By David Colon – Revue Défense Nationale (2024)
Translate By SAKHRI Mohamed
In 1952, Soviet Colonel Anatoly Kitov, an engineer by training, was assigned to explore potential military applications of computers. In a secret library, he found the sole Soviet copy of Cybernetics, published in 1948 by American mathematician Norbert Wiener, a book otherwise removed from Soviet libraries. Unlike Kremlin authorities, who deemed cybernetics a “bourgeois pseudoscience,” Kitov realized that this new science of control mechanisms and feedback opened enormous possibilities for machine learning and automation. In 1955, he co-authored an article with esteemed Russian mathematicians Sergei Sobolev and Aleksei Lyapunov, initiating the slow process of rehabilitating cybernetics. The authors envisioned it both as a “theory of high-speed electronic calculating machines” and as a “theory of self-organizing logical processes.”
In June 1960, as de-Stalinization progressed, Norbert Wiener, whose major works were now translated into Russian, was warmly welcomed in the USSR for an extended visit to Moscow, Saint Petersburg, and Kyiv. Wiener’s cybernetics inspired Soviet scientists, officials, and military leaders, including those who attempted to establish a national computer network, the OGAS (Automated National System of Calculation and Information Processing), beginning in 1962. Others sought, in the early 1980s, to develop a system for automated telephone conversation analysis for the KGB’s 12th Division.
Wiener’s theories most strongly influenced the realm of information. Russian information warfare strategists highly valued what Wiener termed the “combat use of information,” where, he wrote, “it is just as crucial to keep one’s own information channels open as to obstruct those of the adversary.” From a cybernetic perspective, this goal meant promoting entropy—disorder—within the adversary’s informational space, and conversely, negentropy—increased order—in their own. To stabilize Russia’s digital information space, the FSB (Federal Security Service) implemented the Operational Search and Measurement System (SORM) in 1995, a system requiring Russian internet providers to install telecommunications intercepts at their expense. In 1999, FSB Director Vladimir Putin expanded SORM across the Russian internet.
Toward Automated Online Content Regulation
With the advent of Web 2.0, or “participatory web,” the FSB focused on controlling online discussions, initially through human “trolls,” and later using computer programs. In 2012, the Duma passed a law instituting internet filtering through Deep Packet Inspection (DPI) technology, enabling automatic removal or modification of specific content. The 2016 Yarovaya Law, nicknamed the “Big Brother Law,” required internet service providers, social media, and messaging services operating in Russia to retain user data for three years and allow authorities access to encrypted communications.
Kremlin operators sought not only to remove or alter content or slow its spread but to drown critical content in comments automatically generated by bots—software agents that interact automatically or semi-automatically with servers to generate connections, create fake accounts, amplify content through likes and retweets, and disseminate messages. Leaks from Kremlin documents indicate plans to use AI to detect banned content and maintain a minimum of 70% pro-Russian comments on the Telegram channel “ChP Donetsk.” The cybernetic approach to automated content regulation aims to maintain stability in a regime-favorable information environment, applying the physiological principle of homeostasis—organism processes that maintain a stable state, here applied to social stability, despite external disturbances. In recent years, other authoritarian regimes, including China’s, have adopted this approach, investing heavily in automated online content control systems to replace their armies of trolls.
The determination of authoritarian regimes to control the internet has increased significantly with the rise of smartphones and social media, which have empowered oppositional voices. The 2009 Iranian and Moldovan “Twitter Revolutions,” the “2.0 Revolutions” of the Arab Spring, and the 2011 Bolotnaya Square protests organized via Facebook and VKontakte led affected regimes to engage on social media, viewing it as a vital threat to their political survival.
AI in the Kremlin’s Networked Warfare
In Russia, Vladimir Putin perceives that the U.S. uses social media to destabilize his rule and engages the Kremlin in “networked warfare” against the U.S. and its allies. In 2012, Putin appointed former KGB officer Igor Shchegolev as his internet advisor and tasked the state investment fund Rusnano and the Skolkovo research center with heavy AI and nanotechnology investments in Silicon Valley. Concurrently, the Kremlin contracted a dozen Russian “political technology” firms to initiate information manipulation campaigns in the West. The most famous, albeit less discreet, is oligarch Yevgeny Prigozhin’s Concord company, which established the Internet Research Agency (IRA) in Saint Petersburg, a “clandestine information factory.”
Kremlin operators, whether “polit-technologists” or intelligence services, sought to appropriate American technology. The GRU and FSB pursued the behavioral microtargeting tech from Cambridge Analytica, a U.S. firm that used deep learning AI to infer personality traits from Facebook user behavior. Russian intelligence is especially interested in targeting personalities of the “dark triad”—traits that make individuals susceptible to conspiratorial thinking and violence. This psychographic targeting multiplies the effects of online harassment (trolling). Cambridge Analytica’s lead engineer, Moldovan-born Aleksandr Kogan, was invited to present his work in Saint Petersburg and later contracted by the Russian state to aid a psychology research team at Saint Petersburg State University. The GRU acquired the American Facebook user data he collected, encompassing millions of profiles.
