Russian developers of large AI models will soon face a shorter and clearer path to government support after the law on artificial intelligence development comes into force. Authorities plan to simplify the process of obtaining the status of a national or sovereign model, with all decisions verified using a single standardized set of test tasks.
The reference test, according to Izvestia, will be published in open access and updated on a regular basis. This approach will allow developers to understand in advance exactly what criteria the state intends to use when examining their neural networks, rather than discovering requirements only after submitting documents.
New statuses and related support measures will be available to companies including MWS AI, T-Bank, and other market participants. Several organizations that have passed state verification will conduct the expertise. The primary criterion will focus on Russian company control over the entire lifecycle of the model instead of the origin of every line of code.
Developers will be permitted to use foreign components under open licenses, provided they can independently modify, develop, and maintain the solution even after external technologies disappear. Bureaucratic hurdles will also be reduced: missing documents can be submitted after the initial application, and some redundant information will no longer be required.
Computing infrastructure must remain in Russia, although data centers may be rented or used on other legal grounds. The first sectors where domestic models will become mandatory include education and public services, covering educational processes, citizen appeals handling, and remote interactions.
The main provisions of the law will take effect on September 1, 2026, while requirements for sovereign models, content labeling, and intellectual property will begin on March 1, 2027. A complete ban on foreign AI solutions is not planned, although the government may restrict certain sectors to Russian models only.
