Experts warn about EU law that could change the internet
March 12, 2019
Experts warn about EU law that could change the architecture of the internet, forcing websites to install flawed and expensive filters that would block satirical content like memes and lead to digital monopolization.
Article 13 of the EU's new directive on copyright is under sustained criticism from media experts and campaigners, warning of a risk of unintended censorship with no working exceptions for satire or small businesses, which could lead to a filtering of legal content and a further monopolization of the internet.
"User-generated platforms would look completely different from what we know today," warned Dr. Stephan Dreyer at the Leibniz Institute for Media Research. Although the initiative has strong support from publishers like the Axel Springer Group and copyright agencies like the German GEMA service, media rights experts from the Science Media Center Germany have issued a warning about the consequences of the legislation on freedom of speech and in particular satire.
Julia Reda, the German Pirate Party MEP who has dedicated her entire legislative period to reforming this legislation, told DW that “The overall effect of the proposal is that the internet would become more like television, as a smaller number of people and platforms would be able to create and share."
Article 13 is aimed to update the EU’s outdated copyright legislation. Under the proposed legislation, almost all companies which host user-generated content, which includes most platforms that host pictures, could be liable to prevent copyrighted content being uploaded to their websites.
Although installing filters is not mandated by the new law, according to Thomas Matzner, media professor at the University of Paderborn, the only feasible way for companies to enforce this would be to install expensive filters which recognize and block copyrighted content. Matzner also warns that “This will encourage further concentration on the Internet. Many small initiatives and start-ups will suffer because they will not be able to cope with the effort they have to make under Article 13. The supposed exception for small firms only applies for the first three years.”
It’s the algorithms, stupid
Tobias Keber, media rights professor at Stuttgart's Media University HDM compared it to Facebook's nudity detection algorithm, which has been continually improved, but still has difficulty addressing nudity properly in an artistic context. He noted that the issue is not just about photos or illustrations. "Copyright law deals with a large number of very different works: text, images, audiovisual content, computer programs or works of dance art. That an algorithm in this respect can recognize all conceivable forms and contexts in criticism, satire or quotation is absolutely out of the question. This requires the assessment of people, and that is a good thing."
Parliamentarian Reda pointed to a ironic example of how copyright filters are already overzealous: "We livestreamed a protest [against this proposed law] in Berlin last weekend and YouTube recognized it as copyright infringement because of techno music in the background, which YouTube’s filter recognized as belonging to the techno artist. The video was an original piece of video journalism, so these filters are a threat to freedom of expression."
Axel Voss, MEP and the bill’s proponent, points out that memes are specifically protected under the InfoSoc directive. Researchers from the Science Media Center Germany argue, however, that the law would have the unintended effect of blocking satirical content including memes, because there is no foreseeable way that digital filters could recognize the difference between copyright material and satire, and would therefore block them as forbidden material.
Dr. Dreyer warns that "current methods of machine learning can only take such factors into account to a very limited extent." Whereas Florian Gallwitz, professor of media informatics at Nuremberg Tec, says "the reliable automatic recognition of parodies or quotations, is completely out of the question. When automatic upload filters are used, quotes and parodies are inevitably blocked."
Legislators claim that only for-profit companies will be affected, but Reda pointed out that in cases of a copyright infringement both the company and the user would be liable to pay damages. What does this mean? If a fan posted an unlicensed song or remix to a public Facebook group, this could be a copyright infringement.
Artificial Intelligence and Art
Pirate MEP Reda also said that any private company which hosts content uploaded by users such as pictures or audio files would be liable for copyright breaches. Tinder or Trip Advisor, for example, could effectively be forced to buy licences for all the content that exists in the world, massively increasing their costs.
The stated aim is to prevent unlawful breach of copyright so that creators such as publishers and musicians earn more. But Reda warns this could backfire for sites like Patreon, which gives paying users exclusive access to creator’s works, but would have added expenses as a result of the legislation and would then actually pay creators less.
Marcus Liwicki, professor and chair of machine learning at Lulea University in Sweden, warned: "The new law is generally difficult to implement and it is unfair to shift the burden of review to the platforms. Flea markets would have to check all pictures, films and records they sell in the future to see whether they really are originals and not illegal copies. Or telephone providers may have to check all telephone calls to see if they are not transmitting copyrighted content during conference calls with more than X participants."
But Voss disagrees with his fello MEP Reda and others on the scope of the law, telling DW: "We are just concentrating on platforms that are infringing on copyrighted works like YouTube. Not for dating platforms or merchandising or local social platforms. Only 1.5 percent of internet platforms will be affected."
And what if it affects sites that host user-generated content negatively? "We all have legal obligations to fulfill. If you have a massive platform like YouTube you will have to use a technological solution. Everyone has these obligations. They have created a business model with the property of other people – on copyright protected works. If the intention of the platform is to give people access to copyright protected works then we have to think about whether this kind of business should exist. The new legislation is improving the situation for the European creators industry."
He admitted, however, that the current version of the law had few explicit exceptions, only small sites younger than three years old, due to a compromise with France in the European Parliament.
Another controversial part of the legislation is Article 11, which relates to news aggregation sites such as Google News, but which could have more serious unintended effects. Article 11 enforces copyright on text content such as headlines and teasers, and is based on a German law that even Voss describes as “not the best idea”.
This law is aimed at protecting the rights of journalists and news sites, but when Spain implemented a similar law, Google News simply shut down in the country, and news sites suffered a noticeable drop in clicks (and therefore revenue).
According to media researcher Keber: "One has to ask oneself whether a concept that has demonstrably not worked in Spain (where Google News was discontinued) and Germany (where people are afraid that Google will unlist them, and grant free licenses) should now work at the level of EU law."