The Kremlin has also closely followed Michal Kosinski, an early researcher in social media-based predictive personality analysis who refused to work with Cambridge Analytica. On July 14, 2017, Kosinski was flown to Moscow to present his research to Prime Minister Dmitri Medvedev, Foreign Minister Sergei Lavrov, and other officials, impressing Kosinski with their knowledge of his work. Later that year, Putin underscored the significance of AI in national security, declaring, “Whoever becomes the leader in AI will dominate the world.”
China’s Militarization of AI
At that time, China shared Putin’s conviction. President Xi Jinping prioritized information technology among his “Four Modernizations,” investing heavily in Silicon Valley to acquire minority stakes in high-tech U.S. firms. In information warfare, China often draws on Russia’s model but has the advantage of its advanced technology, owning its own equipment and platforms. The 2017 launch of TikTok marked a major step in implementing the “Three Warfares” doctrine (psychological, media, and legal) introduced in 2003. Using AI, TikTok captures user attention, collects data for behavioral targeting, and deploys influencers to shape narratives on a massive scale, advancing China’s influence at minimal cost.
Since 2021, the release of generative AI tools has exponentially increased capabilities for digital propagandists. In June 2024, Google released the first empirical study on misuse of these tools, highlighting that easily accessible generative AI can be exploited for minimal technical expertise. Manipulation of human images and evidence falsification are among the most common tactics, with generative AI also used to manipulate opinion, creating armies of fake accounts, scaling automation, and distributing content.
From Generative to Degenerative AI
Kremlin operatives now leverage generative AI extensively. FBI Director Christopher Wray highlighted, on September 4, 2024, Russia’s dependence on AI in influence operations. In March 2024, the disinformation network CopyCop was identified for deploying disinformation on a large scale, using generative AI to plagiarize, translate, and modify mainstream media content, tailoring it for specific audiences with engineered biases. CopyCop also generates deepfakes and garners significant engagement with targeted, human-generated content. This network amplifies its content through two other Russian disinformation networks, Doppelgänger and Portal Kombat.
The impact of generative AI is evident. In June 2024, the American start-up NewsGuard reported that the top 10 generative AI models reproduced Russian disinformation narratives in 32% of cases, citing fake news sites created in Moscow as authoritative sources. As generative AI systems become increasingly polluted with AI-generated disinformation, they produce more defective outputs. Kremlin operators have thus created an entropic feedback loop, degrading the ability to distinguish truth from falsehood, eroding public trust in information, and depriving citizens of what French historian Marc Bloch termed the “minimum of clear, reliable information essential for rational conduct.”
Generative AI has thus become a powerful tool for Russia’s information warfare doctrine of “reflexive control,” an influence strategy targeting adversaries’ perceptions of their environment. AI pioneer Geoffrey Hinton, who resigned from Google citing generative AI risks, described a “nightmare” scenario: an assertive power like Russia gaining access to improved AI technologies. To avoid this, we must require generative AI platforms to allocate revenue toward security and risk assessment measures and leverage AI to counter adversarial information manipulation.
References
- Sobolev Sergei L., Kitov Anatolii I. et Lyapunov Aleksei A., « Osnovnye cherty kibernetiki » [Caractéristiques de base de la cybernétique], Voprosy filosofii, n° 4, 1955, p. 141.
- [2]Wiener Norbert, Cybernétique et société. L’usage humain des êtres humains, Paris, Seuil, 2014 (1950), p. 156
- [3]Roonemaa Holger, Vunsh Marta, Morozova Anastasiia, Huppertz Carina et Carlsson Mattias, « Kremlin Leaks : Secret Files Reveal How Putin Pre-Rigged his Reelection », VSquare, 26 février 2024 (https://vsquare.org/).
- [4]Lewis Paul, « ‘I was shocked it was so easy’ : meet the professor who says facial recognition can tell if you’re gay », The Guardian, 7 juillet 2018 (https://www.theguardian.com/).
- [5]marchal Nahema, Xu Rachel, Elasmar Rasmi, Gabriel Lason, Goldberg Beth et Isaac William, « Generative AI Misuse : A Taxonomy of Tactics and Insights from Real-World Data », Google Deepmind, 5 juin 2024.
- [6]US Department of Justice, « Justice Department Disrupts Covert Russian Government-Sponsored Foreign Malign Influence Operation Targeting Audiences in the United States and Elsewhere », 4 septembre 2024 (https://www.justice.gov/).
- [7]Sadeghi McKenzie, « Les 10 principaux modèles d’IA générative reprennent les récits de désinformation russe un tiers du temps, en citant comme des sources faisant autorité de faux sites d’information locaux créés à Moscou », NewsGuard, 18 juin 2024 (https://www.newsguardtech.com/fr/special-reports/ia-generative-desinformation-russe/).
- [8]Klenman Zoe et Vallance Chris, « AI ‘godfather’ Geoffrey Hinton warns of dangers as he quits Google », BBC, 2 mai 2023 (https://www.bbc.com/news/world-us-canada-65452940).
