty of Law, University of Windsor

Lucie Guibault, Professor, Schulich School of Law, Dalhousie University

Gregory Hagen, Associate Professor, Faculty of Law, University of Calgary

Cameron Hutchison, Associate Professor, Faculty of Law, University of Alberta

Ariel Katz, Associate Professor, Faculty of Law, University of Toronto

Alexandra Mogyoros, Assistant Professor, Lincoln Alexander School of Law at Ryerson University

Graham Reynolds, Associate Professor, Peter A. Allard School of Law, University of British Columbia

Anthony D. Rosborough, Doctoral Researcher, European University Institute

Teresa Scassa, Canada Research Chair in Information Law and Policy, University of Ottawa

Myra Tawfik, Don Rodzik Family Chair in Law and Entrepreneurship, University of Windsor

September 26, 2021

Sent by e-mail

Minister of Innovation, Science and Industry
Minister of Canadian Heritage
copyright-consultation-droitdauteur@canada.ca

Consultation on a Modern Copyright Framework for Artificial Intelligence and the Internet of Things

Submission by IP Scholars

Copyright and the Internet of Things

Dear Ministers,
Chers Ministres,

In response to the consultation process on the modernization of the copyright framework launched in the summer 2021, we hereby present our analysis and recommendations concerning some of the issues involved on the interaction between copyright an artificial intelligence (AI), and copyright ant the Internet of Things (IoT). Although there are similar structural issues involved in AI and the IoT, for practical purposes, we are submitting two concurrent separate briefs addressing the interaction between copyright and AI on the one hand, and the interaction between copyright and IoT on the other hand.

This submission concerns the interaction between copyright and the IoT. The analysis and recommendations (section 4 of this brief) reflect the opinion of the intellectual property scholars (IP Scholars) signatories to this brief and are informed by years of study, teaching and practice in Canadian and international intellectual property law. Should there be hearings set up pursuant to the written submission phase, the signatories would welcome the opportunity to appear separately before the consultation committee to explain and expand upon particular aspects of this brief and/or other copyright reform proposals not addressed herein.

In what follows, we emphasize:

• The importance of approaching the questions raised in the consultation with a firm commitment to maintaining the appropriate balance of rights and interests in Canada's copyright system, within the broader framework of the Constitution;
• That the modernization of the Copyright Act (the "Act")^{Footnote 1} requires a careful examination of the copyright framework within larger observable trends of dominant positions in the marketplace and anti-competitive practices, of the extraction of big (personal) data, and of market and legal infrastructures' heavy reliance on non-negotiated standard terms (as evidence of consent, and to legitimize various commercial practices).

Our comments and recommendations will occasionally make reference to the consultation paper "A Consultation on a Modern Copyright Framework for Artificial Intelligence and the Internet of Things" https://www.ic.gc.ca/eic/site/693.nsf/eng/00316.html [the Consultation Paper].

1. Introduction: IoT and copyright within the broader context of the Constitution, property, competition, data protection and privacy

Throughout this submission, "Technical Measures" refers to devices deployed for the purpose of blocking the access to, or certain uses of works protected by copyright. "Technological Protection Measures" or "TPMs" refers to the legal regime implemented in Canada (and other jurisdictions) that prohibits various acts in relation to Technical Measures.

The impact of copyright law on the IoT, and more specifically on the right of repair and issues of interoperability revolves to a large extent around the current scope of TPMs in the Act. Before discussing the nature of TPMs within the copyright framework, including Canada's international obligations (section 2), it is à propos to briefly situate copyright and the IoT within the broader context of the promotion of innovation, the Constitution, property, competition law and privacy law.

