ECONOMY
Russia at the Forefront of Front-Line AI
June 9, 2025
-
Ksenia Buksha
Journalist and writer
Journalist Ksenia Buksha explores how the Russian AI sector, having been cut off from the global market, has been redirected to innovate for the battlefield and cyberattacks.
The original text in Russian was published in Republic and is being republished here with their permission.
The state of Russian AI
The Russian AI market is small compared to the US, EU and even China. There are too many obstacles to its development. Back in 2021, a study commissioned by the Russian government identified serious problems with the Russian AI sector: a shortage of key personnel, a weak venture capital market, low penetration of Russian products into foreign markets, dependence on imported products and services, slow implementation of products by business and government bodies, and a weak link between ideas and their implementation as products.
After the start of the war, these problems were compounded by others: the inaccessibility of global infrastructure and advanced technologies (since most government AI development projects are under sanctions) and the emigration of many scholars and IT specialists.
Before sanctions, Russian tech giants like Yandex and Sber bought many important components needed for AI products ready-made: Nvidia and Intel GPUs, storage, and cooling systems, among others. Russia had failed to develop its own semiconductors or competitive GPUs, and it is now impossible to close this gap.
There is a “gray” import market that continues to supply processors to Russia, including such advanced ones as the Nvidia H100. However, old models are what is on offer in the gray market, meaning Russia is perpetually 2-3 years behind. When it comes to AI, such lags are critical.
Small wonder that plans for AI development have been tempered in recent years: whereas in 2019, RUB 56.8 billion was penciled in to be allocated from the federal budget for AI development until 2024, in 2023 it was only RUB 24.6 billion until 2030. Meanwhile, extra-budgetary funding was reduced more than tenfold: from RUB 160 billion to RUB 14 billion.
Yet this does not mean the development of AI in Russia has stopped. Rather, it has been put entirely on a war footing. Sure, Russia is lagging (or rather, not even participating) in the race to create AI for civilian purposes, be it supercomputers or LLMs.
The state of Russian AI
The Russian AI market is small compared to the US, EU and even China. There are too many obstacles to its development. Back in 2021, a study commissioned by the Russian government identified serious problems with the Russian AI sector: a shortage of key personnel, a weak venture capital market, low penetration of Russian products into foreign markets, dependence on imported products and services, slow implementation of products by business and government bodies, and a weak link between ideas and their implementation as products.
After the start of the war, these problems were compounded by others: the inaccessibility of global infrastructure and advanced technologies (since most government AI development projects are under sanctions) and the emigration of many scholars and IT specialists.
Before sanctions, Russian tech giants like Yandex and Sber bought many important components needed for AI products ready-made: Nvidia and Intel GPUs, storage, and cooling systems, among others. Russia had failed to develop its own semiconductors or competitive GPUs, and it is now impossible to close this gap.
There is a “gray” import market that continues to supply processors to Russia, including such advanced ones as the Nvidia H100. However, old models are what is on offer in the gray market, meaning Russia is perpetually 2-3 years behind. When it comes to AI, such lags are critical.
Small wonder that plans for AI development have been tempered in recent years: whereas in 2019, RUB 56.8 billion was penciled in to be allocated from the federal budget for AI development until 2024, in 2023 it was only RUB 24.6 billion until 2030. Meanwhile, extra-budgetary funding was reduced more than tenfold: from RUB 160 billion to RUB 14 billion.
Yet this does not mean the development of AI in Russia has stopped. Rather, it has been put entirely on a war footing. Sure, Russia is lagging (or rather, not even participating) in the race to create AI for civilian purposes, be it supercomputers or LLMs.
“But for military applications of AI, computing power is not so important, and in this area, Russia has proven itself quite competitive.”
The Korsar unmanned aerial vehicle on Red Square during the Victory Day Parade, May 9, 2018. Source: Wiki Commons
Angry birds
Russian AI has been highly militarized from the very beginning. The development of the sector began just after the annexation of Crimea and the start of the war in the Donbas. After the full-scale invasion of Ukraine, military uses of AI became the priority. In autumn 2024, Putin introduced the term “sovereign AI”: “we can certainly join the ranks of leaders, considering our certain advantages. Sovereignty is a highly important component… Sber and Yandex are actively engaged in this area, and overall, their work has been quite successful.”
Putin was referring to AI models developed by Yandex, Sber and possibly other less public firms and organizations that are used for military purposes. This is what Russian officials have in mind when they boast about Russian AI.
There is not much in the way of publicly available information on Russia’s military AI efforts, but researchers have some ideas.
The main thrust of Russian AI innovation is “brains” for drones.
To operate independently from operators and navigate effectively behind enemy lines, a modern drone needs to be equipped with AI. Satellite navigation cannot be relied upon, as the enemy is trying to intercept the drone and putting out fake signals to knock the drone off course. Hence the need for AI in alternative navigation systems, such as optical.
Drones, preloaded with satellite images, constantly compare real images with this reference material: they identify terrain features and objects, plan the route and optimize the trajectory to a target. Other AI models perform motion prediction, dynamic trajectory adjustment and other navigation tasks. The system does not require GPS or communication with operators – it independently makes all the decisions, including target detection and the timing of impact.
These drones are equally effective against moving targets, which they identify using thermal imaging (usually at night) or visual recognition. They can be deployed for operations close to the front lines (since they are immune to electronic warfare systems) or to strike deep into enemy territory.
