Abstract
The deployment of automated content recognition technologies by online platforms has become one of the most contested intersections of intellectual property law, fundamental rights, and digital regulation in the European Union. Article 17 of the Digital Single Market Directive (2019/790) requires online content-sharing service providers to obtain authorisations from rightsholders for works uploaded by their users, and to implement measures, including content recognition technologies, to prevent the availability of unauthorised works. This obligation has generated profound tensions between the legitimate interests of copyright holders in preventing infringement and the fundamental rights of platform users, including the right to freedom of expression (Article 11 of the Charter of Fundamental Rights), the right to the protection of personal data (Article 8), and the right to conduct a business (Article 16). This article analyses the legal framework governing content recognition technologies under the DSM Directive, examines the CJEU's landmark decision in Poland v Parliament and Council (Case C-401/19), which upheld the validity of Article 17 whilst imposing significant constraints on its implementation, and evaluates the practical impact of automated content filtering on fair use, parody, quotation, criticism, and other permitted uses. It argues that the current implementation of content recognition systems produces systematic overblocking of legitimate content, disproportionately affecting individual creators, educational users, and political commentators, and proposes reforms to the Article 17 framework that would better protect user rights whilst maintaining effective copyright enforcement.
Keywords: content recognition; Article 17; DSM Directive; copyright; automated filtering; overblocking; fair use; freedom of expression; platform liability; CJEU
1. Introduction
The regulation of user-uploaded content on online platforms represents one of the most complex and politically charged areas of EU intellectual property law. The internet has democratised content creation and distribution, enabling millions of individuals to create, share, and build upon creative works at unprecedented scale. It has simultaneously created an environment in which copyright infringement occurs at a volume and speed that traditional enforcement mechanisms, individual notice-and-takedown requests, cease and desist letters, and court proceedings, are unable to address.
The EU's legislative response to this challenge, Article 17 of the Digital Single Market Directive (2019/790), fundamentally altered the legal framework governing the relationship between online platforms, their users, and copyright holders. By requiring online content-sharing service providers (OCSSPs) to obtain authorisations from rightsholders or, failing that, to demonstrate that they have made best efforts to prevent the availability of unauthorised works, Article 17 effectively requires the deployment of automated content recognition technologies, colloquially known as "upload filters", that scan user uploads for potential copyright-infringing material and block or remove infringing content before or shortly after it is made available to the public.
The deployment of these technologies raises fundamental questions about the compatibility of automated content filtering with the rights and freedoms of platform users. This article examines those questions through the lens of EU intellectual property law, fundamental rights, and the CJEU's jurisprudence. Section 2 analyses the legal framework. Section 3 examines content recognition technologies and their limitations. Section 4 analyses the CJEU's decision in Poland v Parliament and Council. Section 5 examines the impact on user rights. Section 6 proposes reforms. Section 7 concludes.
2. The Legal Framework
2.1 Article 17 of the DSM Directive
Article 17 of the DSM Directive establishes a liability regime for online content-sharing service providers that departs fundamentally from the "safe harbour" model that previously governed platform liability under the E-Commerce Directive (2000/31/EC). Under Article 17(1), an OCSSP performs an act of communication to the public or an act of making available to the public when it gives the public access to copyright-protected works or other protected subject matter uploaded by its users. The OCSSP must therefore obtain an authorisation from the rightsholder, for instance by concluding a licensing agreement, in order to communicate or make available to the public the works or other subject matter.
Where no authorisation has been obtained, Article 17(4) provides that the OCSSP shall not be liable for unauthorised acts of communication to the public if it demonstrates that it has: (a) made best efforts to obtain an authorisation; (b) made, in accordance with high industry standards of professional diligence, best efforts to ensure the unavailability of specific works and other subject matter for which the rightsholders have provided the service provider with the relevant and necessary information; and (c) acted expeditiously, upon receiving a sufficiently substantiated notice from the rightsholders, to disable access to, or to remove from, its website the notified works or other subject matter, and made best efforts to prevent their future uploads.
2.2 The Safeguards for User Rights
Article 17(7) provides that the cooperation between OCSSPs and rightsholders shall not result in the prevention of the availability of works or other subject matter uploaded by users which do not infringe copyright, including where such works or other subject matter are covered by an exception or limitation. Article 17(9) requires Member States to provide that OCSSPs put in place an effective and expeditious complaint and redress mechanism that is available to users in the event of disputes over the disabling of access to, or the removal of, works or other subject matter uploaded by them.
These safeguards are intended to ensure that the deployment of content recognition technologies does not result in the systematic suppression of legitimate user expression. However, the practical effectiveness of these safeguards depends critically on the ability of automated systems to distinguish between infringing and non-infringing uses, a challenge that, as the following section demonstrates, current technologies are not well-equipped to address.
