In May 2024, Dmitry Medvedev, a former Russian president fond of chastising the West, lashed out at a smart speaker.
He criticized Alice, a device produced by Russian technology giant Yandex, for refusing to answer a question about Ukrainian monuments to Stepan Bandera, a Ukrainian who led an insurgent war against Soviet forces and collaborated with the Nazis during World War II.
Medvedev called Alice's creator "a terrible coward" and hinted that the company's managers should receive the status of foreign agent, a designation Russia uses to clamp down on dissent.
Alice's crime? By refusing to acknowledge monuments to Bandera, she was undermining Russia's narrative that Ukraine is a country full of Nazis. Russian AI had just slammed up against the Kremlin.
As Chinese companies such as DeepSeek prove that advanced AI can be built at lower costs, Russia’s global ambitions are faltering under the weight of ideological restrictions, an exodus of tech talent, and Western sanctions that cut off access to essential hardware.
Promising Beginnings
Vladimir Putin has made it clear that Russia's future is with AI. "Russia must become a world leader, not only in the creation, but also in the...penetration of artificial intelligence into all spheres of our lives without exception," the Russian president said at a government-linked AI conference in Moscow in December 2024.
Five years ago, Putin's words might have sounded more realistic. Russian computer scientists were getting global attention for their work building faster, more efficient AI models. In 2022, Sber, Russia’s leading financial company in AI development, launched Kandinsky, a text-to-image generation model. The model was designed to compete with OpenAI's DALL-E and Midjourney, positioning Russia as a contender in the AI-driven creative space.
Russia also had a pedigree of developing neural network technologies, which power AI by mimicking brain-like learning. Russian computer scientists had experience working with the state. A facial recognition system, for instance, has been linked to CCTV cameras in the Moscow subway since 2020.
An international AI expert with close ties to Sber's AI team, who spoke to RFE/RL's Russian Service on condition of anonymity, said at the time, "Russia had a chance of becoming one of the world leaders in generating images."
And then it all fell apart in February 2022 when Russia invaded Ukraine.
Shortly afterward, the United States and the EU placed sanctions on Russia's technology industry. In an effort to undermine the Kremlin's military capabilities and slow down the economy, the sanctions restricted sales to Russia of semiconductors, dual-use technologies, and advanced computing capabilities.
The effect was dramatic. NVIDIA, a U.S. technology company specializing in the graphics processing units (GPUs) that power AI's backbone, almost immediately stopped supplying chips to Russia. The Taiwan Semiconductor Manufacturing Company, which uses American technology, also ended its supply.
According to the expert with insider knowledge at Sber, the company has managed to accumulate around 9,000 GPUs since February 2022, although those numbers can't be independently confirmed by RFE/RL. By comparison, Microsoft purchased nearly 500,000 GPUs in 2024 alone.
Russia has tried to get around the sanctions by purchasing sophisticated components such as GPUs from China and India, but companies in these countries are wary of doing deals as they don't want to be sanctioned themselves for supplying Russia.
With their options drastically limited, Russia has to fall back on so-called "gray imports" from countries such as Kazakhstan, which cannot get access to chips quickly or in large quantities. To make matters worse, when chips burn out, companies like NVIDIA won't honor the warranties of those that were imported to Russia.
Human Brain Drain
Russia isn't just struggling with a deficit of parts but a deficit of people, made worse by the war in Ukraine.
Shortly after Sber launched Kandinsky in the summer of 2022, many of the specialists who had worked on the project had already left Russia.
"I talked to one of the developers from this team," the AI expert told RFE/RL, "and he joked that, after the war began, almost all the best specialists in neural-network image generation from Russia ended up in one rented apartment in Yerevan. [The developer] soon found a job in a Western company."
While Sber has downplayed the effects of brain drain on its operations, the numbers are stark. Maksut Shadayev, the Russian minister for digital development, communications, and mass media, said that around 100,000 IT specialists -- about 10 percent of the country's tech workforce -- left Russia in 2022. Many left immediately after the war began in February 2022, with another wave of departures after Russia announced military mobilization in September 2022.
Many skilled IT workers have moved abroad, creating hubs of exiled Russian tech talent in countries such as the United Arab Emirates, Georgia, Armenia, and Serbia. With salaries often higher, and a better and more open culture of innovation and collaboration, many end up staying.
Russia has tried to counteract this brain drain by increasing remuneration and creating domestic research centers in places like Kazan and Novosibirsk, but it has just been a drop in the ocean.
Russia's Censor-Bots
Adding to all these challenges are the Kremlin’s ideological constraints. Just as Putin once viewed the Western-dominated Internet as a threat to Russia's sovereignty and values, he now sees AI development through the same ideological lens.
In 2023, at a Sber conference, Putin highlighted the threats that "Western" large language models (LLMs) -- for example, GPT-4 or Gemini -- could pose to Russia. Such LLMs, he said, "simply ignore and cancel Russian culture" and impose ethics "that we oppose."
Medvedev's clash with Alice was just the tip of the iceberg. As Russian journalists from Meduza later found out, Alice doesn't only refuse to discuss Bandera but also Vladimir Putin.
To fight this Western threat, Russia wants to be self-reliant and build its own AI models. In December 2024, Aleksandr Dugin, a leading Kremlin ideologue, said that Russia should "develop a Russian version of artificial intelligence -- that is, a Russian artificial intelligence that will, without thinking, answer not only who Crimea belongs to, but also Kyiv, Kharkiv, [and] Odesa."
At the end of 2024, Putin instructed the government and Sber to deepen cooperation with China in the field of artificial intelligence. And on February 6, Reuters reported that Sberbank is planning to collaborate with Chinese researchers on joint AI projects.
China has been censoring LLMs since 2023. This censorship is carried out not only at the so-called alignment stage -- where the AI can be tweaked to not answer certain questions -- but also at the training stage, where at least part of the training data is preselected.
With Russia's strapped resources, trying to create a tight system of AI censorship would place a great strain on resources and already limited processing power -- not to mention the detrimental effect on freedom of speech. China can do it because its AI sector is vast. Unlike Russia, it has huge datasets, sophisticated infrastructure, and extensive and deep global partnerships.
For now -- or at least while international sanctions are still in place -- Russian AI will develop slowly. In a rating maintained by Chatbot Arena, a benchmarking platform, Russian LLMs created by Sber and Yandex ranked 12th and 17th, respectively, out of 44 participating models. The top three were Gemini Pro 1.5 (Google), Claude 3.5 (Anthropic), and ChatHPT-4o (OpenAI).
Falling behind in the development of AI would affect more than just chatbots and smart speakers. As artificial intelligence (AI) becomes increasingly important to scientific research, industrial innovation, and military capabilities, Russia could lose its economic competitiveness, technological sovereignty, and strategic influence on the world stage.
