Roskomnadzor (RKN) plans to develop and implement a new internet traffic filtering system using machine learning technologies in 2026. 2.27 billion rubles will be allocated for these purposes. This is reflected in Roskomnadzor's digitalization plan, which was submitted to the commission on digital development for consideration.
Currently, all Russian internet traffic is filtered using technical means of countering threats (TSPU). These means are located on operators' networks in accordance with the law on the "sovereign internet." RKN independently maintains and installs TSPU, allowing it to block access to prohibited resources using deep packet inspection (DPI).
According to Oleg Terlyakov, Deputy Head of RKN, more than 1 million prohibited resources have already been blocked. On average, TSPU restricts access to 5,500 new addresses and domains daily.
RKN also manages a register of prohibited sites since 2012, adding pornographic, narcotic, and extremist materials to it. Operators are required to block these resources in accordance with laws protecting children.
RKN wants to use machine learning to find unwanted content, according to Comnews Research partner Leonid Konik. The solution will allow more effective blocking of sites that change addresses to bypass blocking.
Machine learning tools will allow blocking content not by internet addresses, but by words, expressions, sentences, or some other characteristics. And at the same time, find those who copied or quoted it.
In addition, RKN uses AI to search for prohibited information. This has reduced the time to detect problematic content to 6 hours, compared to 48 hours previously. This was stated by the head of the supervisory service, Andrei Lipov.
The use of artificial intelligence is also observed in the "Oculus" and "Vepr" systems, which help find prohibited materials in videos and analyze social networks.