Two weeks after Moscow’s full-scale invasion of Ukraine in March 2022, Russian TV Channel One editor Marina Ovsyannikova made a split-second decision that pierced the state’s tightly controlled information bubble. Bursting onto the set of the live evening newscast, she held up a hand-drawn sign directly behind the anchor’s head bearing a stark warning: “Stop the war. Don’t believe propaganda. They’re lying to you!” She shouted, “No to war!” before security forces violently dragged her away off camera.
That act of live defiance remains an outlier. No one has protested the war on Russian television since then, a silence bought partly through drastically tighter security measures and a sweeping, pervasive climate of fear. But by October 2024, Margarita Simonyan, one of Russia’s chief state propagandists, offered an entirely different explanation for the absence of dissent on air. In a television interview, she casually revealed that a growing number of anchors across the state-backed RT network were no longer real human beings.
“That face doesn’t exist,” Simonyan explained to her interviewer, describing how producers generated the visual likeness, the voice, and every other element using artificial intelligence. In a subsequent interview with the business newspaper Vedomosti, she spelled out the ultimate bureaucratic logic behind the technology: “These anchors don’t need a salary or insurance,” she said. “They won’t get arrested, and the police won’t search their homes. It’s all wonderful and terrifying at the same time.”
This dynamic captures the core promise that artificial intelligence extends to every autocratic ruler across the globe: the tantalizing prospect of maintaining absolute political control without the persistent hazard of delegating vital tasks to unreliable, self-interested humans. For modern dictators, it is an exceptionally attractive proposition. No matter how powerful or entrenched, authoritarian leaders have historically depended entirely on extensive webs of human subordinates—including censors, propagandists, security guards, police officers, intelligence analysts, and local bureaucrats—to execute their daily orders.
Yet, human subordinates inherently pose an enduring, structural threat. They can shirk their responsibilities, steal state resources, lie to their superiors, leak sensitive intelligence, or conspire behind closed doors against the boss. Consequently, one of the oldest dilemmas facing autocrats has centered on how to empower agents to carry out administrative commands without simultaneously enabling those same agents to turn against the supreme leader.
Historically, autocrats managed this treacherous tradeoff by systematically filling key government positions with mediocrities whose professional incapacity and lack of ambition served, as political theorist Hannah Arendt famously observed, as “the best guarantee of their loyalty.” But this deliberate preference for incompetent loyalists inevitably makes it far harder for dictatorships to govern effectively, manage complex modern economies, or survive sudden national crises.
Artificial intelligence now appears to offer contemporary autocrats two distinct pathways to resolve this long-standing governance dilemma. First, it grants leaders the unprecedented capacity to monitor, evaluate, and discipline their followers in real time. A sudden, suspicious shift in a regional official’s spending habits, lifestyle, or meeting schedules can be flagged automatically by an advanced AI system, entirely bypassing the need for self-interested human informants whose motives are often murky.
Second, AI promises to eliminate underlings altogether. By automating complex bureaucratic tasks that once required human discretion and oversight, rulers can replace unreliable intermediaries with faithful algorithms that cannot be bribed, blackmailed, or manipulated. Officials within autocratic regimes have increasingly spoken openly about this disruptive transformation. Artificial intelligence “can definitely replace half of officials,” Maksut Shadaev, Russia’s Digital Development Minister, told a crowded data forum in Moscow, adding for emphasis, “maybe slightly more.”
Yet these grand promises of a self-running, frictionless state are fundamentally illusory, because artificial intelligence systems are never truly autonomous. Complex machine-learning architectures must be initially built and continuously maintained by a specialized cadre of engineers, programmers, and data scientists whose deep technical expertise political leaders are largely incapable of evaluating.
A dictator who turns to artificial intelligence to reduce their historic dependence on unreliable human bureaucrats may ultimately end up relying, far more blindly than before, on the minuscule handful of tech specialists who keep the digital machinery functioning. This newly minted digital Praetorian Guard harbors many of the exact same systemic weaknesses as the historical guards that preceded it, transforming the technical experts into a potent, hidden threat to the ruler.
Eyes and Ears
Globally, China has established a commanding lead in developing and deploying artificial intelligence systems capable of monitoring or replacing human subordinates. Elaborate smart city programs established in Beijing, Shanghai, and other major urban centers utilize AI to continuously track performance and centralize oversight over local officials who once operated in comfortable political obscurity, subjecting them to unyielding algorithmic performance reviews.
