Free beta access. 50 free AI tokens each week. Optional top-ups available. See what is included
Opening Word document…

Russia is weaponizing AI through hybrid warfare---that undermines NATO decision-making

Franz-Stefan Gady 21, Research Fellow for Cyber, Space and Future Conflict, International Institute for Strategic Studies. “NATO Decision-Making in the Age of Big Data and Artificial Intelligence” Editors: Sonia Lucarelli; Alessandro Marrone; and Francesco Niccolò Moro. Sonia Lucarelli is Professor of International Relations and European Security at the University of Bologna, and member of the Board of Directors of the Istituto Affari Internazionali (IAI). Alessandro Marrone is Head of the Defence Programme of IAI and teaches at the Istituto Superiore di Stato Maggiore Interforze (ISSMI) of the Italian Ministry of Defence. Francesco N. Moro is Associate Professor of Political Science at the University of Bologna and Adjunct Professor of International Relations at the Johns Hopkins University Europe Campus. This publication is the result of the Conference “NATO Decision-making: promises and perils of the Big Data age”, organized by NATO Allied Command Transformation (ACT), the University of Bologna and Istituto Affari Internazionali (IAI) of Rome. https://www.iai.it/sites/default/files/978195445000.pdf //pipk

This chapter seeks to offer preliminary answers to two questions. First, to what degree will Artificial Intelligence (AI)-enabled information warfare exacerbate hybrid threats to NATO decision-making? Second, what can NATO countries do to alleviate the threat? To narrow the research scope, this chapter will principally look at threats posed by Russian AI-enabled information warfare operations under the concept of gibridnaya voyna. The study argues that given that AI-enabled information warfare has the potential to amplify societal polarization, elite disagreement within domestic politics, and reshape the perception of the “Russian threat” in NATO member countries, it can have a direct negative impact on national and NATO decision-making. Such operations, however, only amplify existing symptoms of polarization and disagreement found within NATO member states and are not their root causes. The chapter concludes that NATO must better integrate tech-centric so-called “whack-a-troll” tactics with a whole-of-nation strategy to better safe-guard NATO decision-making and alliance cohesion. Renewed discussion of whole-of-society defense concepts to inform national security strategies may be useful in this regard.

Introduction NATO defines hybrid threats as threats that “combine military and non-military as well as covert and overt means, including disinformation, cyber attacks, economic pressure, deployment of irregular armed groups and use of regular forces” (2019). Such hybrid methods are used to blur the line between war and peace and “attempt to sow doubts in the minds of target populations.” NATO further emphasizes that “[t]he speed, scale and intensity of hybrid threats have increased in recent years.” The latter can partially be traced back to various technological advances including in the fields of Artificial Intelligence (AI) and offensive cyber capabilities, utilized in conjunction with deliberate attempts by competitor nations to undermine the political cohesion of NATO member states from within.

Hybrid threats pose a number of unique challenges to political decision-making within both NATO member states and the Alliance’s various deliberative bodies, where decisions are based on the principle of consensus preceded by consultation processes (NATO, 2020). In particular, meddling by outside powers in political processes by means of influence operations, broadly defined as “organized attempts to achieve a specific effect among a target audience,” is one of the top concerns of NATO leadership (Thomas et al., 2020). In particular, Russian tactical-operational influence operations under the concept of gibridnaya voyna (‘hybrid warfare’) have been receiving a great deal of attention since 2014. As Ofer Fridman notes, such operations target society at large, seeking to undermine political cohesion in an adversary state by employing methods that amplify the divisions and polarizations among its citizens (2018).

Definitional confusion has plagued the discussion on hybrid warfare. Fridman, who sketches the conceptual evolution of the term in both the West and Russia, points out that at least three different phenomena have been described as hybrid warfare in past years with no agreed upon definition. Indeed, the term ‘hybrid warfare’ itself may at this point obfuscate rather than clarify modern conflict characteristics. As Fridman argues: “Despite the political usefulness of the term hybrid warfare, it would appear that Russian and Western military professionals now recognize that the term is next to useless for describing the real nature of contemporary conflicts, leading them to promote more specific definitions, such as information warfare, cyber warfare, which are now prevalent in the West, or new-generationwarfare, which is currently prevalent in Russia” (2018: 157).

Indeed, the Russian Chief of the General Staff, General Valery Gerasimov, has publicly aimed to draw a distinction between new-generation warfare and hybrid warfare with the former – given its emphasis on military capabilities – being the main focus of the Russian Armed Forces (Fridman, 2018).

