AI-powered internet censorship

Russia’s internet watchdog wants to use artificial intelligence to block access to restricted information on the Russian internet. AI should help the agency block unwanted content three times faster — within an hour of publication — and strike down content more accurately, the authorities say.

• Russia’s Roskomnadzor communications regulator plans to start using AI this year to create and maintain a register of blocked sites, according to documents seen by Kommersant.
• The agency already uses its own information system to seek out and block access to online content that is prohibited in Russia. Tender documents issued by Roskomnadzor imply that in addition to tracking prohibited content, it has the ability to classify it according to its character (based on a neutral, negative or positive opinion of the author) and also to find copies and duplications of banned material and sites.
• In 2023 the system typically took three hours to identify unlawful content from the moment it was published. This year, the aim is to reduce that to two hours using AI and by the end of 2026 the average deletion time should be down to just 60 minutes. The agency is also targeting a drop in the error rate from 20% to 10%. AI should enable Roskomnadzor to block content both faster and more thoroughly, since it makes it possible to “identify complex contextual connections between text fragments, finding hidden patterns and associations,” the agency said in the documents obtained by Kommersant.
• The agency’s site-blocking operations are already partially automated. In March, it stoppedupdating a public register of forbidden sites, because Roskomnadzor no longer needs to notify telecom operators of what sites to block. Previously, operators were responsible for blocking content but now the agency can do it directly thanks to technological and equipment upgrades.

Why the world should care

The first thing that comes to mind when hearing about plans to use AI to censor the internet is that this is a blatant waste of taxpayers’ money on technology that the authorities cannot begin to understand. But in recent years the internet watchdog has sharpened its teeth. In 2018, Roskomnadzor’s failed attempt to block Telegram prompted ridicule and amusement across the internet. Since then, though, a new management team has begun to introduce more effective means of blocking sites, achieving results in some areas comparable to the Great Firewall of China.