The information on this Web page has been provided by external sources. The Government of Canada is not responsible for the accuracy, reliability or currency of the information supplied by external sources. Users wishing to rely upon this information should consult directly with the source of the information. Content provided by external sources is not subject to official languages and privacy requirements.
Submission to the Consultation on a Modern Copyright Framework for Artificial Intelligence and the Internet of Things
Submitted: September 17, 2021
Introduction and Overview
Access Copyright welcomes the opportunity to respond to this consultation. The fields of artificial intelligence (“AI”) and the internet of things present compelling possibilities for technological advancement. Access Copyright, through its innovation lab Prescient Innovations, is experimenting with AI to better protect the works of its rightsholders and to improve its distributions. While it is desirable to provide a legal and economic climate in which innovative companies can progress in these areas, Canada must also ensure that any steps taken to promote these industries does not harm participants in existing cultural industries, including Access Copyright’s member creators and publishers.
As the Canadian courts have repeatedly held, a copyright regime must attempt to balance user and rightsholder interests. In Access Copyright’s submission, Canada’s current copyright laws are sufficiently flexible to deal with circumstances arising from the use of AI technology, including text and data mining (“TDM”). There is no need to open up new exceptions to copyright, which could cause serious and potentially unforeseen harm to creators, including in the writing and publishing spheres. Technological development should not be used as an excuse to impinge on creator and publisher rights, which have already suffered due to the Copyright Act’s fair dealing for education provision. In particular, the existing market for TDM licensing should not be supplanted by a regime allowing TDM actors to use copyright-protected works without obtaining authorization or paying compensation.
2.1 Text and Data Mining
Writing and publishing industries are harnessing AI technology
Canadian publishers and authors support innovation in AI technology, provided that such innovation does not come at the expense of existing rightsholders. New technologies should be encouraged and furthered only to the extent that they provide a net benefit to Canadians, including our country’s valuable cultural industries.
AI technology is currently being harnessed to provide new opportunities for growth in the writing and publishing sector. Numerous Canadian players in these fields are already developing, commercializing, and exporting AI technology to exploit their creative content.
For example, the Globe and Mail has pioneered award-winning AI software that, amongst other things, selects articles for inclusion on the online front page of the newspaper, using an algorithm to determine what is most likely to appeal to an individual reader and drive new subscriptionsFootnote 1. This made-in-Canada solution has been so successful at driving increases in website traffic and new customers that it has been successfully licensed to other markets – benefiting not only the Globe and Mail, but also creating value for its own readers, licensed newspaper publishers, and the authors of published articles.
AI technology is also being used by rightsholders to complement their existing content and free up human resources for more involved activities. In educational publishing, AI is being used to improve student outcomes and reduce teacher workload. For example, McGraw Hill has developed its ALEKS Platform, which is a student feedback tool integrated into digital learning materials.Footnote 2 In the news media sphere, the Associated Press uses AI to produce routine reports of sporting events, such as previews and recaps of scores,Footnote 3 allowing sports journalists to focus on more creative work such as commentary and critique.
Access Copyright, through its innovation lab Prescient innovation, is exploring the use of AI technologies to better protect the works of its rightsholders and to improve its distribution models. The intersection of AI and blockchain technologies will greatly enhance authentication of works and introduce new monetization models for creators. Prescient’s work in this area, notably through the development of its Attribution Ledger offering and the work that it is doing with its partners Canadian Artists’ Representation/Le Front des artistes canadiens (“CARFAC”), Regroupement des Artistes en Arts Visuels (“RAAV”), and the Copyright Visual Arts (“CVA”) on improved management tools for visual artists are examples of how the publishing, writing, and visual arts sectors are harnessing web 3.0 technology to innovate and make the most of limited resources.
TDM licensing works
There already exists a successful global marketplace for TDM licensing of text-based works. Many publishers directly license their works for TDM purposes, and certain copyright collectives also provide such licences where obtaining permissions on a work-by-work basis would be impractical or involve inefficiently high transaction costs. For example, the Copyright Clearance Center (“CCC”), a copyright collective based in the United States, provides a RightFind XML for Mining licence.Footnote 4 Other industry players such as Elsevier, a journal publisher, have also successfully engaged in direct TDM licensing.Footnote 5 Businesses are willing to pay for TDM licences because they recognize the value of the works licensed.
Both the CCC and publishers (including Elsevier) have invested in improving the quality of TDM technology for their customers by providing the content in XML format.Footnote 6 Making content more usable for customers in this way comes at a cost for rightsholders but can be possible (and a prudent business decision) when the use of works is remunerated. This is illustrative of how allowing the licensing market to continue to develop encourages, rather than stifles, innovation. It results in better outcomes for both rightsholders and users of AI technology.
