Russia’s approach to military AI prioritizes technologies and capabilities that can be used to debilitate the adversary’s command, control and communications systems, as well as gain information superiority.
Since his appointment as defence minister in 2012, Sergey Shoigu, in close coordination with President Vladimir Putin, has successfully shepherded the Russian military modernization effort, with notable accomplishments in ‘the procurement of modern radars, communications equipment, electronic warfare [EW] systems, robotics, unmanned aerial vehicles [UAVs], and high-precision strike assets’. Now, as Shoigu recently proclaimed, ‘it is necessary to ensure the introduction of artificial intelligence [AI] technologies in weapons that determine the future appearance of the Armed Forces’.
The Russian defence establishment is quite enthusiastic about the potential of AI. Over the past few years, the defence ministry has set up a network of research and development (R&D) organizations spanning the military-industrial complex, academia and the private sector to continue work on military robotics and the integration of AI into military systems. Moreover, over 600 new weapons and other items of military equipment have been tested in combat conditions in Syria, with 200 of these items being described as ‘next-generation’.
Western national security decision-makers would be unwise to discount Russia’s potential to use AI-enabled technologies in ways that undermine US and NATO interests.
Western observers, however, are generally sceptical of Russia’s potential in emerging technologies. Russia spends much less on R&D than either the US or China, in terms of both overall value and share of GDP. The private-sector AI ecosystem is relatively small, and there are real problems with talent development and retention. Russia’s AI-related research also lags behind the US and China. For example, between 2010 and 2018, compared to Russian researchers, US researchers published 58 times the number of papers indexed under machine learning (ML) and 42 times the number exploring computer vision. Russia’s micro-electronics industry – the hardware on which all AI runs – is nascent, and despite recent efforts to implement an import substitution strategy, the country’s civilian technology sector is still heavily reliant on semiconductor equipment from the US, Taiwan and South Korea.
These challenges notwithstanding, Western national security decision-makers would be unwise to discount Russia’s potential to use AI-enabled technologies in ways that undermine US and NATO interests. Russian strategists see remotely operated, automatic, autonomous, and AI-enabled technologies as augmenting both traditional and more recently developed advantages in intelligence, surveillance, and reconnaissance (ISR) capabilities, EW, cyber warfare, information operations, and ground-based fires. But how these new technologies may be employed is equally important, with a high priority being placed on disrupting and destroying the adversary’s command-and-control systems and communication capabilities, and a focus on non-military means to establish information superiority during the initial period of war, expanding far into peacetime.
The US and its NATO allies have taken some steps to counter Russia’s more advanced systems and capabilities, including enhancing their EW capabilities; modernizing and hardening command, control, and communications infrastructure; and developing technologies to counter unmanned aerial systems. But the strategic thinking behind these solutions and the operational concepts guiding their potential use could be improved by contextualizing Russian AI-enabled technologies and capabilities within the broader framework of Russia’s way of war.
The chapter proceeds in three parts. The first section reviews the key guiding concepts and principles in Russia’s way of war and the role played therein by emerging technologies and capabilities. The second section covers the key areas for AI and ML investments, focusing specifically on EW, unmanned systems and information warfare. The last section assesses the implications Russian advances in military AI could have for the US and NATO.
Guiding concepts and principles
Russia considers itself both a great power and a nation that is embroiled in an asymmetric competition with more powerful great powers, specifically the US and NATO. The Russian General Staff has therefore repeatedly expressed interest in developing technologies and capabilities that can serve as force multipliers and be employed as asymmetric responses against high-tech adversaries.
While Russia’s military posture is primarily defensive, in the event of a major conflict Russian forces aim to disorient and disrupt the adversary, preventing it from operating in its preferred fashion and slowing its ability to respond to developments on the battlefield. Russia’s emphasis on deception, EW and strikes against command and control, as well as layered air defences and ground-based fires, all play into the broader operational and tactical conception of a disjointed battle. The current focus on the development of UAVs, ISR capabilities, and EW that combine to make the battlefield more visible and controllable and allow Russian forces to mass fires quickly and effectively fits these objectives.
Beyond making the armed forces more mobile, modern, and efficient, Russia’s investments in new technologies also aim to enable a successful confrontation via non-military means during crisis, establishing information superiority over the adversary during the initial period of war. Technological developments have now made it possible to deploy cyber and information tools into foreign systems and societies in peacetime – to conduct reconnaissance, plant viruses and execute wide-scale, targeted information operations.
In this context, the integration of ML techniques into cyber operations could potentially augment existing Russian strengths, enhancing the country’s ability to influence and manipulate potential opponents, undercut democratic institutions, disrupt and disable critical infrastructure, and stir chaos and discord along an array of political and societal vectors. Such efforts can significantly undermine Russia’s potential adversaries’ abilities to organize, mobilize, command and conduct military operations, and, from the Russian perspective, are inherently intertwined with the more conventional aspects of war.