References made in the Consultation Paper to maintaining the right level of incentives to invest and innovate are focused, in the context of the IoT, on right holders and their competitors, while being silent about nurturing innovation and freedom among the users of such innovations.^{Footnote 2} The perils of a selective vision of innovation that justifies strong copyright and other IP protection without due consideration for the innovation of their users have been widely commented upon by copyright scholars.^{Footnote 3} In the IoT/copyright interface, such narrow view of innovation quickly leads to ignoring the basic freedoms that come with, and are enabled by the things that individuals or businesses use and own and which are and should remain out of copyright subject matter and scope. Innovation deserves protection only insofar as such protection does not unduly constrain further (related) innovations, or the ingenuity and freedom of individual users to tinker, to repair, to maintain, to modify, and to build upon the things they own.

Unlike copyright and IP more generally, the IoT largely deals with tangible objects (personal or real property; moveable or immoveable in Québec civil law) that are and should remain predominantly outside the scope of copyright law. Around the time of the introduction of TPMs in 1996 at the international level,^{Footnote 4} the constitutionality of TPMs as a legal device was put in question, as potentially unduly encroaching upon the provincial jurisdiction of property and civil rights.^{Footnote 5} And while the constitutionality of TPMs has never been tested before a court of law, the Ministers should revisit the scope of TPMs, in light of the resulting deeper incursions of copyright exclusive rights into the realm of tangible property in ways that were unimaginable when TPMs were introduced at the international level by WIPO in 1996, and later on in the Act. The Ministers should do so in keeping with the Canadian constitutional framework of federal-provincial division of powers.

The Consultation Paper aptly points to the interaction between IoT, TPMs, interoperability and the competition issues that TPMs may create. We discuss the interface between TPMs and competition further below in this submission.^{Footnote 6} The Ministers should also examine the recalibration of TPMs in light of the broader trend which TPMs directly or indirectly enable: increased surveillance and control by firms through the extraction of business or individual users' personal data.^{Footnote 7} The compounded effects of Technical Measures (reinforced by strong TPM copyright regimes such as in the U.S. or Canada) and the extraction of personal data conferring increased control to suppliers at the expense of the privacy and freedom from interference of their users, are well document in Perzanowski and Schultz The End of Ownership: Personal Property in the Digital Economy.^{Footnote 8} And while privacy and personal data protection are not within the purview of this consultation, there is a parallel to be drawn between the end goals pursued on the one hand, through the strong promotion of TPMs and on the other, the exponential growth of the extraction of (personal) data in the IoT and e-commerce more generally (which contrary to Technical Measures, has been left largely unregulated in Canada and elsewhere). Both trends -strong TPMs, and largely unregulated extraction of (personal) data- have facilitated the acceleration of market dominance, lack of transparency, and various restraints on technology users and their freedoms.

2. TPMs within the copyright legal framework

While there may be debate about the constitutionality of TPMs, there is little doubt that the TPM regime that Canada has adopted, has not only significantly increased the protection of right holders in copyright digital works; it also extended protection and control beyond the copyright works that TPMs were initially allegedly meant to protect. In the IoT, TPMs' reach goes far beyond the medium within which the copyright work is made accessible (CDs, DVDs, streaming services): it extends to cars, smart sensors, smart homes, tractors, medical devices, health tracking devices, all of which have little to do with the original literary, dramatic, musical and artistic works that the Act is meant to protect.

In this part, we address the scope and effect of TPMs, how a sound approach to TPMs needs to be in keeping with balancing competing interests including user rights, and Canada's international obligations regarding TPMs.

• - The scope and effect of TPMs

TPMs were introduced in Canada as part of the Act's last major reform in 2012. This was part and parcel of Canada's efforts to comply to the WIPO Copyright Treaty (1996) and WIPO Phonogram and Performance Treaty (1996).^{Footnote 9} TPMs were among the most contentious issues of that legislative reform, as well as of the several failed reform attempts prior to that.^{Footnote 10} Canada chose to adopt a broad conceptualisation of TPMs as "access controls", borrowed largely from the United States' Digital Millennium Copyright Act.^{Footnote 11} This expanded the purpose and objectives of TPMs envisioned by the WIPO treaties. Rather than be used principally as copyright tools 'in connection with the exercise of rights' and which would restrict acts 'in respect of works', the access control model birthed the 'access right'. In adopting this approach, Canada introduced a distinct set of novel exclusive rights, and a separate regime of exceptions which operate in parallel to the already existing framework under copyright.^{Footnote 12}

