Russia uses AI for cyber warfare

On June 6, 1972, just over a week after Richard Nixon’s visit to Moscow ended, the film Residence Permit premiered in the Soviet Union. On its surface, it was an ordinary drama. In reality, it was a finely engineered piece of state propaganda—a story about a doctor from Leningrad who chooses to remain in Western Europe in pursuit of freedom and professional success, only to discover that stories about Western prosperity are a myth. His decision to leave the Soviet Union, the film concludes, was the greatest mistake of his life.

To spread that message, Moscow needed an entire machinery: studios, censors, distribution networks, cultural institutions. The infrastructure for propaganda was vast, expensive, and slow. More than 50 years later, Russia’s goals remain largely unchanged. It still seeks to project similar messages both domestically and abroad. What has shifted, however, is the cost, speed and scale of pursuing those ambitions.

Today, the Kremlin can do in minutes what once took months, and it can do it across dozens of languages, on hundreds of platforms, at a scale no Soviet propagandist could have imagined. The reason is artificial intelligence. But not in the way the term is usually understood. When analysts speak of an AI race, they typically mean a contest over who builds the most powerful models, the fastest chips and the most capable systems. By that measure, Russia occupies a complicated position. It faces a significant structural constraint: hardware.

Related: Finance and Geopolitics Collide

Hardware constraints and domestic surveillance

Samuel Bendett, an adviser to the Russia Studies Programme at CNA, a Washington, DC–area think tank, says that Russia has a strong pool of talent—STEM-educated specialists and mathematicians capable of developing advanced software. “But hardware has always been the weakness, and this goes back to the early days of the Cold War,” he adds.

That weakness matters enormously in the current AI environment. Cutting-edge machine learning systems depend on specialised chips—graphics processing units (GPUs) and AI accelerators—capable of performing vast numbers of mathematical operations simultaneously. Currently, Russia cannot produce the advanced chips needed for frontier AI. Western sanctions following the invasion of Ukraine have created even more problems. As a result, Moscow relies on smuggled or Chinese-sourced components for more sophisticated systems.

“Russia loves NVIDIA microchips and depends on them for military-related AI applications. The same can be said about hardware such as Raspberry Pi and Orange Pi. That hardware is not produced in Russia or, if its equivalents are actually manufactured domestically, they are already outdated compared to global standards,” Bendett says.

But hardware constraints have done little to curb the Kremlin’s broader ambitions. Instead, it has shifted focus to a different kind of battlefield—one where semiconductor shortages matter far less. In this space, the priority is not building the most advanced systems, but shaping the environment in which they operate: influencing what Western AI models retrieve, controlling what its own citizens see at home and dictating what people beyond the borders believe.

Related: FSSAI halts two food firms over hygiene lapses

Internal security has always been at the forefront of Russian high-tech development in general. A key priority has been how to insulate the country from external influence and limit its impact on the domestic population. AI is now making that approach dramatically more powerful. Where state propaganda once depended on extensive physical infrastructure—studios, printing presses, distribution networks—it can now be managed digitally. This allows the Kremlin to monitor, filter, and shape its information environment with far fewer people and at far greater scale.

In January, the outlet reported that Roskomnadzor—Russia’s federal body for regulating and censoring telecommunications—plans to deploy a machine learning-based system for filtering internet traffic within a year. According to the agency’s digitalisation plan submitted to the government, 2.27bn rubles ($30m) has been allocated to the initiative. The system aims to identify and block prohibited content more efficiently and restrict access to VPN services that Russian citizens use to circumvent censorship.

Cross the Russian Federation, street cameras embedded with AI-powered recognition systems are being used to monitor public spaces and identify individuals in real time. In Yekaterinburg alone, around 1,000 additional cameras are expected to be installed by the end of June, covering streets and public areas. The systems analyse video feeds continuously, significantly expanding the state’s capacity to monitor its population without requiring a proportional expansion of human personnel.

Manipulating the models themselves

Researchers studying Russia’s use of AI believe its parallel efforts in the global AI race are becoming harder to detect, more scalable and more targeted. They not only reach wider audiences but also risk shaping how information is interpreted and reproduced across digital systems.

Related: Tata unveils electric Sierra with strong range

Recent studies suggest that one of Russia’s most consequential AI strategies is aimed not directly at populations, but at the models they increasingly rely on to interpret and understand the world. This strategy targets Western AI models indirectly by shaping the data they are trained on and the sources they retrieve information from.

A network of pro-Kremlin websites has reportedly used AI tools to flood the internet with millions of pieces of Russian propaganda. Much of this content is designed to be picked up by search engines and scraped into large datasets used to train AI systems. Researchers describe this approach as a form of data poisoning by scale, where the aim is not a single piece of misinformation, but a sustained saturation of the information ecosystem. The concern is that, over time, this could subtly shape how AI systems interpret, prioritise and reproduce information.

Sopo Gelava notes that Russia’s AI-driven tactics are becoming more sophisticated over time. AI-generated content used in disinformation campaigns was once relatively easy to spot. But that is changing quickly. “There used to be frequent grammatical errors, and in the past we could often tell from this that the operation had been created by AI. Today, however, it gives more opportunities to creators of disinformation because the translation is much more refined and significantly better adapted to the local context,” Gelava says.