Copyfraud and Overclaiming – What to watch out for when implementing the EU copyright directive
On March 26th the European Parliament voted to pass a draft directive for a reform of EU copyright regulations. On April 15th the Council approved the bill. The EU member states are now required to implement the new directive into their national laws within a span of two years.
Inescapable: Upload filters
According to Article 17 of the new directive (formerly Article 13), platforms will be directly liable to rightsholders when users upload copyrighted content. Under the current safe harbour regulations they are only liable for violations that have been brought to their attention. (This safe harbour rule was introduced in the so-called eCommerce Directive in 2000. Article 14 of that directive has now been rendered inapplicable to platforms by Article 17 of the new directive.)
In order to avoid an incalculable risk of liability, platforms will now have to control all new content the moment it is uploaded. (Those who still claim that Article 17 does not require upload filters are either misinformed or not entirely honest.)
Services that rely on user-generated content have no choice but to use automated filtering systems because the amount of data is too vast to be checked by humans. According to Youtube, 100 hours of video are uploaded every minute to that platform.
Some platforms are already using such filters. The most well-known – or infamous – of these is Youtube’s “Content ID”. This system compares both existing content as well as each new upload against a database of previously deposited reference files. Youtube itself doesn’t provide those files. Rather, certain rightsholders whose content is deemed particularly at risk are given access to Content ID.
First among those rightsholders are, of course, the big players in the music, movie and TV industries. But they also include others, such as major video game publishers or the world football association FIFA. These rightsholders upload their content into the database and set up a so-called policy which regulates what happens to other users’ videos that contain the same content or parts thereof. There are three “escalation levels”: rightsholders can simply track the video and remain informed about the number of views it generates. They can run ads before or during the video and monetise it. Or they can block the video entirely so that it cannot be watched by anyone.
Overblocking and Overclaiming
Since Content ID was introduced there have been complaints about so-called overblocking or overclaiming, meaning that videos are blocked, monetised or simply tracked even though there is no copyright violation.
Cases of overblocking and overclaiming can roughly be sorted into three categories:
1. No match, “false flag”
The system takes effect even though the claimed content doesn’t match the reference file. There have been several instances of Content ID assigning birdsong or even white noise to a music label. Performers of classical music who upload their own recordings to Youtube report a similar problem. Apparently, the system cannot tell new recordings from previously published ones that belong to labels selling classical music.
2. Legal use
The system takes effect because it (correctly) detects the use of claimed content. However, this use is covered by one of several copyright exceptions and limitations. These limitations make it legal, for example, to quote from an existing work or to use it in a new work of parody or satire, without having to ask the rightsholder for permission. Before the new directive was passed, its critics explained again and again that automated filtering systems would not be able to recognise whether or not a specific use might be covered by an exception, such as the right to quote or – in the USA – the right to fair use.
3. Undue claim
Labels, studios and other production companies claim content via Content ID that does not belong to them.
This distinction is not merely an academic one. Rather, it highlights who is responsible for the overclaiming and how it can be resolved.
Mistakes of the first category happen because the filter doesn’t work correctly. However, Google is constantly refining the system. One day, so goes the promise, Content ID will be able to exactly recognise whether it’s Lang Lang playing Mozart’s “Rondo Alla Turka” or my neighbour.
Regarding the second category, the new directive (Article 17, Paragraphs 7 and 9; PDF, Page L130/120 ) directly refers to the exceptions and limitations. According to its provisions, the platforms shall be required to offer mechanisms to enable users to effectively make use of the rights granted to them by those limitations. Any complaints “shall be subject to human review”.
Time will tell how successful these solutions to errors in categories one and two will be. The important thing to note is that the responsibility for solving them is put solely and squarely on the platforms and users and not on the rightsholders. These can counter any complaints regarding the first category with the excuse that it’s the platform’s fault. In cases matching the second category the users have to actively fight back; their rights are not recognised automatically.
Undue claiming of immaterial rights
Now let’s focus on cases of the third category. With regard to those, the directive says … nothing.
The legislature assumes that labels, publishers and studios will feed the filters only with files referencing content which is actually theirs. German – or, as far as I’m aware of, any – copyright law does not impose any negative consequences on the undue claiming of rights, aka copyfraud. (Only in the context of the German law of unfair competition, UWG, is this behaviour penalised.)
Before the advent of upload filters, this made sense, because sanctions were unnecessary. If you wanted to enforce a copyright, you had to prove that it actually existed and was yours to enforce.
Of course, even back then the threat of a cease and desist letter could scare users, even if it was unjustified. (The US researcher Jason Mazzone, who is credited with coining the term “copyfraud”, complained about the practice and argued for penalisation in a paper back in 2005.) In the United States the music publisher Warner/Chapell had for decades collected royalties for the song “Happy Birthday” before a court finally declared it to be in the public domain. (If you ever wondered why so few people sing “Happy Birthday” in US movies or TV shows, now you know: it was too expensive.) But the general rule was: If you didn’t have a copyright, you could not control the use of a work.