Key areas for AI and machine-learning investments
Russian leaders and military strategists identify a broad range of areas for the employment of AI, including command and control, robotic systems, EW, cyberspace and information warfare, military logistics, training, health and medicine, and forecasting. While the Western – and especially American – culture of innovation is marked by high levels of tech optimism and a tendency to view and use cutting-edge technologies as a solution to tactical and strategic problems, Russia’s approach to military applications of AI is more utilitarian and pragmatic. AI technologies are largely discussed as enablers and amplifiers of established, albeit continuously evolving, means and methods of warfare – reflecting an evolutionary rather than revolutionary approach to emerging technologies.
The discussion below elaborates on three key areas where investments in AI fit into the broader framework of Russia’s way of war and could have implications for the US and NATO.
Since 2009, Russia has made significant strides in developing its EW capabilities, with investments in this area representing a critical aspect of a broader effort to implement the Ministry of Defence’s network-centric warfare vision through C4ISR (Command, Control, Communications, Computers, Intelligence, Surveillance and Reconnaissance) integration. Currently, Russia has a range of highly mobile EW systems in its arsenal, and at least one has been publicly discussed as having AI capabilities.
According to news reports, the RB-109A Bylina EW system is an automated decision-support system, capable of independently identifying and selecting targets such as radio stations, communication systems, radars, satellites, and other facilities, and deciding how to suppress them and what jamming stations to use. The system then issues the relevant sequence of orders by automatically interfacing with battalion and company command posts and individual radio-electronic warfare (REB) stations, conducting its operations without interfering with friendly REB stations.
In the summer of 2018, the Bylina, as well as three other distinct EW systems, were spotted in the Donbas region of Ukraine, where they were presumed to be providing ‘valuable information and experience to the Russian Armed Forces for future conflicts’. In April 2020, Izvestiya reported that the defence ministry had approved plans to deliver Bylina systems to military units by 2025. According to experts cited in the newspaper, the systems could ‘increase the efficiency of EW by 40%–50%’.
Russia’s investments in EW capabilities seek to take advantage of the fact that most US and NATO military systems and weapons are hooked to satellite communications, Global Positioning System (GPS) navigation, and high-bandwidth internet. The integration of AI into EW systems could enhance Russia’s already notable capabilities in this area, providing the ability to make faster decisions while simultaneously suppressing the opponent’s decision-making abilities.
For example, integrating AI into EW systems could improve EW effectiveness by more accurately classifying signals, helping translate massive amounts of data into actionable intelligence, concentrating attention on the most important signals, and developing a clear sense of the electromagnetic environment and how it looks from a friendly, neutral, or adversarial perspective. Some experts, however, are highly sceptical of Russia’s ability to develop modern AI algorithms and do not foresee ‘sudden, significant improvement in AI-enabled EW from Russia’ that would provide its forces with an overwhelming advantage.
As the smart software behind autonomous physical systems, AI has the potential to increase the speed, persistence, reach and endurance of unmanned systems in the air, on the ground, under water and in space, as well as to enhance coordination in both human–machine teams and between multiple unmanned systems. Technological breakthroughs that enable a shift from remotely operated unmanned systems to AI-enabled autonomous systems could also allow for a broader range of missions in denied and hostile environments, all while minimizing the risk to military personnel. While the topic of unmanned systems is discussed at length in Chapter Five of this paper, the section below summarizes some of the Russian military thinking and advances in AI in unmanned systems.
Outside of experimental prototypes, all of Russia’s current UAVs and unmanned ground vehicles (UGVs) are remotely operated. Russian strategists, however, anticipate that AI will play an increasingly larger role in air combat platforms, which may lead to the development of fully autonomous combat systems. Some even foresee the greater robotization of war, and future warfare involving more machines and ‘not soldiers shooting at each other on the battlefield’. But in the near future, there is a greater emphasis on human–machine teaming and ‘the rational combination of the capacities of soldiers and military hardware’. Russia’s approach to robotization predominantly entails ‘grafting robotic capabilities to existing platforms’, rather than ‘trying to develop completely new systems’.
Thus far, Russia’s UAV development, especially heavy combat UAVs, has been much slower than Moscow wanted, lagging behind that of both China and the US.
The Altius long-range drone, for example, has been in development since 2011, with the most recent variation promising to be equipped with AI elements for command and control as well as increased autonomy for navigation, target identification, and potentially also target engagement. The S-70 Okhotnik-B heavy combat UAV, meanwhile, could be delivered to Russian forces in 2024. According to defence ministry reports, the Okhotnik recently flew in automated mode for the first time and practiced interoperability with an Su-57 lead aircraft. Ultimately, the goal is to deploy Su-57 pilots alongside these combat drones in swarms enabled by AI.
