On June 6, 1972, a week after President Richard Nixon’s diplomatic visit to Moscow concluded, the Soviet Union premiered the film Residence Permit. Ostensibly a drama, the movie was a meticulously crafted piece of state propaganda, narrating the tale of a Leningrad doctor who, seeking freedom and professional fulfillment in Western Europe, discovers the promised prosperity to be a fallacy. The film’s somber conclusion: his decision to depart the Soviet Union was his gravest error. To disseminate this message, the Soviet Union mobilized a vast and costly apparatus: film studios, censorship bodies, distribution networks, and cultural institutions. This infrastructure, though extensive, operated at a deliberate, measured pace. More than five decades later, Russia’s strategic objectives, in many respects, remain consistent: to project particular narratives both domestically and internationally. What has undergone a profound transformation is the cost, velocity, and sheer magnitude with which these ambitions can be pursued.
In the contemporary era, the Kremlin can achieve in mere minutes what once necessitated months of concerted effort, and it can do so across dozens of languages, through hundreds of digital platforms, and at a scale that would have been unfathomable to Soviet propagandists. The driving force behind this paradigm shift is artificial intelligence. However, this is not the AI typically envisioned in the global discourse, which often centers on a race for the most powerful computational models, the fastest microprocessors, and the most sophisticated systems. By these metrics, Russia occupies a precarious and complex position, primarily hampered by structural constraints in hardware.
Samuel Bendett, a Russia Studies Programme advisor at the CNA, a Washington D.C.-area think tank, acknowledges Russia’s considerable strength in human capital, boasting a robust pool of STEM-educated specialists and mathematicians capable of developing advanced software. "However," Bendett observes, "hardware has consistently been the Achilles’ heel, a challenge that dates back to the early days of the Cold War." This hardware deficit carries immense weight in the current AI landscape, where cutting-edge machine learning systems are critically dependent on specialized chips – Graphics Processing Units (GPUs) and AI accelerators – engineered for parallel processing of vast mathematical operations.
Presently, Russia lacks the domestic capacity to manufacture the advanced semiconductor chips essential for frontier AI development. Western sanctions, imposed in the wake of the invasion of Ukraine, have exacerbated these existing challenges. Consequently, Moscow has become reliant on smuggled components or supplies sourced from China for its more sophisticated technological endeavors. "Russia has a strong affinity for NVIDIA microchips, relying on them for military-related AI applications," Bendett states. "The same holds true for hardware like Raspberry Pi and Orange Pi. This hardware is not produced in Russia, and even if equivalent domestic manufacturing exists, it is already significantly outmoded by global standards."
Despite these hardware limitations, the Kremlin’s broader strategic objectives have not been curtailed. Instead, there has been a pronounced strategic pivot towards a different kind of operational theater – one where semiconductor shortages hold far less sway. In this domain, the paramount objective is not the construction of the most advanced AI systems, but rather the deliberate shaping of the information environment in which these systems operate. This involves influencing the data retrieved by Western AI models, controlling the information accessible to its own populace, and dictating the perceptions of audiences beyond its borders.
Projecting Influence Abroad: The AI-Powered Disinformation Machine
Sopo Gelava, a disinformation researcher with over a decade of experience and affiliated with the Atlantic Council’s Digital Forensic Research Lab since 2020, asserts that the integration of AI into Russian disinformation campaigns has intensified significantly in recent years. "Actors who once manually generated such content now exhibit far less direct human involvement," she notes, highlighting a critical shift in operational methodology. Gelava explains that even a single campaign, originating from a Russian website and subsequently propagating across multiple platforms in various languages, can display distinct indicators of AI utilization throughout its lifecycle.
"Either automation is being employed, or AI is involved in content generation," Gelava elaborates. "This hasn’t necessarily introduced a revolutionary leap in disinformation tactics, but it has dramatically enhanced scalability. It empowers creators with a significantly greater capacity to disseminate content at unprecedented speeds, reaching exceptionally large audiences. In essence, AI enables them to achieve a substantially amplified impact." These campaigns typically exhibit heightened activity in countries where Moscow has vested political interests. They often intensify in the run-up to elections but do not cease once voting concludes, with narratives continuously adapting to new events and target demographics.
In a notable recent instance, the Digital Forensic Research Lab identified a network of TikTok accounts that appeared to coordinate the dissemination of AI-generated content specifically targeting Moldova’s ruling Party of Action and Solidarity and President Maia Sandu, while simultaneously encouraging participation in protests. Gelava further explains that AI is deployed in multifaceted ways within these operations. It can automate the synchronized propagation of narratives across diverse platforms or enhance visual content to amplify emotional resonance and boost engagement metrics. The overarching objective, however, remains constant: to reach the maximum number of individuals with the greatest possible efficiency.
