/ Policy

Article 13: Against European SMEs

The EU Commission prides itself with actively supporting the growth of SMEs and startups and paving the way to a prosperous Digital Single Market. Article 13 of the proposed Copyright Reform runs counter these declared goals. In fact, it identifies a wrong problem in a false narrative of the value gap and proposes a wrong solution in the shape of intermediary liability and automated content filters. The result is a destabilized IT business environment, with SMEs and Startups suffering the greatest economic, legal and technical burdens.

Article 13 of the Commission’s proposal for a new Copyright Directive makes online platforms directly responsible for the copyright infringements of their users, unless platforms implement automated content filters. In legal terms, the only way online platforms can avoid the chilling intermediary liability is by taking “effective measures” to prevent the availability of copyrighted material, remove it "expeditiously" when flagged and ensure future availability of such content is blocked. While these requirements are not directly called “automated content filters” (as it would be illegal to impose a mandatory general monitoring obligation), in practice such strict measures can only translate into automated content filters. To reinforce the hint, Commission’s Vice-President Ansip himself advertised Audible Magic (automated content filtering tool) as a compliance solution.

In order to understand the effects of Article 13 on the EU’s ICT SMEs and startups, it is important to ensure some background in what the “value gap” narrative is, how does it connect to the proposed Copyright Reform and what actors outside the “value gap” narrative would be affected. If you’re in a rush and want to skip the background section, jump to part 4.

1. The Value Gap for music creators

The ‘value gap’ has been defined by the IFPI as the “dramatic contrast between the proportionate revenues generated by user upload services and by paid subscription tiers of services such as Spotify, Deezer and Rhapsody-Napster.” While streaming services produced $2bn from their 68 million subscribers, with an average revenue per user (ARPU) of $29.41, ad-supported services only produced the ARPU of an $634m from their 900 million users, with an ARPU of $0.70. While a multifaceted problem, one of the most commonly used explanation for the emergence of this difference is YouTube's dominant position in user upload services. Indeed very likely, the explanation is based on the poor contractual agreements for music creators. In essence, music creators face two bad options in their licensing negotiations with YouTube: locking their work outside the public’s reach on the dominant platform or making their work available on the platform and receive a payment that is regarded as unsatisfactory. At the end of the day, artists need public attention before revenue. For this reason, they tend to opt for the latter when given an (unfair) choice between the two.

2. Article 13: addressing the value gap through copyright law

IFPI and other copyright holders have blamed the value gap on the liability exemptions in Article 14 of the e-Commerce Directive. Without a licensing agreement with copyright holders, user uploads of music videos would generally be infringing unless one of the limited exceptions in copyright law applies (e.g. parody, education). Article 14 protects platforms (intermediaries) against liability, should their users infringe on copyright. The legal protection requires that the platform is not aware of the infringing activity and when it is made aware, it expeditiously responds to notifications of infringement by disabling access to illegal content.

The liability exemption may give YouTube an advantage in the licensing negotiations with music creators, but this is not necessarily the case. YouTube’s dominant position among user upload services cannot be explained by the liability exemption since the entire industry benefits from this exemption. YouTube’s dominant position is therefore a matter for competition law, not copyright law.

Moreover, it is also worth recalling that the liability exemption in the e-Commerce Directive is a carefully crafted balance between the interests of rights holders and internet intermediaries. If information society services with user-provided content had primary liability for any infringing activity, very few services would exist on the internet.

Despite the obviously complex nature of the value gap problem, Article 13 tries to address the issue solely through copyright law with a carrot and stick approach. The stick is the deliberate weakening of liability exemptions in Article 14 the e-Commerce Directive, without actually amending the Directive. The carrot is that platforms can avoid, or at least reduce, this liability by installing upload filters with the purpose of preventing infringing user uploads.

3. YouTube and beyond

The proposed Copyright Reform is aimed at YouTube for the company’s’ role in the abovementioned “value gap” issue. In other words, YouTube would have to implement automated content filters and be subjected to intermediary liability. What does this mean for the tech giant and how does it affect the rest of EU IT businesses that allow users to upload content on their platforms?