Nonetheless, it is evident from a review of Russian military literature that hybrid warfare focused on weakening societal cohesion of an opponent still retains a prominent spot in Russian national security thinking. For example, a 2015 article on future warfare published in the widely-read journal of the Russian Ministry of Defence stated in strong terms the centrality of information to the Russian understanding of future war, asserting that “it is precisely the information-psychological struggle which will, in the main, create the preconditions for the achievement of victory [in future warfare]”. The article, authored by two frequent contributors on future warfare, continued: “…the achievement of strategic objectives in future wars will be impossible without the establishment of information dominance over the enemy. In future wars, special disinformation operations and [operations for] misleading the military-political leadership of the other side will include a system of interconnected and carefully-agreed upon large-scale measures according to the plan of new-type war (‘hybrid war’), including the use of various means of actions upon the personnel of the armed forces and population of the state with the objective of creating internal tension (schism) in society. Information-psychological operations in future war will pursue by non-forceful means the objective of achieving the significant weakening of the military potential of the enemy by means of affecting his information processes, misleading [him], demoralizing the population and the personnel of the armed forces” (Bogdanov and Chekinov, 2015: 45)

As part of such future Russian gibridnaya voyna campaigns, the effective integration of AI with cyber capabilities that enable faster and more precise weaponization of information has the potential to compromise and undermine NATO decision-making at multiple levels. This chapter consequently seeks to offer preliminary answers to two questions in this regard. First, to what degree will AI-enabled information warfare1 exacerbate hybrid threats to NATO decision-making? Second, what can NATO countries do to alleviate the threat?

To narrow the research scope, this chapter will principally look at threats posed by Russian influence operations under the concept of gibridnaya voyna.

Factors Influencing NATO Decision-Making In order to answer to the above questions, it is first necessary to outline the various influences that impact allied decision-making within NATO. According a recent RAND study, member states’ decision-making processes on their own participation in NATO operations can be broadly divided into three categories: Domestic politics; Perception of the Russian threat; Alliance dynamics (Binnendijk and Priebe, 2019).

Notably, RAND finds that domestic politics and perceptions of the Russian threat have the larger impact in decision-making processes involving an unconventional (i.e. hybrid) attack on a NATO country. In the future, such an attack would undoubtedly feature AI-enabled information warfare as part of gibridnaya voyna embedded within a new-generation warfare campaign. Consequently, this chapter will primarily focus on domestic politics and perceptions of the Russian threat to illustrate Russian tactical-operational AI-enabled information warfare.

According to the RAND study, domestic political considerations that influence decision-making regarding support or opposition to NATO military action in an unconventional environment are impacted by a number of factors, including the proximity of public elections, general public support for a particular course of action, elite agreement or disagreement (coalition governments can be particularly vulnerable in this case), and a centralized or decentralized foreign policy decision-making structure. Perceptions of the Russian threat in turn are also subject to a number of factors including perceptions of Moscow’s aims and motivations, escalation risks, vulnerabilities to Russian economic sanctions as well as military retaliations, and competing national security demands (Binnendijk and Priebe, 2019).

Most relevant for this chapter is that in both instances – domestic politics and perceptions of the Russian threat – AI-enabled information operations have the potential to influence factors in favor of Russia. For example, such operations could expand the audience reach of NATO-skeptical parties during election seasons, or in advance of a parliamentary vote in order to undermine elite consensus on anti-Russian actions. Alternatively, with the help of a carefully crafted ‘what aboutism’ narrative employed part of a wider disinformation operation, Moscow could amplify disagreement over aggressive Russian actions within a target country by emphasizing online that Russia, like any other country, is merely pursuing legitimate selfinterest.

AI-Enabled Information Warfare AI-enabled information warfare entails the use of algorithms capable of processing and learning from big data to execute attacks against specific targets autonomously or semi-autonomously in order to achieve a desired effect in the information space. 2 The terms ‘autonomously’ and ‘semi-autonomously’ describe respectively the ability of the algorithm without or with (limited) human intervention to learn from vast amounts of data in order to execute polymorphic attacks on multiple fronts in the information space (for example, simultaneous attacks on a microblogging site synchronized with emails phishing attacks), in which the algorithm is capable of changing its identifiable features (for example, by creating multiple malicious online identities fitted with unique malware packages). The most relevant media for AI-enabled information warfare are audio, text and video (Giles and Hartmann, 2020). Furthermore, AI-enabled information operations target all interrelated dimensions of the information environment – physical, informational, and cognitive/emotional. Herbert Lin and Jaclyn Kerr define the information environment as, the “aggregate of individuals, organizations, and systems that collect, process, disseminate, or act on information” (2019: 4).

