Politics 5 min read
VK’s behavioural profiles blur ad tech and state surveillance risk
VK’s shared Discovery infrastructure is designed to make user interests legible across content, search and advertising. In Russia, that commercial architecture operates alongside long retention requirements and an expanding security-state role in network control.
The most important part of VK’s new “neural profile” is not the label. It is the institutional architecture behind it. VK is moving toward a system in which recommendations, search and advertising draw from a shared understanding of user behaviour across multiple products.
The new model is described as analysing sequences of actions rather than treating likes, views and comments as isolated events. That allows a platform to estimate persistence. A one-off view may be noise; repeated engagement across VKontakte, VK Video and VK Clips can be interpreted as a stable preference or an emerging intent.
From a business perspective, that is a substantial improvement. Advertising performs better when a platform understands not only what a person has touched but how their interest is evolving. Search becomes more useful when it can incorporate earlier behaviour. Recommendations become stickier when they distinguish a temporary curiosity from a long-term habit.
VK has been consolidating these functions on its Discovery platform since 2025. In July 2026, it announced Discovery AI and said the system could personalize results using interests across services. The company has also described multimodal models that predict reactions and build viewer portraits from content interaction data.
Rollout materials attribute a 15% rise in sharing, a 10% increase in full-screen opens and a 5% increase in likes to the new approach. These remain attributed performance claims rather than audited results.
The larger issue is governance. In the UK and EU, advanced profiling is normally debated through data-protection law, transparency duties and regulatory oversight. Russia combines commercial profiling with a much more coercive state-access environment.
Article 10.1 of Russia’s information law requires organizers of information dissemination to retain specified data about users and communications. The retention period for certain metadata and user actions was extended to three years from January 2026. Separate rules allow message content to be stored for up to six months.
Beyond retention sits SORM, the technical interception regime. An EU regulation adopted in 2026 describes SORM as covering internet use and social networks and says copied traffic is available to the FSB. It also refers to the collection of location and behavioural information. In June, the Financial Times reported that the FSB had acquired a stronger role in controlling Russia’s internet, including decisions affecting connectivity and access.
This does not establish a direct hand-off of VK’s proprietary neural profile to the security service. There is no public evidence in the cited VK material of such a mechanism, and the legal retention duties are framed around categories of communications and user data rather than an internal recommendation model.
That limitation does not remove the institutional risk. The centralisation of behavioural intelligence means that a private platform can create a richer analytical layer from data that may also exist, in different forms, inside a state-access regime. The distinction between commercial and security use becomes a question of governance, not merely of technical capability.
The analytical power of such data is well established. A Scientific Reports study published in 2024 used data from consenting VK users to model personality traits and cognitive attributes. It was an academic project with informed consent, not a commercial or intelligence application. Its relevance lies in demonstrating how much information can be inferred from a social-media history.
For businesses, the case illustrates why data concentration creates both value and liability. A single behavioural profile can power engagement, conversion and personalization across several products. The same concentration magnifies exposure to breaches, coercive disclosure and changes in state policy.
VK’s neural profile is therefore best understood as part of a broader shift in Russian digital infrastructure. Commercial systems are becoming better at modelling citizens at the same time as the state is strengthening the legal and technical mechanisms through which digital activity can be monitored. The risk comes from the convergence of those systems, even where a direct bridge between them has not been publicly proven.