The undesirable ramifications of the 'access right' created by TPMs are numerous and have been thoroughly discussed by copyright experts.^{Footnote 13} As we contend in this submission, these ramifications are particularly poignant in relation to IoT devices. By restricting access to copyright works, TPMs transform copyright protection into controls over the use and modification of tangible property. This control is exacerbated by the proliferation of embedded computer systems in myriad devices and components, all of which integrate some form of software as a protected copyright work. Where such embedded systems control the physical components of devices, TPMs become a functional bottleneck for their operation and use. This sets unprecedented restraints on the use of personal property under the sanction of copyright law.

It is important to stress that this copyright-enabled control over the physical functioning of devices is not limited to a few niche products. Rather, it has granted designers and manufacturers of myriad products a legal basis upon which to fashion a new design paradigm. This paradigm is marked principally by user limitation, control, and permissions. Formerly squarely within the domain of computer scientists, so-called "compliance and robustness rules"^{Footnote 14} are now the common domain of product designers in the IoT space, ranging from smart hairbrushes to complex agricultural equipment. Below we canvass two practical manifestations of these techniques and their consequences; namely, the ability to repair tangible property, and the implications for follow-on innovation.

1. Repair

Product manufacturers have used TPMs to inhibit or prevent repair and servicing in several ways. One approach, demonstrated by Apple as part of its iOS 9.2.1 mobile operating system, can detect whether an authorised service person has repaired the device (including iPhones and iPads). Known as "Error 53", this software Technical Measure will disable the device entirely following a repair by an unauthorised technician. This applies particularly to repairs involving interference or replacement of the device's screen or home button – repairs which are incidentally very commonly performed by independent technicians. The disabling of these devices (or "bricking" as it is informally known) led to the Australian Competition and Consumer Commission launching suit against Apple in 2017 on account of its infringement of Australian consumer law. The suit resulted in fines to Apple in the amount of AU $9 million for its use of TPMs as an anti-repair mechanism.^{Footnote 15}

Another way that manufacturers have used Technical Measures and TPMs to inhibit repair is in the access and disclosure of diagnostic information. While a cracked smartphone screen needs little in the way of diagnosis, many complex computerised devices require servicing where the solution is not immediately apparent. This is the case for the Taylor C602 ice cream makers, which are used in McDonald's restaurants around the world to produce soft-serve ice cream and milkshakes.^{Footnote 16} These are very expensive, computerised machines with intricate parts that must operate at certain temperatures, or the machine will fail. The machine is programmed to periodically enter a self-cleaning process which superheats the machine to kill bacteria. Incidentally, many McDonald's franchisees have found that the machines fail during this process and require maintenance. The machines are equipped with Technical Measures which ensure that only Taylor's certified technicians can perform repairs. The system does not allow a user to perform diagnostics or obtain information regarding why the machine has failed or what must be done to make it operational again. Accessing diagnostic information requires the user to input a code, which is not provided to franchisees and is not found anywhere in the machine's maintenance manual.^{Footnote 17}

Frustrated franchisees found relief from Kytch, a third-party peripheral device that can connect to Taylor machines to provide diagnostic information and to provide a user-friendly interface. Effectively acting as tool to circumvent Taylor's Technical Measures, the makers of Kytch received cease-and-desist letters from Taylor, and franchisees later received letters advising them that the use of Kytch voided warranties and posed risks to employees and customers. Opting for the safer approach, Kytch acquiesced to Taylor's demands in 2020 only to discover that Taylor had replicated the Kytch device and begun distributing it to franchisees.