TL;DR - It doesn’t mean much for YouTube.
According to Article 13, an online platform as YouTube would hold primary liability and be held directly responsible for their users’ copyright infringement, unless they implement the automated content filters. Perhaps it would be worth pointing at YouTube’s (and Google’s, and Alphabet’s) army of lawyers that could survive a long legal battle, should the platform be sued on intermediary liability claims. But well, it’s not worth it. The reason is simple: no legal battle will take place of those claims. YouTube has been implementing automated content filtering (ContentID) since 2009, hence primary liability would be avoided by definition. The system automatically detects uploaded videos that infringe on copyright and allows for their removal when they are flagged, as well as in the future, should infringement repeat.

Very few changes, if any, would be needed for YouTube to comply with Article 13, as the article is arguably based on the current design of content filters like ContentID. The value gap problem would remain unaffected since YouTube is compliant even before the new Directive is adopted.YouTube’s dominant position in the negotiations with music creators will therefore be maintained unharmed.

What the legislators seem to forget is that there are businesses operating with online content other than YouTube. Firstly, content can also be different that audio and video. In fact, text, images and even source code is considered content. As long as a website allows for any type of content to be uploaded by users, it will have to abide by Article 13 if the reform is adopted. Secondly, even if Article 13 is supposed to be limited to platforms where users upload content over which others hold the copyright, the effect of Article 13 will also extend to services that allow users to upload their own content. For example, an online service offering hosting for podcasts will have no way knowing whether users actually upload their own podcasts or somebody else’s music, or whether the user podcasts contain music elements that would require a licence in order not to infringe copyright. The only way that the podcast hosting service can know this is to employ automated content filtering.

Asked about the impact of content filters on non-YouTube actors, EU Commissioner for Market Competition Vestager declared that making “sure part of the value reaches the creator” involves “some costs somewhere”. EDiMA identified the “somewhere” and put together all platforms possibly affected by Article 13.


For YouTube’s competitors and other online services with user uploaded content, Article 13 is a very different story. In best case scenarios, these companies will face additional costs of implementing the same content filters that YouTube already has. In other scenarios, for example when they provide a platform for content types for which no filters are available (such as source code), their entire business model will be threatened.

In the SME and startup scene, several voices pointed at the dangers Article 13 brings for small market players. Allied 4 Startups launched the Copyright 4 Startups campaign, documented the grievings of outspoken startup innovators and pointed at the reform as being a monopoly enforcing tool. Moreover, several MEPs involved in the legislative process have been addressed by an official letter of concern that focuses on Article 13’s negative impact on small businesses.

4. Article 13 meets EU’s IT SMEs and Startups

But wait, what are they all complaining so much about? Now that we have a grasp of the context in which the proposal for a Copyright Reform emerged, here’s a summary of what IT SMEs and startups could face.

Legal uncertainty and costs

The proposed Copyright Reform effectively undermines the foundation of the e-Commerce Directive that many IT SMEs and startups are relying on in conducting their business.

For the past fifteen years, online platforms in Europe have been operating within a legal framework that shields them from primary liability of the content hosted, transmitted or cached. We have explained this protection under Article 14 of the e-Commerce Directive in Section 2. Another notable provision of the e-Commerce Directive is Article 15, an article that echoes European Court of Justice‘s past rulings: it is illegal for Member States to impose general monitoring obligations on internet intermediaries (online platforms) in regards to their users’ content.

Article 13 of the proposal introduces primary (direct) liability for intermediaries (against the spirit of Article 14 of the e-Commerce Directive) and offers general content monitoring as a way to avoid this liability (against Article 15 of the e-Commerce Directive and ECJ rulings). These two conflicting aspects create a undeniable legal uncertainty for online services that allow users to upload any type of content on their platforms (audio, video, text, code). For small market players, this legal uncertainty is particularly dangerous. Here’s why.