The upload filters reverse that rule. They give corporations that only claim to be rightsholders a never-before-seen level of control. At the same time, the law does not attach any negative consequences to the unlawful claiming of copyright (or ancillary rights). Dealing with – or “punishing” – this kind of behaviour is left solely to the platforms.
Platforms as referees
Youtube provides mechanisms for resolving conflicts surrounding Content ID claims. If your newly uploaded video contains claimed content you will receive a notice. The video will be blocked, monetised or tracked, according to the policy set by the owner of the reference file.
If you think that the claim to your video – or, more often, parts of it – is illegal, you can file a dispute. The rightsholder can then release the claim. If they don’t react, the claim will automatically be removed from the video after 30 days. However, the – alleged or real – rightsholder can also uphold the claim or – and this is dangerous for you – take down your video and have it removed from the platform. If they do that, you will receive a so-called copyright strike. If you receive three strikes within a period of 90 days, your account will be terminated and all your channels and videos removed. Independent vloggers that earn money on Youtube will think twice about risking a strike or termination. A lot of them will choose to accept the claim or not use the claimed content, even if they are convinced that their use was legally justified. This opens the floodgates to censorship.
If a common user violates copyrights repeatedly, they will be punished. But no such strict rules exist for corporations that regularly claim content that belongs to someone else or is in the public domain. Youtube’s policy on this reads as follows: “Content owners who repeatedly make erroneous claims can have their Content ID access disabled and their partnership with Youtube terminated.”
Unlike alleged violators of copyright these perpetrators of copyfraud or overclaiming will not be punished automatically. Rather, it’s up to Youtube to decide. And Youtube will never ban a major label or major studio from its platform.
Yet overclaiming isn’t rare. One reason for this is that large corporations often don’t bother removing content from their reference files that is not exclusively theirs.
This led, for example, to five different US TV stations claiming video content provided by NASA which was actually in the public domain. The stations had reported about the Mars mission and then deposited the complete content of their own news coverage as reference material, including the public domain video. (A similar thing happened to the Mueller Report which was flagged as copyrighted by the Scribd platform’s filter BookID.)
In doing so, the TV stations violated Youtube’s policy which states that users of Content ID may only upload such content into reference files to which they own exclusive rights. The platform even provides easy tools for excluding another party’s content from the reference files.
But many rightsholders don’t bother to use those tools. When the band Pinkstinks performed their song “Not Heidi’s Girl” (a feminist song criticising the show “Germany’s Next Top Model” for promoting sexist stereotypes and a harmful body image) on the German TV station RTL, their own music video was blocked shorty after because Content ID recognised the song as belonging to RTL. Only after they lodged a dispute was the video restored. The reasons given for the snafu – according to the band – are symptomatic:
“A speaker for the station said, this happened all the time and could not be avoided because otherwise the station would become the victim of piracy.”
That is not true. This type of overclaiming can very well be avoided. The rightsholders are simply too lazy to correctly configure their reference files, because this configuring would cost time and money, which they are happy to save if they do not have to fear any repercussions. As long as keeping their reference files clear of undue content is more expensive for rightsholders than engaging in wilful overclaiming and only relinquishing claims when faced with protest, the legal framework provides incentives for overclaiming. Not to mention that rightsholders can also make money from claiming other people’s content.
Two recent decisions by the Landgericht (district court) Frankfurt give some reason for hope. In the first case, a major label had claimed a small competitor’s song as their own. The competitor successfully filed for a cease and desist order.
In the second case, a major label had acquired a non-exclusive license for a piece of music and used it in a new work. Despite the fact that – as explained earlier – this violated Youtube’s policy guidelines that rightsholders can only claim content that they own exclusively, the label uploaded the complete new work as a reference file. Other users who had bought a non-exclusive license to the same music from the composer all of a sudden had to deal with a major label’s claim to their content. This, too, was prohibited via court order.
At the last hearing in August 2019 the label – on the advice of the court – withdrew the appeal against the injunction and accepted the decision as final.
The new EU copyright directive explicitly states that users’ recourse to the courts shall not be restricted. The Landgericht Frankfurt has shown that this recourse can be effective. However, in both decisions mentioned above, the court ruled that the undue claim violated someone else’s copyright. This won’t be of any help if the unduly claimed content is not copyrighted at all, either because its copyright has expired or because it is genuinely public domain.
I’m not aware of a case in which users have tried to fight such a claim in a court of law. Anyway, in light of the immense control that filtering systems give to the (alleged) rightsholders, it seems neither just nor reasonable to burden the users with the cost and hassle of a lawsuit.
When implementing the EU directive into national law, the legislature should rather – for the first time in the history of copyright – make it illegal to claim content which either belongs to someone else or is public domain (meaning it belongs to everybody). And violating this prohibition must result in a penalty that is more expensive for labels, publishers, studios and stations than correctly configuring their reference files.
Tl;dr: Upload filters incentivise overclaiming. We should change that.
Disclosure: Marion Goller works as an attorney at the Media Kanzlei Frankfurt which won the cease and desist orders mentioned above.