The Taylor/Kytch example shows a use of Technical Measures and TPMs, in conjunction with exclusive service agreements and warranty terms, to deny access to the information necessary to complete repairs. While Apple's Error 53 code is demonstrative of absolute locks on independent repair, Taylor's approach with the C602 machine demonstrates a use of TPMs as part of a suite of statutory and contractual tools to control the use and management of devices long after the point of sale.

The ability to remotely disable devices and deny access to diagnostic information are merely two of many TPM-enabled practices utilised by manufacturers to inhibit or deny repair activities. As IoT devices and embedded systems become increasingly reliant on co-verification techniques^{Footnote 18} and tethered connectivity to OEMs,^{Footnote 19} we anticipate that these techniques will increase and be even more latent in modern product design.

1. Follow-on Innovation

Modern agricultural equipment is produced by a small group of large manufacturers with global reach, including the likes of John Deere, CNH Industrial and AGCO. To suit a variety of crop-specific needs and soil types, manufacturers of shortline or "add-on" manufacturers have sprouted up. These manufacturers principally develop peripheral devices that plug into large OEM machinery to perform specific tasks or functions. Whether attached to the front (known as a "header"), or towed behind, shortline agricultural machinery has proven itself to be an innovative industry in Canada – building on top of and adding value to the dominant platforms. Canada's shortline industry accounts for over $4 billion in revenues generated by over 500 companies.^{Footnote 20} These companies show promise as innovation leaders for many small towns and rural areas.^{Footnote 21}

Until relatively recently, shortline companies in Canada have been able to produce interoperable equipment with OEM machinery on the basis of the ISOBUS standard, a communications protocol and interface that allows for the exchange and use of information between the central computer and add-on device. They have been able to accomplish this through internal research and development, including reverse engineering activities. With the advent of the John Deere X9 combine, however, this interoperability is no longer possible. Relying on a suite of Technical Measures, TPMs, and a bespoke interface, shortline innovators in Canada are being denied participation in follow-on innovation and secondary markets.

Though the Act contains a clear exception which permits circumvention of TPMs for the purposes of "interoperability", it conceptualises such interoperability as purely a relationship between two computer programs. One consequence of this approach is that it fails to accommodate a broader class of technologies (such as embedded systems within physical devices) and by extension, significantly narrows the type of innovative activity that is permitted by the exception.^{Footnote 22} By extending the effect of TPMs into the realm of physical components and devices, there is a clear need to broaden the interoperability exception to accommodate IoT devices and embedded systems, including agricultural equipment. This is because today's follow-on innovation requires interoperability between a broader class of technologies than merely "computer programs". This is particularly problematic in the case of embedded systems and IoT technologies where the distinction between the computer program and other components of the device is not entirely clear. In the absence of reforms to its interoperability framework, Canada risks further cementing an approach to product design which restricts repair and follow-on innovation in the embedded systems and IoT space.

1. Economic and Ecological Consequences

The negative effects of TPMs on repair and interoperability within IoT devices and embedded systems are numerous. From a market perspective, the use of TPMs in this way effectively affords OEMs in dominant market positions a convenient tool to deny others access to an essential facility for market participation. As Canada's shortline industry reveals, this applies equally to the principal market occupied by OEMs as well as secondary ones. The type of activity or conduct restricted by TPMs is often essential for independent repair and follow-on innovation. By using TPMs to reserve secondary markets for themselves, OEMs are engaging in palpable anti- competitive activities enabled by copyright law. Curtailing these effects are particularly important in a country such as Canada, where many large OEMs are located elsewhere, and domestic industries are to a large extent reliant on secondary markets and follow-on innovation.

Importantly, Canadian competition law and policy is generally deferential to the exercise of statutory intellectual property rights. For this reason, the use of TPMs as copyright overreach – beyond what the scope of what copyright justifies – deserves more scrutiny. Our position is that the failure to do so may undermine Canada's competitive landscape.^{Footnote 23} As such, the Ministers should evaluate the role of TPMs within broader issues of market fairness and dominance. This may add crucial insight into the activities of digital platforms and other high-tech players, and how our blind spot regulatory approaches may have contributed or facilitated anti-competitive practices exercised through market dominance.

