Cropped etching of a wedding scene: Front left two musicians playing the shawm. A bridal couple and several young couples in contemporary and old-fashioned clothing dance to the music. On the right, a man stands against a tree, viewing the scene through his pince-nez. A church tower in the background.

Article 17 – five years later

This article was first published on Kluwer Copyright Blog on June 7, 2024.

Today marks the fifth anniversary of the entry into force of the Directive on Copyright in the Digital Single Market. It is hard to remember how divisive and controversial the Directive was during its creation. The Directive’s most controversial provision – Article 17 – which brought hundreds of thousands of people onto the streets and millions of voices online to warn of the potential of upload filters to kill the Internet, seems to have faded almost completely from public view. Instead, attention has shifted to the TDM exceptions in Articles 3 and 4 of the Directive – so much so that some people have begun to refer to the Directive as the “TDM Directive,” – which received relatively little public attention at the time.

In this situation, it seems appropriate to take another look at Article 17 of the Directive and see what we have learned in the last five years and how the provisions have worked in practice.

Overblocking is real a fairly marginal outcome

First and foremost, it is pretty clear that the upload filters made mandatory by Article 17 (but which had been around for much longer) did not, in fact, kill the Internet.

Looking back over the past five years, it seems clear that the adoption and implementation of Article 17 has not led to more widespread use of automated content detection systems by social media platforms, and that while there continues to be anecdotal evidence of overblocking, it is not a very widespread problem. Most of the evidence for this finding comes from YouTube’s copyright transparency reports, which YouTube began publishing in 2021. And while the first of these reports led me to argue on this blog that “overblocking is real“, the picture has since become much more nuanced.

Starting with its second transparency report, YouTube has pointed out that more than 90% of all ContentID claims are actually related to disputes over monetization (i.e., who gets to claim the revenue generated by the ads displayed around the video), which means that the vast majority of ContentID claims are not related to blocking or removing uploads. There is further evidence from YouTube’s transparency reporting that blocking claims are relatively rare. In September 2022 – likely in response to its obligations under some of the more user rights-focused implementations of Article 17, such as the German one – YouTube introduced an expedited appeals option that is only available for blocking claims (and thus not for takedown claims). In the second half of 2023, 94,343 such appeals were filed, representing just 0.01% of the 1.02 billion(!) ContentID claims processed in the same period. Taking into account that about 45% of such appeals fail, the total number of instances of unjustified automated blocking is likely to be around 50,000.

While not a small number, it is hard to argue that a system that appears to have a false positive rate of 0.005% is fundamentally flawed. This impression is underscored by another figure from YouTube’s transparency reports. According to the reports, automated claims are challenged at about half the rate of manual claims.

While all of the above is based on the somewhat selective transparency reporting of a single platform – but let’s be honest here, the discussion about Article 17 has always been a discussion about YouTube first and foremost – it makes clear that many of the negative effects that the opponents of Article 17 predicted have simply not materialized.

Overblocking is a real problem, but an extremely marginal one. In hindsight, it seems clear that disputes over monetization, rather than blocking, have a much greater impact on users of social media platforms.

So was it worth it?

So if the predicted negative impacts of Article 17 on freedom of expression and other user rights have not materialized, does that mean that the whole fight over the provision was in vain? Of course not.

First of all, the mobilization against upload filters resulted in a final version of the provision that is far superior from a user rights perspective to the Commission’s original proposal and to versions that proponents of the measure would have liked to see. In order to overcome the opposition to Article 17, the co-legislators successively added more and more user rights protections to the article, with the ultimate effect that Article 17 has strengthened the rights of users in the EU.

As COMMUNIA has documented in its analysis of the implementation of the Directive, Article 17 has led to a dramatic increase in the number of Member States that have fully implemented the copyright exception for purposes of caricature, parody or pastiche in Article 5(3)(k) of the InfoSoc Directive. Prior to the adoption of the CDSM Directive, only 9 member states had fully implemented this exception, while today only 3 of the 26 member states that have implemented the Directive have not fully implemented the caricature, parody or pastiche exception.

This outcome is notable because further harmonization of users’ rights (which was a core demand of many civil society organizations for copyright reform) never gained significant support among legislators as a stand-alone demand. Instead, it was achieved through the legislative back door, as a concession to get Article 17 passed in the face of fierce opposition from the same civil society organizations.

Article 17 contains a number of other improvements to user rights. The most notable of these is the introduction of a separate obligation not to prevent the availability of lawful content in Article 17(7). While initially criticized for creating internal contradictions that were difficult to resolve, it was the Polish challenge to the legality of Article 17, combined with strong civil society efforts to push for meaningful user rights safeguards during the stakeholder dialogue, that led the European Commission, and subsequently the CJEU, to clarify that the obligation not to block legitimate content (Article 17(7)) as an outcome obligation overrides the blocking requirements (Article 17(4)) at the heart of Article 17, which are mere best efforts obligations.

Finally there is also reason to believe that the controversy surrounding Article 17 led to the strong emphasis on freedom of expression and procedural rights in the DSA notice and takedown process (instead of a notice and staydown approach).

But what about the value gap?

All of this leaves us with one major unanswered question: Was the whole fight over Article 17 worth it for its proponents? Did Article 17 serve its intended purpose of closing the alleged “value gap” by redirecting more of the revenue generated by user-generated content platforms to authors, performers and other rights holders? Unfortunately, this is a question that cannot be answered from the outside. The fact that neither the platforms nor the organizations representing rightholders have publicly commented on this question suggests that, despite all the efforts and hopes invested in the fight for Article 17, its impact on improving the income situation of authors and performers has probably been limited at best.

Here it would be interesting to see a more thorough evaluation based on real data from the Commission, for which we will have to wait at least until June 7, 2026.


A satire on the art business in which art experts and dealers who assess paintings are depicted as donkeys. After the drawing by Trémolières in the Hessisches Landes Museum in Darmstadt (cropped).
Featured Blog post:
The Post-DSM Copyright Report: The Meme Supplement
Read more
Older post
The Post-DSM Copyright Report: The Meme Supplement
June 7, 2024