26.07.2026, Russia.
The introduction into the law on the development of so-called artificial intelligence (AI) technologies in the Russian Federation of the concept of a “national” large foundational model (LFM) of AI, which allows the use of foreign components, contradicts the goals of state security and technological independence, notes the IT editorial board of the Rossa Primavera News Agency on July 25.
On July 17, the Federation Council of the Russian Federation approved a law on the development, implementation and regulation of AI. The law is aimed, among other things, at ensuring state security, including the sphere of its spiritual and moral values, technological independence and Russia’s leadership in the development and application of AI.
The document introduces the concepts of “sovereign” and “national” AI models, for whose developers measures of state support are introduced. Financial, property, guarantee and information support are envisaged, the specific measures of which will be determined by the Government of the Russian Federation. Developers are also granted access to information in federal information systems and other state databases for training AI.
The key difference between “sovereign” and “national” AI models is the developer’s ability to use third-party components and AI models, including foreign ones, in the latter. Whereas for a “sovereign” model, “full technical and technological reproducibility of the development cycle by the developer … including its training” is required.
It is obvious that the easiest way to obtain a working AI model and start monetizing it is to create a “national” model, roughly speaking, made according to the well-known principle of sticking a “made in Russia” label on a Chinese product. The key point is that potential developers of such models gain a competitive advantage over potential developers of “sovereign” models. In other words, the law clearly does not create conditions that stimulate the creation of “sovereign” models and accelerate the achievement of technological independence and reduce risks to state security.
Moreover, the need to adjust and control the responses of foreign models that are trained in the interests of foreign entities — their owners — increases the risk of fatal errors in final responses related to the sphere of domestic spiritual and moral values. This is a particularly delicate area where even one mistake may be enough to undermine a person’s patriotic feeling and make him a potential accomplice of hostile states. This problem is already relevant in itself: reports that Russian citizens, including young people, are detained for ties with hostile intelligence services are heard regularly.
It should also be noted that the law provides for a six-year moratorium (until September 2032) on possible prohibitions on the use of AI models that are neither “sovereign” nor “national” if they are deployed on Russian computing power. This factor additionally levels out the desire for breakthrough development of sovereign technologies.
There is no doubt that the task of developing domestic artificial intelligence technologies is a state strategic task. However, the scheme for solving it outlined in the law, where the emphasis is primarily on quick tactical achievements, raises serious concerns.
Source: Rossa Primavera News Agency