Beyond the anti-competitive effects of TPMs used to curtail repair and follow-on innovation, there are environmental implications as well. We gladly take note of these concerns in the Consultation Paper. TPMs that preclude independent repair and follow-on innovation can reduce product lifespan and increase maintenance and repair costs. The result is often a market incentive to "throw away and buy new". In this way, TPMs can act as not only impediments to market fairness, but also to the establishment of a circular economy.

Both the manufacture and disposal of IoT devices can take a massive toll on the environment. From the extraction of raw materials from ecologically sensitive areas, manufacturing techniques, shipping and packaging, there are enormous environmental costs for failing to extend the lifespan and repurpose devices. Perhaps even more troubling are the end-of-life impacts and harms to human health through the recovery of rare earth elements in electronics waste.^{Footnote 24} Given the Intergovernmental Panel on Climate Change (IPCC)'s recent Climate Report signalling a "code red for humanity",^{Footnote 25} Canada should make every effort to prioritise the reduction of unnecessary electronics waste and consumption. We believe that an essential step toward this goal is to enable independent repair and follow-on innovation through expanded exceptions to Canada's TPM framework.

Canada would not be alone in enabling the right to repair through an environmental lens. In November of 2020, the European Union adopted a suite of measures under its Directive 2009/125/EC (the "EcoDesign Directive") to promote the repairability of various consumer products. While these measures do not address TPMs per se, they do require manufacturers of certain products to provide access to repair information, replacement parts, and to design products that can be repaired using commonly available tools.^{Footnote 26} Given the global reach of many consumer products and IoT devices, we believe that Canada should take a similar approach and make revising its TPM framework a key part of its strategy.

The necessity of TPM reforms are not merely academic or theoretical musings. Manufacturers of myriad products have shown a keen interest in strictly enforcing their TPMs, and Canadian courts have supported that approach. The most notable example is the Federal Court's 2017 decision in Nintendo v King.^{Footnote 27} It was the first (and remains the only) case centred around TPMs in Canada. It illustrates the compounding effects of broad statutory language coupled with broad judicial interpretation. The case involved "mod chip" devices sold and installed by the defendant which allowed users to play user-generated games as well as infringing copies of Nintendo's games. By affirming that Nintendo's encryption technology and the design of physical components of their consoles constitute TPMs, it emboldened manufacturers' ability to lock out users and market competitors in ways that extend well beyond preventing copyright infringement. In the end, the Federal Court awarded Nintendo over $12 million in damages (including punitive damages) on the basis that the defendant was held to have trafficked circumvention devices in contravention of s 41.1(c) of the Act.^{Footnote 28}

Nintendo v King also shed light upon the inadequacy of Canada's TPM interoperability exception for follow-on innovation. Though not raised by the defendant, the Court inferred an "interoperability defence" based on the defendant's claim that modifications to Nintendo's consoles were for the purposes of allowing users to play user-generated games. In response, Campbell J introduced several factors to establishing the 'defence' of interoperability. One of which is that the primary purpose of the circumvention must be for the purposes of interoperability. Another factor is whether there are "legitimate paths" for developers to achieve interoperability without circumventing TPMs.^{Footnote 29}

The result is that Canada is saddled with both statutory language for an interoperability exception which fails to accommodate broader modalities of innovation, and potentially, an additional factor of necessity.^{Footnote 30} The effect of this requirement is to significantly narrow the application of the existing interoperability exception to be one that fails to provide essential breathing room to innovators within the IoT and embedded systems space to develop innovative products and solutions.

The above demonstrates that Nintendo v. King is instructive on many levels. It shows how the Act's TPMs confer legal entitlements which under the cloak of IP protection, may facilitate commercial practices with anti-competitive effects, including use of exclusive rights to restrain trade, for exclusive dealing, or tied selling restrictions.^{Footnote 31} Overall, the foregoing decision demonstrates the need for statutory revision. It makes clear that market forces alone will not resolve the negative market and ecological consequences posed by TPMs given the unequivocal language in the Act and narrow judicial interpretation. Revisions to TPMs and the related exceptions are essential for the health of Canadian innovation, robust market competition and the environment.