Another example of rightsholders successfully licensing their works for AI purposes comes from the news media industry. Newspapers license their articles for TDM for media monitoring and news aggregation purposes. This licensing market needs to be protected, particularly in light of the challenges faced by newspapers whose content is routinely scraped by third-party web crawlers without authorization. It is important that newspaper publishers retain the ability to monetize and control the uses of their content. Consequently, a TDM exception should not be implemented without further consultation with the newspaper sector. Otherwise, there is a strong likelihood that harm will occur to an industry that the government has recognized is currently at risk.Footnote 7
The existing marketplace for TDM licensing should be protected
Licensing preserves the integrity of copyright by ensuring that creators can control how their works are used. This mechanism is fundamental to ensuring that copyright owners can exploit their works, and that they remain incentivized and able to create and invest in the creative economy. Licensing also provides legal certainty to users of copyright-protected works and, in a modern copyright framework such as Canada’s, is not a meaningful barrier to doing business.
Establishing a new TDM exception could seriously harm or even destroy the Canadian licensing market. It would result in a transfer of revenues from content creators to corporations exploiting the content for profit including, for example, offshore technology giants that have little to no investment in content creation.
There is no evidence that a TDM exception would be more effective in increasing Canada’s competitiveness or fostering innovation than if Canada allowed the existing TDM licensing market to grow organically. If the market remains intact, content creators will be able to afford to continue investing in formatting improvements to make content more easily accessible for TDM uses.
Furthermore, even if Canada created a new exception and became a safe harbour for the free use of works for TDM, it does not follow that there would be an increase in jobs or investment here. Canada is bound by international treaties to extend any copyright exception granted to Canadian users to foreign nationals, according to the principle of national treatment. There would be no competitive advantage to Canadian AI companies if a TDM exception were legislated. It would be extremely simple for a foreign company to locate a server in Canada for the purpose of conducting TDM, without otherwise investing here – essentially taking advantage of weak copyright laws to harvest content for profitable exploitation overseas. Implementing weaker copyright laws than our trading counterparts is not the type of “innovation” that would benefit the Canadian economy.
Rather, a robust copyright regime is necessary to ensure that society can fully benefit from the AI field without harming existing cultural markets. Advances in AI technology, and the creation of high-quality AI-generated works, require access to suitable input data and information. This can only be achieved equitably through a balanced legislative approach that treats content creators fairly. An exception could financially harm creators and publishers including, for example, in the news sector, which could seriously hamper Canada’s creative economy.
The fact that a wide variety of works may be used in a given TDM application should not absolve users from needing to get authorization to exploit the works. Rightsholders – who have expended considerable efforts and resources in developing their works – should be able to determine how these should be licensed, whether directly or through voluntary collective management. Collective licensing provides an efficient means of granting licences for uses of a large quantity of content where obtaining permissions on a work-by-work basis is impractical or would result in unacceptably high transactional costs. As mentioned above, the CCC in the United States already has an established collective licence for TDM.Footnote 8
A targeted TDM exception in the Copyright Act is not necessary
A blanket exception for TDM is not necessary and would be inconsistent with Canada’s international obligations. The Berne Convention for the Protection of Literary and Artistic Works establishes a three-step test for exceptions to copyright, namely, that any exception be (1) limited to certain special cases; (2) not conflict with the normal exploitation of the work; and (3) not unreasonably prejudice the legitimate interests of rightsholders.Footnote 9 Canada is bound to respect this test, as it has ratified not only the Berne Convention, but also other treaties that incorporate the three-step test such as the Agreement on Trade Related Aspects of Intellectual Property Rights (“TRIPS Agreement”)Footnote 10 and the WIPO Copyright Treaty.Footnote 11 A broad exception for TDM uses of copyright-protected works would fail to meet this test, as it would not be sufficiently specific and, by replacing licensed uses of works for TDM purposes, would conflict with the normal exploitation of relevant works to the prejudice of rightsholders.
Instead of taking steps to further limit copyright protection, Canada’s focus should be on ensuring that rightsholders are remunerated when their works are used for AI purposes. Compensated use of copyright-protected works in a functioning marketplace is preferable to allowing mass ingestion of works. The goal should be to find a balance between the interests of users and copyright owners. Fostering a functioning marketplace will pay dividends in terms of increased competition and innovation, wider access and dissemination of works, and the possibility of a more secure future for the voices of Canada’s creators and storytellers.
Laws in other jurisdictions should be studied
While the consultation paper cites laws of certain other countries, these laws need to be studied carefully to ensure they do not undermine licensing opportunities that currently exist or may develop for rightsholders in the Canadian marketplace.
As an example, the TDM exceptions in the EU Directive on Copyright in the Digital Single Market are narrow and limited in scope and the drafters were mindful of the potential harm to rightsholdersFootnote 12 and recognized the importance of ensuring rightsholders are able to license the uses of their works.Footnote 13 As noted by Neil Turkewitz, the emphasis of the EU “Directive as a whole [is] to encourage licensing, including where appropriate via collective management” and the framing of the EU Directive “clarifies that theories of innovation cannot override the need to ensure the effective protection of copyright or be permitted to undermine licensing opportunities for creators.”Footnote 14
Fair use should not be adopted and TDM is not always considered “fair use” under US laws
Access Copyright does not support the adoption of a fair use regime for TDM. Adopting a fair use model would result in significant and expensive litigation.Footnote 15 All businesses, including those in the cultural and high technology spheres, require a legal environment with clear and understandable rules to thrive and attract investment. Establishing fair use doctrines through the common law may take decades of lengthy court cases – not an efficient outcome for AI investors or an inviting prospect for Canada’s already struggling writing and publishing industries.