The legal risk outlined above may seem unlikely to materialise in legal threats (lawsuit) while the company is small and does not produce considerable profits. For this reason, entrepreneurs might be tempted to ignore it and continue with doing business-as-usual. But this is a dangerous strategy. As the company grows and becomes more profitable, copyright holders (or specialised copyright collection companies acting on behalf of copyright holders) can secretly collect evidence on uploads by users that infringe on copyright, and sue when prepared.

European IT SMEs/startups risk a situation in which legal uncertainty would morph into legal action as a small business grows profitable enough to pay for its users’ infringements. Even if we assume legal action would take place earlier in the life of the startup, the company would most likely be pushed out of existence as a result of costs associated with the suit, before the court even has the chance to make a decision. Whatever strategy copyright holders decide to pursue, the end result is greater legal costs for companies offering online services with user uploaded content.

Such a scenario is detrimental to investors too, who already face a large risk due to the nature of startup investments. In this new context, any platform that would be distributing content, regardless it being legal or illegal, would be doomed by definition due to inability to support potential legal costs..

Operational costs

In order to avoid an unquestionable death sentence, most IT SMEs and startups might choose to completely secure themselves against intermediary liability by implementing automated content filters. The problem is, filters cost good money.

If you want to buy it

EU Commissioner Andrus Ansip claims that “400-500 bucks” would be enough for a small company that filters 10.000 songs using the automated content filter Audible Magic. In reality, EU Commission’s own report suggests that “(…) on the basis of the information available, it is estimated that a small scale online service provider with a relatively low number of monthly transactions can obtain such services as from €900 a month.” In the reality outside Brussels, this price is challenged again. In the study “The Limits of filtering: A look at the functionality and shortcomings of content detection tools”, the authors note that the €900 / month figure applies to max 5000 songs. In fact, the study quotes a survey (pg 26) that reported medium-sized companies paid between 10,000 and 25,000 dollars a month in licensing fees alone for Audible Magic’s filtering tool. Conclusion? Cheap filters myth - debunked!

Should a company choose to look for filters alternatives on the market, it will soon notice there is only one other option available: YouTube’s Content ID. Yes, the same YouTube this whole Copyright Reform was trying to punish for provoking the ‘value gap’. Not only are we talking about a monopolised market with 2 players (currently in a lawsuit over, heh, trademark claims), but small ISP could be obliged to empower their greatest competitor - Google.

If you want to create it

Perhaps IT SMEs and startups would like to develop their own filtering solutions. In the end, EU is full of talent. That would not be a problem, should the company have 60 million euros to spare, a figure YouTube invested in developing Content ID. If 60 is too much of a hassle, maybe Soundcloud’s 5 million and 7 active maintenance employees sounds less effort. Either way - few, perhaps zero, European startups and SMEs would be able to even consider these figures for the development of a complex automated content filtering solution.

Technical feasibility

What legislators often seem to forget when trying to impose technical measures is that technology is never fault-free.

Technically, it fails

Technology fails. Especially in its incipient form. This is where automated content filters are right now. MEP Julia Reda documented some of ContentID’s failure, and Daphne Keller of Stanford Law School warned that

Developing monitoring tools that do not over-filter (remove more information than intended) or under-filter (remove less) is a significant, perhaps asymptotic technical challenge. Even YouTube’s Content ID, [...] is routinely faulted by Internet users for misidentifying and removing lawful content.

On top of the above, one should remember different filters are needed for different types of content. If a platform allows its users to comment on hosted videos, there are automatically at least 3 filters required: one for video, one for audio and one for text. In The Limits of Filtering, Engstrom and Feamster point at the technical limitations of existent automated content filters:

all content filtering technologies are at best capable of simply identifying the contents of a file, not making the often complex determination as to whether the use of a particular file constitutes an infringement. Furthermore, no filtering technology can process encrypted files.

...and then there was code.

