Draft Rules Push Firms Toward Sovereign Domestic AI Models
Russian officials are debating a draft government resolution that would push companies and organisations to prefer “sovereign” large AI models when those models cost no more than 1.5 times a broader “national” alternative. The discussion follows Federal Law No. 243-FZ, which took effect on 1 September 2026 and created legal categories for sovereign and national foundation models.
Filed under Policy and dated September 10, 2026, this AI4Russia briefing notes that the Ministry of Digital Development has confirmed talks are ongoing and that no final mandatory sector list is approved. Reporting around the draft often cites Sber’s GigaChat as fitting the stricter sovereign definition, while Yandex’s Alice AI is discussed as a national-model example that may include open-source components. Both companies have urged further deliberation and competition.
Why it matters: procurement teams in Moscow and across Russian industry need clarity on when domestic-only infrastructure is required versus when mixed stacks remain allowed. A 1.5× price rule would reshape vendor shortlists even before enforcement dates firm up. Overseas model access and chip constraints already narrow options; regulatory preference would narrow them further.
What it means in practice
For careful operators, the response is documentation first: inventory which models touch customer, employee, or industrial data; record where inference runs; and assign a human owner for each high-impact workflow. Run time-boxed pilots with logging and override rights. Treat draft rules as planning signals, not finished obligations, until the government publishes a final list of sectors and cases.
Caveats come first. Working-group summaries can overstate consensus; vendor labels of “sovereign” can shift as criteria tighten toward March 2027 secondary rules. AI4Russia presents the draft as directional context only.
What to watch next: an official sectors list; how price comparisons will be measured; and whether industrial buyers publish migration timelines. Readers can continue on the AI4Russia homepage for related stories, or browse the Newsroom for additional briefings.
Bottom line: treat this update as orientation, not instruction. Domestic AI policy in Russia is moving quickly and unevenly. Organizations that benefit most will measure pilots honestly, keep people accountable, and avoid locking critical workflows to a single vendor before rules stabilise.