AI-enabled information warfare methods may plausibly be deployed in various other ways. For example, AIenabled algorithms permit the building of realistic simulations of individuals “to test every individual’s reaction to events (both virtual and real), advertising, political campaigns, and psychological operations, and even to guess what might go viral through person-to-person interactions” (Libicki, 2017: 52). Such simulations could be created by deploying an algorithm capable of pattern recognition to study the social media behavior of a select target group of individuals in a specific country. This in turn could be used to create personalized phishing emails for social engineering attacks, but also to create so-called ‘deepfakes’ to gain access to sensitive systems or to facilitate the spreading of disinformation. Another example is the use of bots in order to ’hijack’ public perceptions: “Bots, trolls, and sock puppets can invent new ‘facts’ out of thin air leading to a polarized society and a culture of mistrust” (Wright, 2019: 318). The main objective behind creating alternate facts would not be to create a convincing alternative narrative – which is an objective of strategic communications – but rather to erode the solidarity of groups facing the same threat by playing on the fears and anxieties of individuals (Libicki, 2017).

In sum, AI-enabled information warfare differs from ‘regular’ information warfare in four crucial areas: 1. Speed: AI can accelerate the pace of operations by, for example, faster identifying targets due to the expedited operationalization of cyber intelligence, surveillance and reconnaissance (ISR) data as a result of increased pattern recognition capabilities of algorithms. 2. Scope: AI can expand the scope of information warfare operations by semi-autonomously or autonomously executing polymorphic attacks on multiple platforms with the help of bots, trolls, or sock puppets. 3. Scale: AI can increase the chances of a particular message or narrative going ‘viral’ in the information space by, for example, self-learning algorithms ‘wargaming’ the potential impact of specific content online, or AI-enabled facial recognition software that is capable of recognizing emotional states of individual humans. 4. Sophistication: AI can not only increase the sophistication of microtargeting processes, but also facilitate the creation of ever more convincing synthetic media products (such as ‘deepfakes’) with the support of Generative Adversarial Networks (GANs) where it is increasingly difficult to separate real from synthetic identities (Rocca, 2019).

Since AI is both scalable and effective, it is likely to be of significant utility to those seeking to undermine Western democracies because it will permit adversaries to avoid making tradeoffs between scale and effectiveness. In particular, AI will likely allow adversaries to expand exponentially the scale and rate of attacks, while also increasing the number of targets (Rocca, 2019). In addition, the following factors are all likely to enhance the utility of AI in information warfare: societies’ growing dependence on cyberspace as a news source; citizens’ increased reliance on social media; the difficulties individuals face in distinguishing between fake and genuine news sources; the unprecedented levels of access to information; and the speed at which information can be spread between individuals (Giles and Hartmann, 2020).

AI, however, is no panacea for influence operations. The impact of AI-enabled information warfare will largely depend on the quality of input data on the one hand, and continue to be contingent on analogue factors that may be too difficult for algorithms to effectively process on the other (for instance, various cultural and historical factors that may impact a society’s reaction to a particular event).

Furthermore, an overreliance on AI-enabled technology, in particular for conducting information warfare operations in cyberspace, may be misguided given the uncertainty surrounding the degree to which AI will favor the offense or defense (Dafoe and Garfinkel, 2019b). Building on the literature analyzing ‘regular’ or conventional cyber attack and defense operations (given that AI-enabled information warfare operations will mainly be conducted through this domain), broad generalizations on the subject are misplaced. As one analysis regarding the offense-defense balance in cyberspace puts it: “Sweeping claims about the offensedefense balance in cyberspace are misguided because the balance can be assessed only with respect to specific organizational skills and technologies” (Slayton, 2016: 72).

Nevertheless, there is some evidence that certain methods appear to favor the offensive. As a digital forensics expert notes regarding deepfake technology, “the adversary will always win, you will always be able to create a compelling fake image, or video, but the ability to do that if we are successful on the forensics side is going to take more time, more effort, more skill and more risk” (Giles and Hartmann, 2020: 236). Another analysis notes that scaling effects will likely be an important factor in determining the offensedefense balance (Dafoe and Garfinkel, 2019a). Nonetheless, it is unclear whether, for example, increased automation alone will ultimately favor the attacker or defender. Given that AI-enabled information warfare will be carried out principally through cyberspace, the uncertainty surrounding presumed technical solutions should therefore be kept in mind by policymakers.

If no discernible offensive advantage can be achieved, there is the risk of an information war of attrition with no clear winner but with excessive collateral damage sustained by society overall. The series of Chinesebacked Covid-19 disinformation campaigns executed in 2020 are a good example of this, having inflicted lasting damage on the information spaces of a number of Western democracies despite the adoption of successful countermeasures by governments (Ha & Cho, 2020). As these campaigns have shown, an information warfare stalemate would likely be more harmful to democratic societies that assume the presence of free speech as a norm than to authoritarian societies where the flow of information is under tighter control, and where consequently a disruption of the information space will have fewer consequences.

Sign in to recut, update, or save this card.