MMOOZZGG.com

Basics

What is sovereign AI in Russia and what does the law require from business?

3 min read7 sources

Sovereign AI in Russia has two meanings. The legal one is the status of a model under the law on supporting the development of AI technologies, which came into force on 1 September 2026. The practical one is AI that works inside your own network and sends no data out. Both are below, with dates and caveats.

On this page · 4 sections

Under the law

The State Duma, Russia’s lower house of parliament, passed the law on 8 July 2026, and it came into force on 1 September 2026. A sovereign model: the developer is a Russian legal entity that runs the development itself and can fully reproduce its cycle, training included, and the model’s answers and data are processed in data centers in Russia. A national model: it allows components under open licenses, foreign foundation models included, but the servers and data are also in Russia. The classification criteria and rules of use take effect on 1 March 2027; systems already built have a transition period until 1 September 2032, if the data is processed in Russia. From 1 March 2027 the government gets the right to define cases where only sovereign or national models are allowed; there is no general ban on foreign models for business.

The draft on mandatory use

At the end of August 2026 a working group on the topic discussed a draft government decree, which became public on 8 September: organizations would have to use sovereign models if they cost no more than one and a half times as much as national ones. Fields: strategic planning and public administration, taxes and customs, healthcare and education, critical information infrastructure. No decision has been made; large developers pointed to risks for competition. According to the outlet, at that time only one developer’s model met the sovereign model criterion.

In practice

For most companies the question is not the status of the model but where the data lives and who sees it. Large foreign APIs are not officially available from Russia: Russia is not on the lists of supported countries of two large providers, and one of them has blocked traffic from unsupported countries since 9 July 2024. Russian cloud models solve data localization, and a local model on your own servers also solves the question of third-party access.

What a business should do

Split your data into three classes: may go outside, only to a Russian cloud, only inside the perimeter. Any API fits the first class, Russian cloud models fit the second, and a local model fits the third. If the company works in the fields named in the draft decree, plan a move to sovereign or national models for 2027.

Questions and answers

Must a business already use Russian models now?

No. There is no general ban, the classification criteria take effect on 1 March 2027, and the draft on mandatory use in certain fields has not been adopted.

Is an open model deployed on my servers sovereign AI?

Under the law, the status is given to a model of its developer, a Russian legal entity, after conformity is confirmed; someone else’s open model deployed at your company does not get that status by itself. In practical terms, it is the most closed option: the data goes nowhere.

Sources and fact-checking

Checked by the editors on 29 September 2026.

  1. CNews, which AI models will get sovereign status, 28 September 2026, an article marked as advertising (corp.cnews.ru)
  2. 3DNews, business may be required to use sovereign AI, 8 September 2026 (3dnews.ru)
  3. Telesputnik, the initiative on mandatory use of sovereign models, 8 September 2026 (telesputnik.ru)
  4. Claude Platform, supported regions (checked 29 September 2026) (platform.claude.com)
  5. OpenAI, supported countries for the API (checked 29 September 2026) (developers.openai.com)
  6. Caixin, OpenAI blocks API traffic from unsupported countries from 9 July 2024, 26 June 2024 (caixinglobal.com)
  7. Federal Law No. 243-FZ of 26 July 2026, text on ConsultantPlus (consultant.ru)

Found a mistake? Tell us

Publications

Mike Mozg

Agentic AI use cases at AT&T: nearly 500 processes now run on agents

AT&T no longer uses AI only to help staff work faster: it is rebuilding whole business processes around agents. SiliconANGLE reported on 14 November 2026 that nearly 500 agentic AI use cases at AT&T are already in production. Here is how the company rebuilds them, where a human still checks the agent, and what other large companies can take from it.

News4 min
Contact us