by Mitchell Longan[1]
Introduction
The European Parliament is currently considering a proposal to address problems of piracy and other forms of copyright infringement associated with the digital world.[2] Article 13 of the proposed Directive on Copyright in the Digital Single Market would require information society service providers to adopt filtering technology capable of preventing users from posting or sharing copyrighted materials.[3] While the European Parliament has correctly identified a problem — a failure in the law to adequately address online infringements, especially those of low economic value — the proposed solution would represent a regression in legal evolution. The digital world has revolutionised the way creative works are created and disseminated. That revolution has likewise extended to perceptions as to how copyright law should function. As copyright law in the European Union has gone unchanged since 2001, the content industries have set in place sophisticated pseudo-legal systems designed to combat these issues of unauthorised uses. However, these systems are generally focused on allowing low-value uses while establishing mechanisms to monetise them — both for the users that publish them and the copyright holders themselves. Article 13 undermines these pseudo-legal systems that are functioning successfully on the internet. Moreover, Article 13 stands to create more problems than it solves as filtering technology is currently incapable of addressing many fundamental aspects of copyright law such as exceptions to infringement and legitimately-licensed materials. Forcing internet portals to adopt copyright filtering software is a wrong solution not only because the available technology is incapable of adequately determining copyright violations but also because current industry developments show that monetisation, not take-downs or blocking, should be the focus of legislators.
Inadequacies of the Technology
Legislating a requirement that internet information society providers adopt filtering technology to prevent copyright infringements by their users relies on the existence of technology capable of doing so. While digital fingerprinting technology that can recognise copyrighted works does exist, copyright law is both complex and, at times, indefinite. An analysis to determine infringement requires far more than simply recognition. Determining infringement often requires subsequent analyses relating to statutory exceptions or fair dealing for courts in the United Kingdom. Moreover, copyrighted works are licensed to users on a rather regular basis. This can occur through traditional channels involving contracts or, more often, through tolerated use initiatives by content owners in the form of no action policies. Current examples of filtering technology are not only incapable of determining if a publication would qualify for an exception to infringement, but are also unable to recognise if the work has been properly licensed or if the use is tolerated by the content owner. Furthermore, claims are even, at times, issued via filtering software by those who are not the rightful owners of a copyright. These failings to distinguish the intricacies of copyright law become especially problematic when considering that the current leading filtering software still fails to simply identify 20-40% of music recordings within its database.[4]
False Positives and Exceptions to Infringement
While the technology that powers existing content identification software is undeniably complex and sophisticated, the resulting capabilities of that technology remain limited. Content identification software systems rely on compiling massive databases of content and using fingerprinting technology to mark that content in order to determine whether a piece of uploaded content matches in whole or in part any piece of content in their database. This process is performed solely under a match or no-match condition. In other words, the sole objective, and only capability of the tech, is to determine whether or not the scanned content matches a piece of content from the database.
The first step of an infringement analysis will invariably involve a determination of similarity and access — regardless of jurisdiction. In theory, content identification software will correctly identify direct copying which would satisfy the required analysis for both of these factors. However, to reduce the process of analysing infringement solely to these two steps is, at best, unfair and, more likely, illegal because infringement analyses typically require far more considerations than simply access and similarity. All jurisdictions in Europe provide exceptions to infringement whereby users may legally make use of protected works without permission. The simple match/no-match process of content identification software fails completely to address this more complex aspect of copyright law and works that should be treated as a legal use without permission are regularly flagged as infringements.[5]
Exceptions to infringement are not only foundational aspects of copyright law, but also represent uses of copyrighted works that are not illegal and therefore should not blocked from publication or monitisation for a third party by any entity — software or otherwise. To allow such a system to exist is unjust but to require its adoption would be outrageous.
Licensed and Tolerated Uses
The inability to perform complex analyses regarding exceptions to infringement is not the only failure seen by content filtering software systems. Current filtering software will also not take into consideration whether a work has been licensed through legitimate channels or falls under a broader category of tolerated use.
With regards to licensed material, YouTube’s Content ID software has blocked content containing legally obtained stock audio with such regularity that companies are offering specific advice for how their customers may dispute these claims.[6]
Video game companies are at the forefront of tolerated use policies, with a particularly active and engaged fan base that bolsters the value of their copyrighted materials through various digital fan works such as Lets Play videos. These videos are tolerated by most game development companies with specific policies listed on each company’s website. Ubisoft, a game company that has supported user engagement in the form of fan videos and otherwise, spoke out in support of its fans who were issued Content ID claims for videos containing permitted footage from its games.[7]
Taking down or monetising these legitimate forms of creation establishes content filtering systems as indiscriminate tools of usurpers. Furthermore, such a system would create deterrence to illegitimate uses of copyrighted material as much as to legitimate uses. Being aware that that the content will be flagged up regardless of the legal status, the incentives to acquire license legally would be weakened.