Most of EU’s IT Startups and SMEs use software repositories as Github to store proprietary code, as well as share and build on knowledge through open source code. Article 13 would force Github to implement automated content filters and endanger the backbone of EU’s IT landscape.

The technical issues that filters are often associated with become a lot greater in the case of platforms hosting code. On one hand, the risks of misidentifying copyrighted material increases and on the other, the damage done by such misidentification is considerately more dangerous.

A research paper co-authored by Free Software Foundation Europe and Open Forum Europe points that the only way Github could avoid primary liability for their users’ copyright infringements would be through conducting permanent software compliance assessments. Several issues arise from this.

Firstly, software compliance assessments involve proprietary code licenses, as well as open source code licenses. Checking whether a piece of software aligns with the license it is meant to is therefore a human intensive process that cannot be supported by Github’s team at the scale content circulates on the platform. Secondly, should Github be constrained to conduct these assessments anyway, the only realistic way to do it would be through automated content filters. At the moment, no such filters exist for monitoring code content and should any be developed, a functional level would only be reached in years. Combining the intrinsic need for human judgement in software compliance assessment, the high failure rate of existing automated content filters and the fact that no such filters exist for code lead to a sumbre conclusion: Should Github be forced to implement automated content filters for code that are prone to failure, entire repositories could crash out of nowhere because the filtering algorithm “mis-judged”. In the words of FSFE and OPF, Article 13 could

cause developers’ dependencies randomly to disappear and so literally “break” their builds, resulting in lost business, lost productivity, less reliable software, and less resilient infrastructure.

Thirdly, as automated filters can not scan through encrypted content, Github would need to gain access somehow to such encrypted content to ensure no licensing infringement take place. Such a request would mean an entire reconsideration of Github’s contractual agreements, a deterioration of (especially European) customer relationships and eventually, the platform’s withdrawal from the European market.


Article 13 of the proposed Copyright Reform could cause significant harm to European SMEs and startups. From the basis it has been founded on, to actors it impacts and its relationship with existing laws, this provision’s negative effects are of legal, economical and technical nature.

First of all, Article 13 creates legal uncertainty due to its conflicting relationship with Article 14 and Article 15 of the eCommerce Directive. This uncertainty can be strategically used by the copyright industry against new online platforms by either suing the company while a small player, either by building up evidence throughout time. In both cases, the online platform would go bankrupt. Secondly, avoiding primary legal liability through automated content filters might not be economically possible for most SMEs. Buying a compliance solution is limited between Audible Magic and Google’s Content ID and can cost the medium sized company between 10,000 and 25,000 dollars a month in licensing fees alone. Developing an internal automated content filter involves millions of dollar, a figure European SMEs and startups would not afford. Finally, automated content filters are technically fallible. Existing ones can not distinguish between copyright infringement and copyright exceptions and, as a result, they over/under filter content. Platforms hosting code (such as Github) will have to choose between conducting continuous software compliance assessments or introducing content filters for code (not invented yet). Business hosting their code on such platforms risks having their developers’ work blocked due to algorithmic failure, resulting in lost business, lost productivity, less reliable software, and less resilient infrastructure. Moreover, hosting platforms obliged to validate the legality of hosted content would need access to their clients’ encrypted content. Code hosting platforms would initially reconsider their contractual agreements with European customers and be obliged to eventually exist the European market.

The scenario painted above makes capital investment even a riskier business than it already is. In addition to the inevitable financial risk of investing in startups, investors will also have to consider the risk that a large part of the profits from a successful investment will end of in the hands of copyright holders. Overall, online businesses that allow users to submit audio, video, image, text or code content on their platform will not be able to survive on the European market.

There is still time to stop Article 13 from destroying EU's small ICT businesses and startups. The JURI Committee will vote on this matter at the end of March. The Committee members need to hear YOU!

  • Contact your representatives and let them know how terrible Article 13 is for EU's small businesses.

  • Is your business directly impacted? Write us an email.

This article has been written in collaboration with IT - Politisk Forening