AI models in Russia will undergo a "spirituality" check for access to government procurement.
The Russian artificial intelligence market is on the brink of tectonic regulatory changes. Developers claiming the status of a "sovereign" or "national" model will have to confirm that their algorithms comply not only with technical requirements but also with traditional spiritual and moral values. This initiative, actively discussed in late July and early August at the ANO "Digital Economy" platform, radically changes the rules of the game for the entire AI sector.
New Reality: Two Statuses and a Strict Filter
The AI law that came into force on July 26 is framework-based, but it is already clear that key rules will be spelled out in subordinate regulations. This involves introducing two special statuses for large fundamental models. Obtaining either of them—sovereign or national—will become a mandatory condition for access to government preferences, including benefits in public procurement.
The assessment procedure involves a multi-stage review. The developer submits an application to an independent testing laboratory, providing a description of the architecture and request filtering mechanisms. Experts will check the model's responses against special benchmarks—sets of tests that have yet to be developed jointly with businesses. Particular attention will be paid to resilience against so-called prompt injections—attempts to "hack" AI with malicious instructions hidden in input data.
Who Issues the Verdict and How to Avoid Subjectivity
The compliance conclusion will be sent to the government commission on digitalization, and only a positive verdict will open the doors to state support. At the same time, no one intends to check all models on the market—only those claiming high status. The first implementation areas will be education and public services.
Notably, the review can be conducted by several accredited laboratories at once, rather than a single authorized organization. The scheme is built in two stages: first, the developer independently tests the model using a general risk-based methodology, then an accredited laboratory evaluates these findings and may selectively verify the system's resilience. The final decision will remain with the Ministry of Digital Development, while security for government systems will be overseen by the FSB and FSTEC.
Expert Opinion: The Main Risk Is "Subjectivity"
The key problem market participants warn about is the vagueness of the criteria. Categories such as "civic-mindedness" or "historical memory" are extremely difficult to formalize in a technical specification. Business architects and lawyers agree that the only workable approach is procedural: a unified reporting format, reproducibility of results, and a public registry of conclusions. Without this, there is a risk of selective application of the rules.
My analysis: introducing "spiritual" certification for AI is not just a regulatory barrier but a powerful signal to the market. On one hand, it creates a precedent where compliance with values becomes a technical characteristic of the product. On the other, without clear metrics and a transparent appeals system, this initiative risks turning into a tool of administrative pressure. Investors and developers should closely monitor how the benchmarks are defined—they will determine whether this filter becomes a driver of development or a brake on innovation.