Leading analyst at Mobile Research Group, Eldar Murtazin, stated that Russian neural networks will not be able to develop without competition with the global market. He expressed his opinion on the air of the radio station "Govorit Moskva" on March 20–21. According to him, neither state subsidies nor blocking foreign services will replace the competitive environment — it is the only real incentive for improving the quality of domestic models.
Why blocking foreign neural networks won't help
Foreign neural networks do not collect user data beyond what they themselves provide, according to Murtazin. Blocking will deprive Russians of tools that increase labor productivity, while domestic models have not yet reached a comparable level. Users will continue to use foreign services despite the restrictions. The analyst considers open models like DeepSeek to be an acceptable alternative — they can be deployed within the country without transmitting data outside.
Why subsidies to Yandex and Sber will not accelerate the development of Russian AI
Yandex and Sber are requesting tax deductions from the state for 500–700 billion rubles. Murtazin sees this as an attempt to reduce their own expenses: both companies are profitable and have sufficient funds for independent investments in AI.
Without competition, Russian companies will not be able to develop their neural networks at a good level. This is simply a desire to reduce their expenses. Moreover, both companies are profitable
The Ministry of Digital Development is preparing a bill on AI regulation with a priority for domestic models and possible restrictions on foreign services — the document is planned to come into force in 2027.
Read more materials on the topic:
- The Ministry of Digital Development has published a draft law on sovereign AI: neural networks should be created in Russia and trained on Russian data
- A neural network for finding breakdowns in engines with 99% accuracy was created at the Higher School of Economics
- A service for helping with floods was created at Tyumen State University: the AI service builds a route to the evacuation point by geolocation
