This past Tuesday, the European Union Parliament approved a heavily amended version of the copyright reform legislation that has been generating much controversy over the past couple of years. We’ve been looking specifically at Article 13 of the proposed Directive, which has generated the most controversy for its inclusion of provisions that would require online services to either take licenses to content or filter it out using automated content recognition methods. The draft approved this week is substantially watered down and eliminates any filtering mandates.
First, the Directive changes the definition of applicable service providers from the very broad “information society service providers” to the much narrower “online content sharing service providers,” which it defines as services that “store and give access to the public or … stream significant amounts of copyright protected content uploaded / made available by its users, and that optimise content, and promote for profit making purposes, including amongst others displaying, tagging, curating, sequencing, the uploaded works or other subject-matter”. It contains exemptions for small businesses, noncommercial service providers (Wikipedia), open-source software development platforms (GitHub), cloud services for individual users (Apple iCloud), educational and scientific repositories (Sci-Hub?), and online physical goods retailers (Amazon?).
Second, the new Article 13 says that “special account shall be taken of … ensuring that … automated blocking of content is avoided.” It simply requires that online service providers sit down with content industry representatives in good-faith negotiations to agree on license terms and best practices for enforcement.
On the other hand, the text stipulates that online content sharing service providers “perform an act of communication to the public” that requires licensing agreements. This is legally significant language that gives copyright owners more legal leverage against the Googles and Facebooks of the world. And even this watered-down version of the law will introduce potential legal risk, cost, and overhead that could make entrepreneurs and investors think twice about building or investing in online services that involve content. That’s not good for innovation.
But the current draft of this legislation does not portend a future of pervasive automated copyright filtering; at least not more than exists today. Nevertheless, the open-Internet crowd is still clinging to this narrative, although EU Pirate Party MP Julia Reda — by far the most outspoken legislator against the bill — has at least modulated her rhetoric from “this law requires filtering” to “[t]his law leaves sites and apps no choice but to” filter.
The law gives online service providers a choice of taking licenses to content or “cooperat[ing with rights holders] in good faith in order to ensure that unauthorised protected works or other subject matter are not available on their services”, with the proviso that such “[c]ooperation … shall not lead to preventing the availability of non-infringing works or other protected subject matter, including those covered by an exception or limitation to copyright.”
This is not going to lead to pervasive filtering. Instead it will most likely lead to service providers choosing between implementing notice-and-takedown style schemes (where they haven’t already) and paying blanket license fees to collecting societies.
First, it’s complex and expensive — if not downright impossible — to build filters that meet the above criteria. In fact, this language could lead some services that already use filtering — like YouTube and SoundCloud — to abandon their current filtering schemes because someone decides that they don’t meet those criteria and that the Directive gives them leverage to sue. Online services are just not going to engage in envelope-pushing R&D programs just for the sake of meeting some vaguely-worded legal criterion to the satisfaction of copyright owners.
Second, service providers know that any service that filters content is always going to be less desirable to users than one with similar features that doesn’t filter; users will always choose the latter if they have a choice. In fact this has already happened. In the mid-2000s, music file-sharing service iMesh implemented a fingerprint-based filtering scheme approved by the music industry — and languished in obscurity while similar services like LimeWire and Grooveshark stayed popular (before their eventual shutdowns from lawsuits). More recently — and closer to home in Europe — the Swiss cyberlocker service RapidShare suffered a rapid decline in users after it started manual removals of copyrighted material in the early 2010s.
In contrast, notice-and-takedown schemes like those implemented in the United States pursuant to the Digital Millennium Copyright Act are straightforward to implement, not requiring R&D-level technology. Copyright owners say that they aren’t effective, but online services will argue that they fit the definition of “cooperation … to ensure that unauthorised protected works or other subject matter are not available on [online] services” in the latest version of Article 13. Many online services already have notice-and-takedown schemes in place, and they know (even if they don’t like) the costs of operating such schemes.
Some online services will decide to take licenses rather than implement enforcement schemes. But what does it even mean to take licenses to “all copyrighted works,” given the mess that licensing is in today? Who would online services negotiate with?
The answer is collecting societies. Collecting societies in much of Europe are pervasive and highly influential — more so than they are in America. They would be happy to step up and represent copyright owners in their accustomed roles as royalty intermediaries. In fact, the newly-added Article 13b states that collecting societies should manage license fees for images reproduced or referenced on online content sharing services.
Just as they do today with private copying levies on consumer devices and blank media, collecting societies are likely to propose blanket license schemes that are based on some coarse metric, such as revenue or data volume, and apportion the resulting royalties to rights holders using their current black-box methodologies. Online services will be happy to pay according to such simple schemes instead of building expensive, complex, and error-prone systems to track which content passed through their systems, when, and how. (And of course there is no authoritative database of such information.)
In other words, Article 13 could make general online service providers subject to the same types of blanket licenses as commercial music services like Spotify and Deezer have with music PROs today. This will certainly cost them money. How much? If the wildly divergent private copying levy schemes throughout the EU are any indication, it’s anyone’s guess at this point.
It’s also possible that some online services that already do operate enforcement schemes will opt to shut them down and pay license fees instead, because it’s cheaper and easier that way. That would effectively shift money from enforcement technologies and staff to collecting societies, and eventually the pockets of copyright owners and creators. But the lack of precision involved in such arrangements would continue to be a problem, especially for long-tail indie creators, who would have to continue to trust collecting societies to pay them a fair share.
These cost questions will certainly inform the negotiations that will take place between online services and content owner representatives if and when this law is enacted. And that’s a process that will drag on for years and years — a point that favors the big tech companies and is surely not lost on them. The law requires the European Commission and Member State governments to facilitate the negotiations once the law comes into force, but it gives no timeline for their conclusion; the tech industry will make sure that it takes as long as possible. And these processes won’t even start until EU Member States implement the Directive in their national laws, a process that typically takes two years by itself.
In other words, it will be a long time before we see any practical result of Article 13 in the European market; but it’s a safe bet that it won’t involve pervasive automated copyright filtering. In the meantime, the text will get another round of revision before another vote next January.