• - TPMs and the balancing of competing interests within the copyright law framework

Earlier prognosis about the implementation of TPMs being substantially at odds with the scope of copyright, and the Supreme Court jurisprudence affirming that copyright law was not just about securing rights and remedies to copyright holders, but to secure legitimate access and use to copyright works, are even more applicable today.^{Footnote 32} The use of TPMs by manufacturers to control and protect uncopyrightable features of physical devices in no way furthers the objectives of copyright and the delicate balance to preserve between competing interests, and significantly narrows the scope of the property rights and uses of such physical devices.^{Footnote 33} Meanwhile, the Supreme Court has continually reaffirmed the prominence of the interests of users alongside the ones of copyright holders within copyright law, as stated in Théberge and in CCH.^{Footnote 34} Similarly, the Supreme Court has highlighted the "limited nature" of copyright holder's rights,^{Footnote 35} and warned against giving "excessive control" to "holders of copyrights and other forms of intellectual property".^{Footnote 36}

The modernization of copyright efforts initiated by the Ministers should be an opportunity to revisit the important décalage that the TPM provisions continue to create, relative to the highest court's continuous affirmation of copyright law's role to promote and preserve the interests of users and the public alongside the interests of copyright holders.

• - Canada's international obligations

The primary source of Canada's international obligations to enact TPMs is found in the WIPO Copyright Treaty ("WCT") and the WIPO Performances and Phonograms Treaty ("WPPT").^{Footnote 37} Concluded in 1996, these international agreements require state parties to enact provisions preventing the "circumvention of effective technological measures that are used by authors in connection with the exercise of their rights...and that restrict acts, in respect of their works, which are not authorized by the authors concerned or permitted by law."^{Footnote 38}

These agreements left few (if any) functional limitations on what types of mechanisms or instruments should be covered by TPM regimes. Nevertheless, the WCT and WPPT set some conceptual boundaries on the purposes and use cases for TPMs. With the caveats that Technical Measures would be used 'in connection with the exercise of rights', and which restrict unauthorised acts 'in respect of works', the WCT/WPPT retained a conceptual link to the exercise and protection for copyright works and unauthorised acts in that context.

When Canada added TPM provisions to the Act in 2012, the United States had already pushed ahead with a more robust and restrictive approach from the WCT and WPPT in the DMCA.^{Footnote 39} The DMCA in many ways divorced itself from the conceptual links to copyright set by the WCT and WPPT.^{Footnote 40} Instead, the TPM regime extends to any Technical Measures which 'effectively control access' to works. The result is to grant the manufacturers of technologies (as opposed to merely rightsholders) a new de facto right to determine user and consumer conduct. It was this "WIPO Plus" approach that Canada ultimately followed in implementing its obligations under the WCT and WPPT.

In addition to the WCT and WPPT, Canada's international obligations over TPMs are governed by the Canada-United States-Mexico Agreement ("CUSMA").^{Footnote 41} CUSMA's effect on Canada's TPMs are threefold. First, it reiterates and formalises Canada's 'access control' approach to TPMs mirrored in the DMCA. Secondly, it requires Canada to enact criminal penalties for wilful circumvention of TPMs where the purpose is commercial advantage or financial gain. Finally, CUSMA requires that Canada confine its exceptions and limitations to Technical Measure protections to a fairly rigid list of purposes or use cases. The one exception to this rigid list is a general "adverse effect" provision found at Article 20.66(1)(h), which enables state parties to:

"...provide additional exceptions or limitations for non-infringing uses of a particular class of works...when an actual or likely adverse impact on those non-infringing uses is demonstrated by substantial evidence in a legislative, regulatory, or administrative proceeding in accordance with the Party's law."^{Footnote 42}

The result is that Canada's international obligations regarding TPMs are now stronger and more comprehensive than they initially were under WCT and WPPT. Through the ratification of CUSMA, Canada has moved far beyond the original minimum requirements for TPMs and rationale found in the WCT and WPPT. Today's TPM framework effectively creates a sui generis right over technological design and use which is bolstered by criminal penalties.