As the US and NATO invest in ways to counter unmanned aircraft systems and capabilities, attention should be paid to technologies and skill sets for detecting tactical drones that have small infrared and electromagnetic signatures and can evade current air tracking systems.
Assets such as the heavy strike drone Okhotnik are particularly relevant for large-scale operations against peer adversaries, where they can be tasked with suppressing long-range air-defence systems, hitting targets deep inside enemy territory, and providing cover to manned aircraft from ground-based fires. It is therefore notable that some Western and Russian analysts doubt Russia’s ability to deliver on these large-scale combat drones, predicting that Russia’s ‘developers of UAVs will continue to focus on reducing the radar signature of UAVs, their further miniaturization, lower prices, increased autonomy, reliability and accuracy of output to the target’. With this in mind, as the US and NATO invest in ways to counter unmanned aircraft systems (UAS) and capabilities, attention should be paid to technologies and skill sets for detecting tactical drones that have small infrared and electromagnetic signatures and can evade current air tracking systems.
Research, development, testing, and prototyping of UGVs has climbed higher on the defence ministry’s priority list over the past few years, and a broad range of systems have advanced through the technology development cycle and been used for ISR, demining, breaching operations, and other combat support tasks. While attending the Army 2020 International Military-Technical Forum, for instance, Defence Minister Shoigu described seeing ‘robots that are installed on heavy hardware, robots that can do mine clearance’, and robots that ‘effectively have neural networks of control, artificial intelligence elements’. This technology, according to Shoigu, can pose ‘a serious threat and represents a serious weapon today’.
Some of Russia’s UGV projects, like the unmanned version of the T-14 Armata tank and the Soratnik mid-sized combat UGV, end up as demonstrators of advanced robotics technologies, and while these particular prototypes are not mass-produced, the technical data and testing results inform their redevelopment or even their redesign in new roles. Other systems, like the Nerekhta and Marker UGVs, serve as a test bed for AI applications. The Advanced Research Fund, for example, has used the Nerekhta UGV to test collaborative behaviour with other UGVs or UAVs, while the Marker UGV served to test computer vision, autonomous movement and navigation, and swarming technologies.
While progress in AI applications for UGVs, and particularly the larger combat UGVs, has seen some setbacks, Russia is not alone in this. US, Chinese, and many other scientists and developers have also had to contend with the challenges of autonomous ground navigation, mobility in complex terrain, communication in a contested electromagnetic spectrum, and coordination with humans and other unmanned systems. The US military, for example, has had a number of programmes dedicated to the development of remotely-controlled and semi-autonomous robotic vehicles meant to provide logistical or fire support to dismounted soldiers, which ultimately failed to progress beyond the testing and experimentation phase.
The Russian military, however, tends to move faster than the US military when it comes to testing new technologies in operational conditions – arguably exhibiting a higher risk tolerance in the event of accidents or failures, or possibly a lower regard for what is ethical or permissible under the laws of war. In 2018, for example, the Russian military sent the Uran-9 – an armoured UGV the size of a small tank – for its first ‘near-urban combat’ mission in Syria, where it encountered some problems, including repeated communications outages and failing to fire effectively while on the move. These challenges and delays in delivering the promised UGVs to the force have led some analysts to conclude that such systems can only be used in a limited capacity in combat, primarily for ISR missions. Such assessments fall far behind Russia’s ambitions to deploy these systems as part of combined arms formations, in manned-unmanned operations, and with increasingly autonomous functions.
Information superiority and cyber warfare
Russian strategists see information warfare as a central tenet of contemporary conflicts, and while thinking and approaches to information warfare are continuously evolving, there is a general consensus that information superiority could play a key role in the outcome of wars. Some scholars argue that Russian strategists have come to view AI-enabled information warfare as a ‘strategic war-winning asset in peer-state conflicts’. According to this view, as militaries leverage AI to ‘exponentially increase the power of information warfare’, AI will usher in ‘the third revolution in military affairs’. Former deputy minister of defence Yury Borisov has articulated a somewhat less revolutionary view, stating that the development of AI will allow Russia to more effectively contest the information environment and win cyber wars.
While it is well known that Russia relies on cyber warfare to advance and support its military, political, and strategic objectives, it remains to be seen exactly how the integration of AI, and more specifically, ML-based automation could augment existing Russian capabilities in cyber warfare. Generally speaking, as a recent report from the Center for Security and Emerging Technology’s CyberAI project explains, ‘machine learning could improve discovery of the software vulnerabilities that enable cyber operations, grow the effectiveness of spearphishing emails that deliver malicious code, increase the stealthiness of cyber operations, and enable malicious code to function more independently of human operators’.