At the time of analysis, the identified TikTok accounts collectively commanded a following of 158,556 users, with total engagement across all interaction types exceeding 26.3 million. AI is not only facilitating the scaling of disinformation but is also intensifying cyberwarfare efforts. In April, Dutch military intelligence issued a warning that Russia is leveraging AI to accelerate cyberattacks, a threat projected to escalate. The transformative aspect extends beyond the velocity of these operations to their fundamental structure. AI is transitioning cyberattacks from labor-intensive endeavors to highly automated processes, enabling the simultaneous identification and targeting of multiple entities. Tasks that previously demanded sustained human effort can now be executed in seconds, thereby significantly expanding both the scale and reach of these operations.
Domestic Control: AI as a Tool for Information Management
The underlying logic driving Russia’s AI utilization abroad – predicated on automation, scale, and efficiency – is increasingly being replicated within its own borders. "Internal security has always been a paramount concern in Russia’s high-tech development trajectory," states Bendett. "A key priority has been to insulate the country from external influence and mitigate its impact on the domestic population." AI is now rendering this approach exponentially more potent. Where state propaganda historically relied on extensive physical infrastructure, such as studios, printing presses, and distribution networks, it can now be managed entirely through digital means. This allows the Kremlin to monitor, filter, and curate its information environment with a considerably smaller human workforce and at a vastly expanded scale.
In January, Forbes reported that Roskomnadzor, Russia’s federal body responsible for regulating and censoring telecommunications, intends to implement a machine learning-based system for internet traffic filtering within a year. According to the agency’s digitalization plan submitted to the government, a sum of 2.27 billion rubles (approximately $30 million) has been earmarked for this initiative. Media reports indicate that the system is designed to more efficiently identify and block prohibited content and to restrict access to Virtual Private Network (VPN) services, which Russian citizens utilize to circumvent censorship.
Bendett posits that the current restrictions on Telegram, broader internet blocking, and limitations on VPNs run counter to long-term strategic logic. He argues that isolating the majority of Russians from international IT applications and global messaging platforms will ultimately hinder development, given the relatively small size of Russia’s IT and high-tech sector. "Russia’s government policies, which currently aim to restrict the population’s access to some of these international components, are likely shooting themselves in the foot," Bendett contends. "This is delaying numerous projects and developments that would have otherwise unfolded if Russian developers and users had unimpeded access to Western applications, databases, and algorithms."
The surveillance architecture extends beyond the digital realm into physical spaces. Across Russian cities, street cameras equipped with AI-powered recognition systems are being deployed to monitor public areas and identify individuals in real-time. In Yekaterinburg alone, an estimated 1,000 additional cameras are slated for installation by the end of June, covering streets and public spaces. These systems continuously analyze video feeds, substantially augmenting the state’s capacity to monitor its population without necessitating a proportional increase in human personnel.
Subverting the Foundation: Steering AI Model Training Data
Russia’s operational endeavors, both domestically and internationally, are largely observable, even if challenging to counter. However, another dimension of its AI strategy is far more insidious and considerably more difficult to detect. Recent studies suggest that one of Russia’s most impactful AI strategies is not directly targeting populations but rather the very AI models that individuals increasingly rely upon to interpret and comprehend the world. This strategy discreetly targets Western AI models by influencing the data upon which they are trained and the sources from which they retrieve information.
A network of pro-Kremlin websites has reportedly utilized AI tools to inundate the internet with millions of pieces of Russian propaganda. A significant portion of this content is strategically designed to be indexed by search engines and subsequently scraped into the large datasets used for training AI systems. Researchers characterize this approach as a form of "data poisoning by scale," where the objective is not the dissemination of isolated pieces of misinformation but rather a sustained saturation of the information ecosystem. The underlying concern is that, over time, this tactic could subtly alter how AI systems interpret, prioritize, and reproduce information.
Sopo Gelava observes that Russia’s AI-driven tactics are continuously evolving and becoming more sophisticated. AI-generated content previously employed in disinformation campaigns was often relatively rudimentary and easy to identify. However, this landscape is rapidly transforming. "There were frequent grammatical errors in the past, which often indicated AI involvement in the operation," Gelava states. "Today, however, it presents disinformation creators with greater opportunities, as translations are far more refined and significantly better adapted to the local context." Researchers analyzing Russia’s use of AI concur that its parallel efforts in the global AI race are becoming increasingly covert, more scalable, and more precisely targeted. These efforts not only reach broader audiences but also pose a significant risk of shaping how information is interpreted and reproduced across digital systems worldwide.