Despite this seemingly inflexible and privileged status afforded to TPMs, CUSMA makes clear that Canada retains the right and ability to enact exceptions and limitations to TPMs when an actual or likely adverse impact […] is demonstrated. We urge the Ministers to consider all flexibilities that CUSMA offers to better curtail the unintended detrimental effects of TPMs in the context of the IoT and beyond.

3. Looking Abroad: The United States TPM framework

As the Consultation Paper is soliciting comments about various possible ways to constrain the effect of TPMs on the right of repair and interoperability, looking at the decades-long history of regulatory exemptions under the anticircumvention provision of the DMCA^{Footnote 43} is instructive.

When Congress passed the DMCA in 1998, it included a number of narrowly tailored statutory defenses. These provisions targeted circumvention undertaken by law enforcement^{Footnote 44} and non-profit libraries,^{Footnote 45} as well as circumvention necessary for encryption research,^{Footnote 46} security testing,^{Footnote 47} some acts of reverse engineering,^{Footnote 48} and the protection of personally identifiable information.^{Footnote 49} With the exception of the broad law enforcement carveout, these defenses are exceedingly narrow, offering precious little comfort to researchers and device owners in practice.

Recognizing the likelihood that other exceptions would prove necessary, Congress provided for a triennial rulemaking, conducted by the Librarian of Congress and the Register of Copyrights, to identify classes of copyrighted works, the non-infringing uses of which are likely to be adversely affected by the prohibition on circumvention.^{Footnote 50} While Congress provided the Librarian and the Register with a list of factors to determine whether an exemption was warranted, the substantive and procedural rules were left largely undefined.^{Footnote 51}

The history of the triennial rulemaking reveals the struggles of the Copyright Office. It has applied a set of shifting and unpredictable standards that have yielded inconsistent and sometimes surprising results. The Copyright Office has answered fundamental questions—like the definition of a "class of works," the consideration of fair use, and the standard of review for existing exemptions—in radically different ways from rulemaking to rulemaking.^{Footnote 52} In the most dramatic example, the Register's refusal to renew a twice-granted exemption permitting owners of mobile phones to remove digital locks that prevented them from lawfully using their devices on competing carrier networks, led to over 100,000 signatures on a White House petition,^{Footnote 53} a Federal Communications Commission (FCC) investigation,^{Footnote 54} a private agreement among carriers to allow unlocking,^{Footnote 55} and an act of Congress overturning the Register's decision.^{Footnote 56} Equally troublingly, exemption proposals commonly raise contested questions of fact and law—from election security and consumer privacy to environmental protection and patients' rights—that are far beyond the expertise of any single agency or official.^{Footnote 57}

In recent years, the rulemaking has yielded a proliferation of narrowly defined exemptions that apply to exceedingly narrow circumstances. It is lawful to circumvent a Technical Measure to repair a smartphone, but not a smart TV, or a car, but not a boat. Crucially, under U.S. law, these exemptions apply to acts of circumvention, but not to the creation or distribution of software tools that enable circumvention, rendering these legal rights all but useless to the vast majority of consumers.^{Footnote 58} Moreover, the process for proposing and securing exemptions is resource intensive, requiring hundreds of hours of legal work, often provided pro bono by law school clinics and non-profit organizations.^{Footnote 59}

While the U.S. experience can serve as a learning opportunity to craft a better process for regulatory exceptions, it also helps illustrate the need for a narrow scope of TPMs, including through broad, permanent exceptions that accommodate reasonable uses from the outset.

