- Category
- World
Russia Says Its AI Lacks the Brains—but Excels at “Traditional Values”

Russia is promoting a model of "sovereign" artificial intelligence to countries across the Global South, pitching state-aligned systems to governments wary of the privacy practices and content controls of Western large language models, according to The Moscow Times on June 3.
Alexander Vedyakhin, first deputy chairman of the management board of Sberbank—Russia's largest bank and a leading domestic AI developer—outlined the strategy in an interview with Reuters ahead of the St. Petersburg International Economic Forum (SPIEF).
We bring you stories from the ground. Your support keeps our team in the field.
Russia trails the US and China in the global AI race, with Sberbank and the technology company Yandex developing their flagship large language models, GigaChat and YandexGPT. Vedyakhin indicated that "friendly" states are seeking networks trained on local content that reflect their own traditions and outlook.
"What we have offered and will continue to offer is, of course, the creation of sovereign models—for partners and those interested," Vedyakhin stated. "Proposals were coming to us: install it, come, we are ready to pay money."
He acknowledged the systems would lag their rivals at first. "Yes, it will initially be slower, in some areas not as smart as Anthropic, Grok, or DeepSeek, but it will correspond to traditional values," he added.

GigaChat is available to the public with open code, and Sberbank's partners can download it themselves, Vedyakhin explained. The obstacle, he noted, is that those partners lack the specialists needed to configure and train the model so it does not generate content a given country considers undesirable.
"This is highly intellectual work—turning a raw product into a model that a specific country needs is a major IT project in itself," he continued. "There is strong demand for exactly this from the Global South and from countries that want, but cannot afford, to develop sovereign AI."
The Russian leader, Vladimir Putin, declared last week that Russia ranks among the three countries capable of developing its own AI models for sensitive uses, such as government operations and the defense sector.
During Putin's visit to China in May, Herman Gref, the head of Sberbank, discussed purchasing Chinese chips to run GigaChat, as Western sanctions continue to block Russian access to advanced foreign hardware. Vedyakhin declined to detail those talks, characterizing them as commercial negotiations.
-77cf0a1e23b793fd430ddab3310b5a8a.png)
He also pointed to Nvidia's CUDA programming software, now an industry standard underpinning existing large language models, as a barrier for newcomers, warning that a rival platform would struggle to displace it.
Vedyakhin predicted the field would shift toward smaller, specialized models that consume fewer resources. Sberbank's GigaChat Ultra exceeds 700 billion parameters, he indicated, but he argued that users want concrete problems solved at a clear price rather than raw parameter counts, making model compression the next step.
Moscow's drive to project AI self-reliance has repeatedly run up against its dependence on foreign technology. Russia's first officially registered AI assistant for government employees, added to the national software registry in recent weeks, was found to run on open-source engines built by American developers, including the text-generation system Ollama and the database project ChromaDB.
Discuss this article:
-9a7b3a98ed5c506e0b77a6663f5727c5.png)
-111f0e5095e02c02446ffed57bfb0ab1.jpeg)

-72b63a4e0c8c475ad81fe3eed3f63729.jpeg)



-605be766de04ba3d21b67fb76a76786a.jpg)

