What VK’s “neural profile” means inside Russia’s surveillance system

A single like is noisy data. It can reflect a joke, an accident, a passing curiosity or a topic a person will never return to. A sequence of behaviour is different. VK’s new “neural profile” approach is intended to look at repeated actions across VKontakte, VK Video and VK Clips and use them to build a more stable picture of what a user actually cares about.

The product logic is straightforward. If someone watches one video about buying an apartment, the platform knows very little. If the same person repeatedly watches mortgage explainers, opens neighborhood guides, comments on renovation clips and shares property posts, the pattern begins to look like a persistent interest. Recommendation systems can react earlier and advertising systems can become more precise.

VK has been moving toward this architecture for some time. Its Discovery platform combines search and recommendation technologies across the company. In July 2026, VK said Discovery AI could use interests drawn from multiple group services. The company had previously told investors that one of its multimodal models predicts user reactions and creates a viewer portrait from interactions with content.

The rollout materials cite three engagement changes: users sharing content 15% more often, opening it full screen 10% more often and liking it 5% more often. These figures are rollout claims from the supplied launch description and should not be read as independently confirmed measurements.

If this were only a recommendation story, it would be another step in the global race to make feeds more addictive and ads more efficient. Russia’s legal environment changes the significance of the same technical capability.

Article 10.1 of Russia’s information law imposes retention duties on organizers of information dissemination. The current rules require certain information about users and communications activity to be stored inside Russia for three years, a period that took effect at the start of 2026. Message content can be subject to retention for up to six months under separate government requirements.

The state also has a technical surveillance layer. The European Union’s 2026 regulation on Russian surveillance technology describes SORM as a system used to monitor phone calls, internet use, messages and social networks. The document says user traffic can be copied to infrastructure accessible to the FSB and that the system can collect information about location and behavioural patterns.

The Financial Times has separately reported that the FSB’s power over Russia’s internet increased substantially from mid-2025, with the security service playing a larger role in restrictions and shutdowns. That trend makes the governance of large domestic platforms more important, not less.

There is still a line that should not be crossed without evidence. Public VK material does not establish that its internal neural profile is directly transferred to the FSB. Russia’s retention law does not say that a proprietary recommendation embedding or interest model must be delivered as a finished file. It is therefore more accurate to describe a surveillance risk created by the surrounding system than to claim a specific data-sharing mechanism that has not been demonstrated.

Why does that distinction matter? Because the concern is not dependent on a single secret interface. The more information a platform can infer from routine activity, the more valuable its stored event data becomes. A sophisticated profile can be reconstructed or approximated from many underlying signals even if the exact commercial model remains internal.

Academic work points in the same direction. A 2024 Scientific Reports study used VK data from 1,358 consenting participants to infer personality-related traits. The researchers operated under informed consent and ethical review, and their project was unrelated to VK’s commercial recommendation system. Still, it shows how behavioural traces can support conclusions far beyond the visible content of a profile page.

The practical effect for users will probably feel pleasant: a feed that adjusts faster, fewer irrelevant videos and ads that better match current interests. The political effect is less visible. Russia is building a digital environment in which private platforms generate increasingly detailed behavioural models while state institutions retain strong legal and technical access to communications data.

That combination is what makes VK’s neural profile consequential. The feature is not proof of a new spy database. It is evidence that the raw material of surveillance — detailed, longitudinal behaviour — is becoming richer and more coherent by design.