Recommendations

The growing prevalence of the IoT reveals more clearly than ever before how TPMs need to be recalibrated in keeping with the objectives of copyright, the Constitution, property rights, and of promoting competitive markets. As such, the most effective way to achieve this goal would be to:

1. Narrow the scope of the TPM prohibitions under the Act, whereby the circumvention of access controls or copy controls for non-copyright-infringing purposes would be lawful, with a non-exhaustive list of such purposes to provide greater legal certainty. The same treatment would apply to the dealing in TPM circumvention technology enabling the exercise of non-copyright-infringing purposes.

In the alternative, bearing in mind the limits but also the flexibilities available under CUSMA, the Act should be amended to:

1. Introduce a new exception that would confirm that the TPM provisions (and other relevant exclusive rights in the Act) do not apply to the right to repair, including for maintenance and diagnostics purposes. We note that private member's Bill C-272 proposing an exception to the prohibition against circumventing TPMs, as well as against dealing in TPM circumvention technology in connection with exercising the right of repair is one approach to consider.^{Footnote 60}
2. Introduce a new exception to encourage follow-on innovation. The existing interoperability exception's affirmation of the "computer programs" language in CUSMA leaves little room to widen its application to broader modalities of innovation. Rather, Canada should look to article 20.66 of CUSMA's "adverse effects" provision to enact an entirely new exception. This would provide innovators with the breathing room to engage in circumvention activities for the purposes of product development and testing, even where the prohibition of such activities does not have immediate market or competition consequences.
3. Additionally, just as copyright holders should not be allowed to contract out of exceptions to copyright infringement through non-negotiated standard form agreements,^{Footnote 61} neither should they be allowed to opt out of exceptions to TPM prohibitions by contract.^{Footnote 62}

In the end, adding more exceptions to the already long list of exceptions to TPMs will only amplify the convolutedness and disjunction that the parallel TPM regime of prohibitions and exceptions has introduced within the Act. Therefore, realigning the TPM provisions to encompass no more than what copyright protects remains the preferred approach for the reasons outlined in this submission.

The Governor in Council should also make use of its regulatory powers under s. 41.21 of the Act, to introduce exceptions to the application of TPMs, in instances where TPMs unduly restrict competition in the aftermarket sector in which the technological protection measure is used,^{Footnote 63} and beyond such instances.^{Footnote 64} This includes the power to require copyright holders to grant access to copyright works restricted by TPMs. To that effect, we invite the Ministers to look to other jurisdictions that have implemented such mechanisms, including France.^{Footnote 65} While those regulatory powers of the Governor in Council confer additional flexibility under the Act, we acknowledge that such powers are restricted in scope, and may not be suited to address several of the issues identified in this brief, including the obstacles TPMs impose on the right of repair.

We thank the Ministers for having the opportunity to submit this brief in the context of this important consultation on the modernization of the Act. Nous vous remercions à l'avance pour l'attention que vous porterez à ce mémoire.

Respectfully, Respectueusement,

Signatories and affiliations:

Pascale Chapdelaine, Associate Professor, University of Windsor Faculty of Law

Anthony D. Rosborough, Doctoral Researcher, European University Institute

Aaron Perzanowski, Professor of Law, Case Western Reserve University

Bita Amani, Associate Professor, Queen's University, Faculty of Law

Sara Bannerman, Canada Research Chair in Communication Policy and Governance

Carys Craig, Associate Professor, Osgoode Hall Law School, York University

Lucie Guibault, Professor, Schulich School of Law, Dalhousie University

Cameron Hutchison, Professor, Faculty of Law, University of Alberta

Ariel Katz, Associate Professor, Faculty of Law, University of Toronto

Alexandra Mogyoros, Assistant Professor, Lincoln Alexander School of Law at Ryerson University

Graham Reynolds, Associate Professor, Peter A. Allard School of Law, University of British Columbia

Teresa Scassa, Canada Research Chair in Information Law and Policy, University of Ottawa

Myra Tawfik, Don Rodzik Family Chair in Law and Entrepreneurship, University of Windsor