In December 2015, for example, Russian attackers executed the first known cyberattack on an electric grid, hitting three power companies in Ukraine. Although the attackers used automated systems to conduct reconnaissance within the network and delete data, the attack was ‘decidedly manual’, and ‘each manual attack at each substation required a distinct human operator’. During a 2016 attack on Ukraine’s power grid, however, the malicious code CRASHOVERRIDE could automatically find circuit breaker controls, switching them on and off and creating a blackout. This attack offers insights into the increasing role of automation in offensive cyber operations and its significance expands beyond Ukraine, as ‘the creators of CRASHOVERRIDE had developed an automated weapon that they can easily adapt for electrical grids all over the world, and that they could use, in theory, to generate blackouts at the flip of a switch’.
Looking ahead, developments in ML have the potential to make cyber operations more efficient, far-reaching, and widespread, while shielding the identity of the perpetrators and making it more difficult to defend against incursions. AI can be used to automate, accelerate and scale synthetic accounts and content, or, as one senior adviser to the Russia military has put it, to ‘supplement the information space with a large volume of artificially created data’ and ‘virtual truth’. In this sense, technological advances in AI have the potential to ‘hyperpower Russia’s use of disinformation’. For the US and NATO, technical solutions buttressing cyber defence are necessary, but are also unlikely to be sufficient in countering the broader effects of Russian information warfare, which expand beyond the cyber realm, threatening to erode public trust in democratic institutions and deepen social divisions.
Implications for the US and NATO
The strategic shift from counterinsurgency operations in the Middle East to the new era of great power competition against China and Russia marks the end of operations in permissive environments where US and allied forces have enjoyed essentially uncontested technological superiority and freedom of manoeuvre across the different domains and the electromagnetic spectrum. The rising popularity of the Joint All-Domain Command and Control (JADC2) concept across the US Department of Defense over the past three years reflects this reality, as do the investments in enhancing EW capabilities and modernizing command, control and communication networks.
This multi-domain approach, the push toward connectivity between sensors and shooters from all of the military services, and the changes to command and control are meant to offset the sophisticated anti-access/area denial capabilities of potential adversaries, including Russia’s more advanced systems. But closer attention to how Russia envisages using some of these new technologies could help reveal gaps in current thinking and potentially improve the overall strategy for dealing with a potential Russian threat.
For example, Russia sees EW as an asymmetric way to counter high-tech opponents. This suggests that EW could be integrated with deception and reflexive control techniques not only for a greater but for a qualitatively different tactical, operational and psychological impact from that which EW can accomplish on its own. A Russian EW contingent using AI-enabled systems to debilitate the opponent’s frequency and communications capabilities, disorganizing the command and control, could also alter the correlation of forces on the battlefield. In this sense, tactical applications can have operational, if not strategic, effects.
Russia has repeatedly demonstrated innovation in drone tactics – using tactical drones to provide near-real-time intelligence and targeting information for supporting artillery units.
The US and some of its NATO allies have also taken steps to deal with the growing threat from UAS technologies, including investments in counter-UAS capabilities. Here, it is worth noting that Russia has repeatedly demonstrated innovation in drone tactics – using tactical drones to provide near-real-time intelligence and targeting information for supporting artillery units. In July 2014, Russian forces used this technique to destroy four Ukrainian army brigades. While passive air defence measures could have minimized the damage from such an attack, continued evolution and innovation of tactical drone tactics, including AI-enabled swarming capabilities, will require an equal if not greater degree of adaptability and ingenuity on behalf of the US and NATO.
Now, because AI is a new technology that may fail when confronted with tasks and environments different from those for which it was trained, Russia’s experience testing its high-tech weapons in near-operational and combat conditions in Syria and Ukraine is also consequential. Even the struggling Uran-9 provided key insights into capabilities, limitations and changes needed before this system can be integrated into the force. Data collected during experiments and testing in near-operational conditions can prove imperative for training new AI algorithms, while the experience Russian soldiers now have in operating and working alongside these advanced systems is important for future advances in human–machine teaming.
Finally, the US and some NATO allies are also investing in R&D related to counter-autonomy and counter-AI technologies, including as part of efforts to buttress their cybersecurity measures. The amalgamation of technical and psychological elements in Russia’s approach to information warfare, however, presents a challenge that extends beyond what technological solutions alone can solve.
As a whole, there are good reasons to question Russia’s ability to develop the modern AI algorithms that fuel sophisticated EW systems, to scale prototypes of heavy combat drones and unnamed ground combat vehicles, or to use AI-enabled information and cyber warfare to cause strategic effects. Russia, however, does not need to be an AI superpower to successfully employ AI-based technologies and capabilities against US and NATO interests.
A holistic understanding of Russian military AI developments must pay close attention to how AI-enabled systems fit within broader Russian military thinking about the strategic, operational, and tactical approaches to modern warfare. The emphasis on how new technologies may be used is therefore equally consequential to the technological advances themselves.