We further note that, even under the fair use laws developed in the United States, TDM is not always considered fair use of copyright-protected works. The consultation paper states that “certain appellate-level courts have provided guidance about the extent to which copies made as part of TDM activity for commercial purposes would constitute fair use”Footnote 16 (emphasis added). This is an oversimplification of the Authors Guild v Google, Inc. (“Google Books”) and Authors Guild v. HathiTrust (“HathiTrust”) cases cited. In both cases, although the issue of text mining was briefly mentioned, the focus of the decisions was on copying for purposes of search, which is different from text mining and involves different legal considerations.
The Google Books and HathiTrust cases were specific and fact-driven and should not be relied on to address broader questions about copying for purposes of TDM. These decisions focused on the digitization of printed materials to be searched for information about the text of copied works and the courts concluded that the copying was a transformative use that, largely because of certain extreme limitations imposed on the use of the copies, produced no direct commercial benefit to the copier, there was no delivery of significant amounts of the expressive content of the scanned works to the searchers and no meaningful evidence of market substitution that would result in market harm to the copyright owner. The only conclusion that can be drawn from these cases is that the specific copying in those particular circumstances is “fair use” and as such, they provide little value in making generalizations about TDM itself.Footnote 17
A fair use analysis of other acts of copying, particularly those done for commercial purposes or providing substantial access to protected expression from the copyrighted works, could easily produce a different result. It is important to keep in mind that the minimal discussions on text mining in these cases were in the context of converting text from print to digital form in order to make it mineable. That conversion rarely occurs with respect to most text mining today, where the materials being searched often are “born digital” and already possess the characteristics that the print-to-digital conversion added (i.e., the ability to be indexed and searched). As a result, it is open for a court in a text mining case today to come to a different conclusion. In one case in the United States, Fox News Network, LLC v. TVEyes, Inc., the Court found that extraction of copyright-protected content in order to provide access to a database of TV clips owned by a third party was not fair use.Footnote 18 In that situation, the Courts recognized that a plausibly exploitable market for the clips had been usurped from the rightsholder.Footnote 19 It would be unfortunate if the government of Canada were to make a legislative change which required rightsholders to expend significant resources litigating to protect their rights, instead of investing in developing markets to exploit their works.
Conclusion on TDM
A blanket exception is not suitable in situations involving emerging technologies, where the precise application of the technology is unclear. There may be unforeseen repercussions to legislative change, which could undermine both current and future licensing markets. Given the rapidly evolving nature of AI technology and its applications, further review and dialogue, particularly with creators, publishers and organizations who are currently engaged in licensing TDM activities, is warranted once the evidence provided in this consultation is received.
2.2 Authorship and Ownership of Works Generated by AI
Further evidence of how AI is used to generate works is needed before questions of authorship and ownership can be properly addressed. Changes to these legal concepts should be approached with caution, to avoid any unintended negative consequences.
Any regime to recognize authorship and ownership of AI-generated works must ensure that appropriate rights, including moral rights, are vouchsafed to the authors and owners of any works used as inputs by AI technology applications. This is important to maintain the ability of owners and authors to benefit from, and be credited for, the creation of human-generated works requiring genuine originality and skill.
2.3 Infringement and Liability Regarding AI
Evidence of how AI technology is used in practice is needed before a legislative framework relating to infringement and liability can be developed. Access Copyright notes, however, that it is far from certain that maintaining the status quo would have a chilling effect on the market for AI-generated works. On the other hand, there is a high likelihood that carving out exceptions to liability for infringement conducted using AI technology would result in negative impacts on copyright owners. As such, further consultation should be conducted before any changes are made in this regard.
3. Internet of Things
Strong provisions related to TPMs are essential to adequately protect creative content. The current TPM regime for creative content works well and assists rightsholders in monetizing and preventing unfair and exploitative misuse of their works. Any amendments to the Copyright Act should not weaken TPM restrictions for creative content.
Access Copyright recognizes the importance of collecting evidence and studying the interplay between AI and copyright. At the same time, however, we urge the government to act expeditiously on the urgent need to reform the Copyright Act to address the significant problems associated with fair dealing, which have negatively impacted the creative and publishing industries since 2012. In particular, government action is needed and long overdue to restore fair compensation to creators and publishers for the uncompensated copying of their works by the education sector. Consistent with Recommendations 18, 19, 20 and 21 of the Shifting Paradigms report, Access Copyright urges the government to amend the Copyright Act so that:
- Fair dealing for education only applies to educational institutions where a work is not commercially available under licence by the owner or a collective;
- Copyright Board tariffs are enforceable by collectives; and
- Statutory damages are available to all collectives.