China also developed and tested a specialized “Zero Trust AI system,” deploying it across approximately 30 counties and cities over the past decade. This system cross-referenced data from more than 150 separate government databases to automatically catch instances of embezzlement, corruption, and nepotism among local officials. In practice, the system arguably worked a little too well, aggressively flagging 8,700 officials before local governments, buckling under intense political pressure from the very bureaucrats the technology was monitoring, began rolling it back.
Even so, Beijing has retained its advanced e-government portals, which employ automated systems for issuing official documents and permits, thereby drastically reducing face-to-face interactions between citizens and bureaucrats that historically bred petty corruption and favoritism. Meanwhile, the Chinese Communist Party has aggressively pursued research into what it terms “thought management”—exploring how to leverage AI to generate hyper-personalized, microtargeted propaganda. China’s vast army of internet commentators, which was once composed of paid human workers, is already being systematically replaced by automated bots.
Russia has pursued these exact same technological goals with similar institutional fervor. Moscow’s sweeping citywide facial recognition system, deployed across an infrastructure of 200,000 closed-circuit cameras, was heavily utilized to identify, track, and apprehend political protesters during major anti-government demonstrations. The central government in Moscow has increasingly relied on automated data collection and aggregation methods to completely bypass regional officials who might otherwise taint administrative data or manipulate it to promote their own parochial interests.
In 2023, Russia’s federal internet regulator launched Oculus, a sophisticated AI system designed to scan hundreds of thousands of images and text files per day for prohibited content, including unauthorized political commentary—operating at orders of magnitude beyond what human censors could possibly process. Simultaneously, the Russian security service’s proprietary Meliorator tool has been deployed to generate over a thousand profiles of fictitious American citizens, primarily operating on the social media platform X, to effectively spread official Kremlin narratives at a fraction of the financial cost required to run traditional human troll farms.
Other authoritarian governments are rapidly following in the footsteps of Beijing and Moscow. The United Arab Emirates established a Regulatory Intelligence Office utilizing artificial intelligence to autonomously draft and amend federal laws—work that was previously handled exclusively by legislative staff. A year later, UAE Prime Minister Sheikh Mohammed bin Rashid al-Maktoum announced that autonomous AI agents would take over half of all federal government operations within a span of two years, declaring that “the performance of ministers, directors general, and entities will be assessed based on their ability to adopt this transformation.”
Despite these sweeping administrative innovations, such systems cannot actually create what dictators desire most: a truly self-running state. Even advanced AI-generated television anchors still require human technicians to build, update, and maintain them, just as automated censorship systems require specialized personnel to continuously retrain algorithms as the linguistic vocabulary of political dissent inevitably shifts. Behind every sophisticated AI model stands a small, indispensable group of engineers and data scientists performing essential maintenance, and their daily work remains so deeply technical that supreme political rulers cannot possibly comprehend it.
This inherent technological dependence quickly morphs into a potent new form of political power. The engineers who maintain an autocrat’s expansive surveillance apparatus, for instance, retain the ability to fundamentally shape what the ruler actually sees. They decide precisely what information is important enough to pass up the chain of command and in what visual or analytical format to present it. They determine which emerging threats get flagged for immediate attention, and they can covertly adjust the underlying algorithms that select which political content gets suppressed and which citizens get arrested.
They can execute these adjustments without anyone inside the palace noticing. Autocrats who turn to artificial intelligence to escape their historic dependence on unreliable subordinates have merely transferred their core institutional vulnerabilities onto a brand-new group of technical officials.
In many ways, this new digital Praetorian Guard could prove significantly more dangerous to a regime than previous ruling elites. This heightened danger stems partly from the inherent opacity of complex AI technology, but also from the fact that this technical clique is remarkably small. Roman emperors typically maintained tens of thousands of dedicated soldiers serving directly within the Praetorian Guard, while modern autocrats rely heavily on massive standing armies.
With advanced artificial intelligence, however, an autocrat requires only a tight-knit clique of software engineers to build and maintain massive, nation-wide surveillance and control systems. While this initially benefits rulers by reducing the sheer number of people they must personally watch, it dangerously concentrates points of systemic failure. A devastating operational disruption requires only a handful of disgruntled engineers, and a small group makes it far easier for members to scheme collectively against the leader.
Furthermore, members of this new digital guard are virtually indispensable. While a dictator can easily replace high-ranking military generals during a crisis without destroying the underlying armed forces, running complex machine-learning infrastructure requires such a rare, specialized set of technical skills that engineers are exceptionally difficult to replace without causing catastrophic operational disruptions, granting these actors immense political leverage.