Russian AI has been highly militarized from the very beginning. The development of the sector began just after the annexation of Crimea and the start of the war in the Donbas. After the full-scale invasion of Ukraine, military uses of AI became the priority. In autumn 2024, Putin introduced the term “sovereign AI”: “we can certainly join the ranks of leaders, considering our certain advantages. Sovereignty is a highly important component… Sber and Yandex are actively engaged in this area, and overall, their work has been quite successful.”
Putin was referring to AI models developed by Yandex, Sber and possibly other less public firms and organizations that are used for military purposes. This is what Russian officials have in mind when they boast about Russian AI.
There is not much in the way of publicly available information on Russia’s military AI efforts, but researchers have some ideas.
The main thrust of Russian AI innovation is “brains” for drones.
To operate independently from operators and navigate effectively behind enemy lines, a modern drone needs to be equipped with AI. Satellite navigation cannot be relied upon, as the enemy is trying to intercept the drone and putting out fake signals to knock the drone off course. Hence the need for AI in alternative navigation systems, such as optical.
Drones, preloaded with satellite images, constantly compare real images with this reference material: they identify terrain features and objects, plan the route and optimize the trajectory to a target. Other AI models perform motion prediction, dynamic trajectory adjustment and other navigation tasks. The system does not require GPS or communication with operators – it independently makes all the decisions, including target detection and the timing of impact.
These drones are equally effective against moving targets, which they identify using thermal imaging (usually at night) or visual recognition. They can be deployed for operations close to the front lines (since they are immune to electronic warfare systems) or to strike deep into enemy territory.
“The entire system is resilient to changes in the surrounding environment: even if a forest has burned down or a bridge is destroyed, it can usually figure out how to continue to a target.”
Of course, these drones require powerful chips: at least Nvidia Jetson or better yet programmable logic devices, such as the AMD/Xilinx Zynq UltraScale+, which support real-time inference of neural networks. Russia can easily import such chips through the gray market.
AI can also teach drones to coordinate a swarm and attack together. They observe nearby units like birds in a flock. It is almost impossible to stop these swarms, unless they are intercepted by another swarm. Prototypes of such systems are currently in development.
Managing swarms requires advanced AI models capable of making decisions in real time. Rostec is attempting to develop them.
Cyber(in)security
Another key area of Russian AI development is analysis of data stolen in cyberattacks. The goals of such analysis vary, one of them being better planned military operations.
Russian hackers have always been good at stealing data. But only now have they got their hands on a tool that can analyze it.
In addition, AI is used to hack users through deepfakes and phishing schemes to extract passwords, among many other methods.
Ihor Malcheniuk, director of the Department of Cyber Protection of the Ukrainian State Service of Special Communications and Information Protection, told the Munich Cybersecurity Conference how Ukrainian military personnel are targeted on encrypted messaging platforms: they receive personalized messages designed to trick them into clicking malicious links that hack their accounts and reveal classified information.
Russian cyberattacks on Ukraine are increasingly about cyberespionage. Ukrainian cybersecurity experts have also observed growing cooperation between Russian state-backed hackers and cybercriminal groups.
It’s not just Ukraine that is under attack: UK government minister Pat McFadden said Russia is trying to use AI to enhance cyberattacks on the country’s infrastructure. UK intelligence believes AI will speed up automated reconnaissance, large-scale data analysis and the adaptation of attack strategies in real time. In other words: it will become easier to prepare and carry out cyberattacks, and their efficacy will rise.
There are concerns about AI broadly and its impact on human life. Currently, however, much more real and relevant is the intersection of AI and “natural” human cruelty, lust for power and desire to dominate others.
Perhaps the AI market in Russia appears so small because some expenditures are simply not advertised. The good news is there are not only players like Russia in the AI market, and they are making their own moves in this game.
AI can also teach drones to coordinate a swarm and attack together. They observe nearby units like birds in a flock. It is almost impossible to stop these swarms, unless they are intercepted by another swarm. Prototypes of such systems are currently in development.
Managing swarms requires advanced AI models capable of making decisions in real time. Rostec is attempting to develop them.
Cyber(in)security
Another key area of Russian AI development is analysis of data stolen in cyberattacks. The goals of such analysis vary, one of them being better planned military operations.
Russian hackers have always been good at stealing data. But only now have they got their hands on a tool that can analyze it.
In addition, AI is used to hack users through deepfakes and phishing schemes to extract passwords, among many other methods.
Ihor Malcheniuk, director of the Department of Cyber Protection of the Ukrainian State Service of Special Communications and Information Protection, told the Munich Cybersecurity Conference how Ukrainian military personnel are targeted on encrypted messaging platforms: they receive personalized messages designed to trick them into clicking malicious links that hack their accounts and reveal classified information.
Russian cyberattacks on Ukraine are increasingly about cyberespionage. Ukrainian cybersecurity experts have also observed growing cooperation between Russian state-backed hackers and cybercriminal groups.
It’s not just Ukraine that is under attack: UK government minister Pat McFadden said Russia is trying to use AI to enhance cyberattacks on the country’s infrastructure. UK intelligence believes AI will speed up automated reconnaissance, large-scale data analysis and the adaptation of attack strategies in real time. In other words: it will become easier to prepare and carry out cyberattacks, and their efficacy will rise.
There are concerns about AI broadly and its impact on human life. Currently, however, much more real and relevant is the intersection of AI and “natural” human cruelty, lust for power and desire to dominate others.
Perhaps the AI market in Russia appears so small because some expenditures are simply not advertised. The good news is there are not only players like Russia in the AI market, and they are making their own moves in this game.
Share this article
Read More