Third Party Claims
Not only do these systems have difficulty ascertaining what constitutes material that should and should not be taken down, but they also struggle to appropriately determine who has the right to make a claim on the copyrighted material. YouTube’s Content ID system is rife with reports of automatic claims being issued by entities that should not have the power to issue a copyright claim.
Among these examples include a claim made for a song in the public domain;[8] a magazine claiming ownership of a screenshot from a video game on the basis that it was published in one of their editions;[9] and Universal Music Group licensing a song from an indie artist to use as a backing track in an audiobook and then using that audiobook as a proxy to claim ownership for the original song.[10]
Current Industry Demands
As legislative reform concerning copyright in the digital era has been stagnant, content industries have been left to find their own solutions to infringements online. An analysis of the evolution of tolerated use and similar adopted initiatives illustrates a growing favour for pecuniary rights over exclusivity among content owners.[11] It is difficult to tell whether this emerging trend is a result of the impracticality of preventing users from publishing unauthorised uses of copyrighted works online or simply content owners preferring to establish new revenue streams from copyrights. Yet, legislators have a responsibility to acknowledge this change in perception regarding copyright law. Moreover, while legislation is needed to address the issues associated with unauthorised uses online, any legislation passed should expand on the industry norms rather than invalidate them. Here that means finding ways to help content owners monetise low-value infringements efficiently — not preventing users from sharing the content.
Conclusion
The current available software is not only inadequate in addressing intricate aspects of copyright law but it regularly fails to simply identify copyrighted material in its database. As a result, the technology is far from sufficient to merit a compulsory adoption. Furthermore, such a regulatory approach would demonstrate absolute ignorance to the changing tides of perception towards copyright law and its function online from both users and content owners alike.
[1] University of Sussex.
[2] Commission, “Proposal for a Directive of the European Parliament and the Council on copyright in the Digital Single Market”, COM (2016) 593 final.
[3] Ibid., art. 13.
[4] Tim Ingham, “‘YouTube’s Content ID Fails to Spot 20%-40% of Music Recordings’” (Music Business Worldwide, 13 July 2016), available at https://www.musicbusinessworldwide.com/youtubes-content-id-fails-spot-20-40/ (accessed 25 July 2018).
[5] Tim Cushing, “YouTube Kills Livestream of Convention When Audience Starts Singing ‘Happy Birthday’” (TechDirt 15 October 2013), available at https://www.techdirt.com/articles/20131014/15323524876/youtube-kills-livestream-convention-when-audience-starts-singing-happy-birthday.shtml (accessed 25 July 2018). See Also Paul Resnikoff, “Sony Music Says a Harvard Copyright Lecture Violates Its Copyrights…” (Digital Music News, 25 April 2017) available at https://www.digitalmusicnews.com/2016/02/17/sony-music-takes-down-a-youtube-lecture-about-copyright-issues/ (accessed 25 July 25 2018). (Where a Harvard law school copyright lecture was taken down for including parts of Hendrix’s “Little Wing” despite being a textbook example of fair use.) See also Gita Jackson, “Nintendo Flags YouTuber For Using Switch Sound” (Kotaku Australia, 8 March 2017), available at https://www.kotaku.com.au/2017/03/nintendo-flags-youtuber-for-using-switch-sound/ (accessed 25 July 2018). (Where a Content ID claim was made for a “Ready… Fire!” sound effect from the Nintendo Switch.)
[6] Jordan McElwain, “Solving YouTube Content ID Claims for Stock Audio” (Storyblocks Blog, 13 June 2018), available at https://blog.storyblocks.com/storyblocks-features/youtube-content-id-claims-stock-audio/ (accessed 25 July 2018).
[7] Stephen Totilo, “YouTube Copyright Chaos Continues. Game Publishers to the Rescue?” (Kotaku, 12 December 2013), available at https://kotaku.com/youtube-copyright-chaos-continues-game-publishers-to-t-1481517758 (accessed 25 July 2018).
[8] “YouTube’s Content ID System Brings Humbug to the Holidays” (Adam the Alien Productions, 7 January 2014), available at http://adamthealien.com/2013/12/23/youtubes-content-id-system-brings-humbug-to-the-holidays (accessed July 25, 2018).
[9] Totilo, supra n. 7.
[10] “UMG Licenses Indie Artist’s Track, Then Uses Content ID to Claim Ownership of It” (Techdirt, 18 March 2015), available at https://www.techdirt.com/articles/20150317/10513630347/umg-licenses-indie-artists-track-then-uses-content-id-to-claim-ownership-it.shtml (accessed 25 July 2018).
[11] Youtube, “Creator Academy” (2018), available at https://creatoracademy.youtube.com/page/lesson/cid-optimize?hl=en-GB#strategies-zippy-link-1 (accessed 31 May 2018). (Showing that 90% of all claims issued by YouTube’s Content ID software result in monetisation, not takedown.)