The Indispensables
The critical importance of tech workers to modern autocracies has already become plainly apparent. When tens of thousands of IT specialists began fleeing Russia immediately following the full-scale invasion of Ukraine in February 2022, the Kremlin responded not with immediate threats or mass arrests, but with sweeping inducements. The government offered tech workers official deferments from military conscription, exempted their corporate firms from taxation, and heavily subsidized their mortgages.
When Russian President Vladimir Putin subsequently ordered the mobilization of 300,000 military reservists seven months later, domestic IT workers were granted comprehensive, full-on waivers from active service. While this might initially seem like an excessive precaution given that an individual AI system requires only a handful of engineers to operate, an authoritarian regime cannot know in advance precisely which specific engineers it will need during a crisis, nor can it train replacements quickly. Consequently, the state is forced to hold on to the entire national talent pool from which those few indispensable individuals are selected.
These generous financial measures ultimately failed to stem the outward migration of roughly 100,000 highly specialized technology workers—representing about a tenth of Russia’s total tech workforce—over the course of 2022. Yet Moscow steadfastly stuck to bribery and economic privilege because deep technical expertise is extraordinarily difficult to conscript effectively at gunpoint, and conscripted experts are inherently far less likely to follow instructions reliably.
Even so, vast amounts of money and social privilege cannot guarantee that these technical elites will remain politically loyal during a severe crisis. Autocrats would do well to recall the historical lesson of the original Roman Praetorian Guard: for a significant period, it provided Roman emperors with reliable protection and personal security. But eventually, the praetorians discovered the true value of their own indispensability, and by the second century, they were openly auctioning off the imperial throne to the highest bidder.
This new technological elite possesses the capacity to do precisely the same thing. In June 2023, when a heavily armed column led by Yevgeny Prigozhin and his Wagner paramilitary company moved aggressively up the highway toward Moscow, President Putin depended entirely on his internal security services to provide accurate, real-time intelligence on what was happening.
Future autocrats in Putin’s position will receive such critical warnings directly through machines: intercepted communications automatically sorted by linguistic software, live security camera feeds filtered through facial recognition systems, and regional intelligence reports compiled into automated executive dashboards. The software engineers who build and run those underlying systems could easily strike a clandestine deal with an upstart political challenger and quietly drag their feet.
They could intentionally delay critical data streams, allow security alerts to arrive a few hours late, degrade video feeds at the most inconvenient operational moments, or cause a facial recognition system to temporarily stop functioning altogether. In such a scenario, the supreme ruler would be operating completely blind. The Praetorian Guard did not actually need to assassinate Emperor Nero to replace him; they simply had to abandon him in his hour of need for a politically viable rival.
These computer programmers are unlikely to seize direct political power for themselves, primarily because they lack the coercive force that coup leaders ultimately require: physical weapons. Yet there is already clear historical precedent for software engineers utilizing their technical leverage to shape major leadership challenges.
In August 1991, when hardline elements of the Soviet military attempted to seize political control and overthrow Mikhail Gorbachev, a small group of computer programmers operating the Relcom network kept vital information flowing abroad and successfully relayed Russian President Boris Yeltsin’s decrees to audiences both inside the country and across the international community. The coup plotters proved entirely powerless to stop the rapid spread of news regarding Yeltsin’s open defiance, and the political conspiracy collapsed entirely within three days.
Slimming Down
Ultimately, artificial intelligence will not allow autocrats to completely dismiss all of their human enforcers. Someone must still physically make arrests, guard borders, and run the prison system, and authoritarian rulers will continue to buy domestic political loyalty by offering supporters lucrative state jobs.
However, artificial intelligence will enable authoritarian governments to significantly downsize their vast bureaucracies, and the absolute number of officials ultimately matters far less than how effectively the ruler oversees them. Keeping track of scheming, corrupt, or incompetent subordinates has always stood as the autocrat’s primary administrative burden. Artificial intelligence promises to lighten that heavy burden, allowing supreme rulers to watch their officials far more closely without having to invest as much active, personal attention.
In the near term, therefore, artificial intelligence may provide substantial benefits to autocracies. Despite the numerous structural obstacles and technical hurdles involved in widespread deployment, the technology is already delivering the tangible advantages of significantly cheaper surveillance, smarter censorship, and a reduced number of human subordinates to fear and distrust.
Yet for autocratic rulers, the deep temptations of artificial intelligence may ultimately re-create the exact political trap they are desperately seeking to escape. The more an authoritarian regime depends on artificial intelligence to maintain its grip on power, the more it depends on the specialized humans who keep that intelligence running. And those technical individuals, much like every other praetorian class throughout recorded history, will quickly discover precisely what their absolute indispensability is worth.
Cyber Security & Cryptography
Cyber Security & Cryptography
Cyber Security & Cryptography