2.3 The EU Copyright Exceptions
The DSM Directive requires Member States to ensure that users in each Member State are able to rely on the exceptions or limitations for quotation, criticism, review, caricature, parody, and pastiche (Article 17(7)). These exceptions and limitations, derived from Article 5 of the Information Society Directive (2001/29/EC) and supplemented by Article 17's own provisions, reflect the fundamental rights considerations that underpin the permitted uses of copyright works: freedom of expression, the right to education, and the public interest in the free circulation of ideas.
3. Content Recognition Technologies and Their Limitations
3.1 How Content Recognition Works
Content recognition systems operate by comparing uploaded content against a reference database of copyrighted works provided by rightsholders. The most widely deployed system is YouTube's Content ID, which uses digital fingerprinting to identify audio and video content that matches reference files submitted by copyright owners. When a match is identified, the system can apply one of several automated actions: block the upload entirely, monetise the upload by displaying advertisements and directing the revenue to the rightsholder, or track the upload for informational purposes.
Other content recognition technologies include audio fingerprinting systems (used by music platforms), image recognition systems (used by stock photography agencies and visual content platforms), and text comparison systems (used by academic plagiarism detection services and publishing platforms).
3.2 The Overblocking Problem
The fundamental limitation of content recognition technologies is their inability to assess the legal context of a match. A content recognition system can determine that a user's upload contains audio or visual material that is identical or substantially similar to a reference file. It cannot determine whether the user's use of that material falls within a copyright exception or limitation: whether it constitutes quotation, criticism, review, parody, pastiche, or educational use.
The assessment of whether a particular use falls within an exception or limitation requires a contextual, qualitative judgment that considers the purpose and character of the use, the nature of the copyrighted work, the amount and substantiality of the portion used, and the effect of the use on the potential market for the original work. This assessment is inherently fact-specific and requires the exercise of human judgment. Automated systems, which operate on the basis of pattern matching rather than contextual analysis, are structurally incapable of performing this assessment with the accuracy that the legal framework demands.
The result is systematic overblocking: the automated removal or blocking of user uploads that do not infringe copyright. Studies have documented the scale of this problem. An analysis of YouTube's Content ID system found that a significant proportion of disputed content claims are resolved in favour of the uploader, indicating that the initial automated decision was incorrect. The Communia Association has documented numerous cases of legitimate educational, journalistic, and parodic content being blocked by automated systems, including cases in which content clearly falling within a copyright exception was removed from platforms without human review.
3.3 The Asymmetry of Enforcement
Content recognition systems create an enforcement asymmetry that favours rightsholders over users. Rightsholders submit reference files to the content recognition database; the system then automatically blocks or monetises matching uploads. The user whose content is blocked must initiate a complaint through the platform's dispute resolution mechanism, which may take days, weeks, or months to resolve. During this period, the user's content remains unavailable to the public, and the user may lose the commercial opportunity, audience engagement, or public discourse participation that timely publication would have provided.
This asymmetry is particularly acute for individual creators, small businesses, and educational institutions, which lack the resources to navigate dispute resolution processes and may be deterred from challenging automated decisions by the complexity and delay of the process. The practical effect is a chilling effect on legitimate expression: users learn to avoid content that might trigger a content recognition match, even where their use would be lawful under a copyright exception.
4. The CJEU's Decision in Poland v Parliament and Council
4.1 The Challenge
Poland brought an action for annulment of Article 17(4)(b) and Article 17(4)(c) in fine of the DSM Directive, arguing that these provisions required OCSSPs to carry out prior automated filtering of content uploaded by users, in violation of the right to freedom of expression and information guaranteed by Article 11 of the Charter of Fundamental Rights.
4.2 The Court's Ruling
The CJEU, sitting as the Grand Chamber, dismissed Poland's action but imposed significant constraints on the interpretation and implementation of Article 17. The Court held that Article 17 does entail a limitation of the right to freedom of expression and information, but that this limitation is justified, proportionate, and accompanied by sufficient safeguards, provided that the measures adopted by OCSSPs do not result in the blocking of lawful content.
The Court emphasised that Article 17(7) and (9) require that the specific content recognition tools adopted by OCSSPs must not have the effect of preventing the availability of content which does not infringe copyright, and that users must have access to an effective complaint and redress mechanism. The Court held that Member States, when transposing Article 17, must ensure that the measures are implemented in a manner that does not lead to preventive filtering of all or almost all content, and that sufficient safeguards are in place to protect users' fundamental rights.
4.3 Implications
The Poland v Parliament and Council decision establishes that Article 17 is valid but that its implementation is subject to fundamental rights constraints. OCSSPs may deploy content recognition technologies, but those technologies must be calibrated to avoid the blocking of lawful content. Where automated systems are unable to make this distinction, the system must err on the side of availability, and the user must have access to effective redress.
