The Algorithmic Architecture of Authoritarianism

I

was thirteen when I first read George Orwell’s 1984. Growing up in a politically unstable region, the novel felt like a commentary on the world outside my window.

Throughout human history, every iteration of state power deployed mechanisms of control to consolidate authority and resources for the ruling elite. Orwell’s genius lay in capturing this universal mechanics of state power.

Orwell imagined a state that demanded obedience through surveillance and controlling flow of information. Generative AI, state-controlled algorithms, digital identity systems, and automated public infrastructure now determine what citizens should believe, what resources they survive on, and whether authentic public opinion exists at all.

Unlike Orwell’s oppressive telescreens, contemporary technologies are carried voluntarily. Citizens willingly purchase, update, and carry the devices of their own oversight, surrendering their autonomy for digital convenience. In fact, digital architecture continuously fosters dependency of present life on these tools.

Consider the control of information. Orwell’s Ministry of Truth rewrote the past by destroying records and manufacturing official narratives. Today, mass reliance on generative artificial intelligence is quietly conditioning populations to trade cognitive friction for friction-free answers, outsourcing basic tasks, writing, and critical reasoning to algorithms. As individuals become dependent on quick machine outputs, they are systematically conditioned to think, articulate, and structure arguments in uniform mannerisms dictated by the statistical averages of large language models.

Modern digital control

This homogenization of thought acts as an organic counterpart to Newspeak. By funneling human expression through standardized stylistic templates and algorithmic consensus, generative AI flattens individual idiosyncrasy, critical divergence, and creative rebellion, conditioning society to accept machine-curated outputs as neutral truths rather than biased reflections of corporate training data.

Modern digital control decides which information gains visibility, which narratives are deprioritized, and which voices are quietly shadow banned. Information simply gets buried under automated streams of synthetic noise or throttled at the network level. In 2025 alone, human rights monitors documented over 300 internet shutdowns across more than 50 countries during periods of civil unrest and armed conflict.

When access to the internet is restored, the medium itself acts like a double-edged sword. For populations living under military occupation or authoritarian rule, the smartphones required to seek aid or document atrocities simultaneously expose their locations, networks, and identities to state intelligence apparatuses.

The danger of algorithmic governance is hardwired into the administrative machinery of Western democracies as well. Across the globe, governments are outsourcing public administration to automated risk-scoring models that lack basic transparency and due process.

In the Netherlands, the court system was forced to strike down the government’s SyRI (System Risk Indication) algorithmic risk model, which was deployed to detect welfare fraud by aggregating personal data across public agencies. The system disproportionately targeted low-income and racial minority neighborhoods with opaque scoring metrics, offering citizens no clear avenue for legal appeal.

Fraud-detection platforms

In the United States, automated fraud-detection platforms, such as Michigan’s MiDAS system, wrongly flagged tens of thousands of unemployed citizens for fraud, stripping them of benefits and imposing financial penalties without human oversight.

Algorithms inherit the assumptions, limitations and biases of the systems in which they are built. When an opaque risk score determines access to public benefits, or when a person cannot discover why a decision was made, technological efficiency can become a substitute for due process.

Unlike the centralized state apparatus of 1984, contemporary surveillance relies on a complex, transnational architecture connecting tech monopolies, data brokers, and defense contractors.

Silicon Valley tech giants control the cloud infrastructure and generative artificial intelligence models that feed public consciousness, while states purchase surveillance capabilities off the shelfs. This creates an asymmetric alignment of commercial profit and state security. Even as corporate frontier labs race to deploy increasingly autonomous models, acknowledging catastrophic safety risks and deceptive alignment failures in their technical disclosures, policymakers around the world actively resist binding international regulations in favor of technological dominance.

Former Anthropic safety researcher Jacob Coxon publicly warned of a 10 percent chance of AI causing human extinction within the decade, echoing warnings from pioneers like Geoffrey Hinton and institutions like the Centre for International Governance Innovation. Yet at the United Nations General Assembly, U.S. President Donald Trump explicitly dismissed existential concerns, declaring that official U.S. documentation would rebrand “artificial intelligence” as “Super Intelligence” (SI) and vowing not to rein in growth.

When public institutions become dependent on privately controlled infrastructure, who sets the limits on its use? Who can audit the systems making consequential decisions? And what happens when a technology becomes so deeply integrated into public administration, financial networks or national security that abandoning it carries prohibitive costs?

Is AI weightless?

Crucially, the popular illusion that AI is a weightless, virtual cloud conceals an intensely material reality, massive industrial facilities guzzling water, coal, and gas to train and run frontier models. This computational thirst is reviving fossil fuels, keeping aging coal plants online and burning gas to guarantee 24/7 baseline electricity. In regions across South Asia, including India, Pakistan, and Bangladesh, where groundwater tables are already severely depleted and over 1.3 billion people face water stress, hyperscale data centers receive guaranteed municipal water concessions for evaporative cooling, prioritizing corporate server heat management over agricultural irrigation and drinking water. In developing nations across Sub-Saharan Africa and South Asia, data centers operate thousands of diesel generators as power backups, emitting nitrogen oxides and fine particulate matter directly into adjacent residential areas and driving up local rates of respiratory illness.

While compute power, proprietary software, and market capitalization are overwhelmingly concentrated in wealthy Northern nations, the physical, environmental, and human burdens are offloaded onto the Global South. The hardware supply chain depends on rare earth minerals like cobalt, lithium, and nickel extracted from open-pit mines in the Democratic Republic of Congo and across Southern Africa, destroying rainforest ecosystems and contaminating watersheds.

The lifecycle ends where it begins: millions of tons of high-performance server hardware, discarded after short 2 to 4-year operational lifespans, are shipped as electronic waste to toxic dumping grounds like Agbogbloshie in Ghana, where local workers burn components in open fires and face severe dioxin, lead, and mercury poisoning. Furthermore, the human labor required to train these models relies on low-wage workers in Kenya, Uganda, India, and the Philippines, paid fractions of Western minimum wages to review and filter toxic content without healthcare support.

The ultimate power of modern surveillance lies in its psychological efficiency and structural asymmetry. The modern Big Brother does not need a telescreen in every living room, it is enough to control the interconnected digital infrastructure through which society communicates, accesses knowledge, and survives.