Last November, major media and tech companies in Russia signed a landmark memorandum in order to tackle the rise of pirated content on the Internet.
Central to the agreement was the creation of a database populated with links to material deemed copyright-infringing by entertainment industry groups
Operators of search platforms agreed to query the database every five minutes and then, within six hours, remove links to the content from their search results. The same applies to sites that actually host content, such as Yandex.video and RuTube, for example.
Population of the database got quickly underway and according to the Media Communication Union (MKC), which represents the interests of major media and telecoms companies, now contains around 300,000 links. However, the companies involved feel that the system can be much improved with the addition of custom software.
To that end, this week the MKC revealed that it has begun testing a new anti-piracy system that will allow content to be added to the database more quickly and efficiently. The tool not only allows URLs to be entered manually but also accepts input from “specialized search systems” that are able to identify illegal content.
“An automated solution based on specially trained neural networks is used to analyze the content of sites specified in the rights holders’ reports,” MKC announced.
MKC says that manual testing is also used in a number of cases, with the results being sent to the neural network for “additional training.” As the project develops, the aim is to require the intervention of human operators on much fewer occasions.
“A modern software solution based on self-learning systems will significantly increase the effectiveness of the fight against Internet piracy and will further increase the consumption of legal video services,” said MKS President Mikhail Demin.
An almost fully-automated anti-piracy seems like a big ask, particularly when machines are often blamed for erroneous takedowns. However, for the head of Russian telecoms watchdog Roskomndazor, removing humans from the equation where possible will make the system more effective.
“This is a significant event for both sides of the Memorandum,” Alexander Zharov says.
“An automated solution for interaction within the framework of the Memorandum will help to increase the reaction times and reduce the risks associated with the ‘human factor’.”
It is not yet clear whether the system under development represents anything drastically new in the anti-piracy space, or whether the “self-learning” component will amount to anything more than scraping allegedly-infringing URLs and then sending these to the registry.
Nevetherless, beta tests are already underway and it’s expected that the finished product will be with rightsholders before the end of July.
The memorandum and supporting technical efforts are currently in operation voluntarily until the terms of the agreement run out November 1, 2019. However, given the level of commitment being shown by the parties involved, it’s expected to continue until the terms of the memorandum can be written into local law.