5. Impact on User Rights
5.1 Freedom of Expression
The deployment of content recognition technologies has a direct impact on the right to freedom of expression guaranteed by Article 11 of the Charter and Article 10 of the ECHR. The automated blocking of user content before or shortly after it is made available to the public constitutes a prior restraint on expression, a form of interference that fundamental rights law subjects to particularly strict scrutiny.
5.2 The Right to Quotation, Parody, and Criticism
The copyright exceptions for quotation, criticism, parody, and pastiche serve essential functions in democratic discourse. Quotation enables academic scholarship and public debate. Criticism enables artistic and political commentary. Parody enables satirical commentary on cultural products and public figures. These uses are protected by both EU copyright law and the fundamental right to freedom of expression. Content recognition systems that cannot distinguish between infringing reproduction and lawful quotation, criticism, or parody suppress these protected uses, undermining both copyright law and fundamental rights.
5.3 Educational Use
The DSM Directive introduced a mandatory exception for the use of works and other subject matter for the purpose of illustration for teaching (Article 5). Content recognition systems that block educational content, such as lecture recordings that incorporate excerpts from copyrighted works for the purpose of commentary and analysis, undermine this exception and may have a chilling effect on the adoption of digital teaching methods.
5.4 Human Rights in Platform Economies
The concentration of content distribution in a small number of dominant platforms (YouTube, Facebook, Instagram, TikTok) means that the content moderation decisions of these platforms have a disproportionate impact on public discourse. A user whose content is blocked on YouTube may have no effective alternative channel for reaching a comparable audience. The automated enforcement of copyright by platform algorithms therefore has implications for freedom of expression that extend beyond the bilateral relationship between the rightsholder and the user, affecting the broader public interest in access to information and cultural participation.
6. Proposals for Reform
6.1 Mandatory Human Review for Disputed Content
Article 17(9) requires effective complaint and redress mechanisms, but it does not mandate human review of disputed content decisions. The article proposes that all disputed content decisions should be subject to mandatory human review within a specified timeframe (for example, 48 hours), and that content should remain available during the review period unless the rightsholder demonstrates that immediate removal is necessary to prevent irreparable harm.
6.2 Pre-Filtering Exceptions
Content recognition systems should be programmed to recognise and exempt certain categories of use that are likely to fall within copyright exceptions: short quotations, parodies, educational uses, and news reporting. This "pre-filtering exception" would reduce overblocking by creating a safe harbour for uses that are presumptively lawful.
6.3 Transparency Reporting
OCSSPs should be required to publish annual transparency reports detailing the volume of content flagged, blocked, and disputed through their content recognition systems, the proportion of disputed decisions resolved in favour of the user, and the average time to resolution. Transparency reporting would enable regulators, researchers, and the public to assess the effectiveness and proportionality of content recognition systems.
6.4 Rightsholder Accountability for Overclaiming
Rightsholders who systematically submit overbroad reference files or who repeatedly claim content that falls within copyright exceptions should be subject to sanctions, including the temporary suspension of their access to content recognition systems. This accountability mechanism would counteract the incentive for rightsholders to maximise the scope of their claims at the expense of user rights.
7. Conclusion
Content recognition technologies are a necessary component of copyright enforcement in the digital age. The volume of user-uploaded content on online platforms makes manual review impracticable, and the legitimate interests of copyright holders in preventing mass infringement cannot be dismissed. However, the current implementation of content recognition systems under Article 17 of the DSM Directive produces systematic overblocking of legitimate content, disproportionately affecting individual creators, educational users, and political commentators. The CJEU's decision in Poland v Parliament and Council establishes that Article 17 is valid but that its implementation must respect fundamental rights, including the right to freedom of expression and the rights of users to rely on copyright exceptions. The reforms proposed in this article, mandatory human review, pre-filtering exceptions, transparency reporting, and rightsholder accountability, would strengthen user rights protections whilst maintaining effective copyright enforcement in the platform economy.
Bibliography
Angelopoulos C, 'On Online Platforms and the Commission's New Proposal for a Directive on Copyright in the Digital Single Market' (2017) CIPIL Working Paper
Balkin J, 'Free Speech Is a Triangle' (2018) 118 Columbia Law Review 2011
Communia Association, Implementation of the DSM Directive: User Rights Under Attack (2022)
Crémer J, de Montjoye YA and Schweitzer H, Competition Policy for the Digital Era (European Commission 2019)
Frosio G, 'Reforming Intermediary Liability in the Platform Economy' (2017) 112 Northwestern University Law Review Online 18
Husovec M, Injunctions Against Intermediaries in the European Union (CUP 2017)
Back to contents ↑ 
