We’re living through a quiet, unsettling shift in how power works and how it shows up in daily life. The Enlightenment taught us to trade a ruler’s whims for rules and procedures, to trust the assembly over the autocrat. But the assembly has changed shape. It has slipped out of the chamber and into code, standards, and interoperable databases that now decide how we prove who we are, what we can say, how we earn, and where we can go. The question is no longer just who governs, but which architectures make us legible and manageable in the first place.
“Code is law,” Lawrence Lessig warned, and we now live with its grown-up version: code as policy, process, and precedent—executed at machine speed, at planetary scale. Voltaire’s provocation—that the tyranny of one can be more bearable than the tyranny of many—lands differently when the “many” are risk engines, compliance rails, and platform protocols.
From Beijing to Brussels, Bengaluru to Brasília, a mesh of systems promises efficiency while stripping away discretion, turning governance into an ambient climate that’s hard to escape. The pandemic simply made its tools ordinary: QR gates, health passes, automated triage. How does such power thicken? How does accountability fade into consensus and software? What becomes of democracy when mercy is treated as an error to fix? And can we rebuild the operating system of our civilisation so that code serves life rather than administering it?
We’ve long framed the choice as tyrant versus crowd. Since the Enlightenment we’ve preferred the assembly: checks and balances, due process, the stabilising rhetoric of the common good. But there’s a discomfort under that reassurance. Concentrated power can be seen, named, and resisted. Power diffused into procedures and standards can turn invisible as it tightens its grip. The line often attributed to Voltaire (and sometimes Chamfort) is a useful irritant: if you must choose, the tyranny of one is easier to spot than the tyranny of many. This isn’t an ode to monarchy. It’s an invitation to change the lens—from who holds power to how power is encoded and spreads.
When authority resides in a person, it stays human. It feels hunger and grief. It can be distracted, softened, or simply wear out. Singular power has a failure point: the limits of an individual. A tyrant’s cruelty is personal and patchy; that very humanity—however warped—leaves a narrow alley for courage, cunning, and luck.
Bureaucracy dissolves that drama into mechanism. Domination becomes architectural: cured into process and guarded by consensus. Harm isn’t delivered in anger; it emerges from spreadsheets, flowcharts, risk models, and the letter of the law. A legislature votes, an agency drafts, a department implements, a contractor builds, a platform enforces. The system then purges inconsistency in the name of fairness and efficiency. Dissent becomes a bug to patch; human softness, a variance to eliminate. We’ve seen this moral psychology before. Hannah Arendt called it the banality of evil. James Scott showed how states simplify complex lives to make them legible and, in doing so, crush what matters. Today those impulses don’t sit in files and forms; they’re compiled into code and executed at machine speed.
Lessig’s insight is no longer a metaphor. Code has become policy, procedure, and precedent—plus something colder: relentless indifference. Decisions once made by committees and courts are now made by risk engines and recommender systems. The “assembly” has expanded to include standards bodies, interoperability protocols, blacklists, app store rules, content policies, credit scores, sanctions screens, and compliance rails—a socio‑technical lattice mediating identity, speech, livelihood, and movement. Scholars call it algocracy. The danger isn’t the spectacle of a tyrant; it’s the implacable calm of a distributed machine, stitched together from public and private parts, justified by convenience, and cloaked in scientific objectivity.
The American experiment offers a clear view of the shift. The founders feared what Tocqueville later called the tyranny of the majority, so they built in friction: federalism, bicameralism, separation of powers, a Bill of Rights. They designed for safe‑to‑fail, not fail‑safe—systems that absorb shocks by slowing down and spreading authority. In our era, the speed of information and industrialised outrage have turned those safeguards into weak points. A new assembly has formed: not Congress alone, nor the old caricature of the administrative state, but a matrix of agencies, contractors, platforms, and data brokers that turn orthodoxy into procedure. Power doesn’t announce itself from the Oval Office anymore; most days, only farce does. Real power lives in algorithm thresholds, policy updates, sanctions engines, and content moderation queues. Citizens now find their prospects, speech, and financial survival conditioned by tools that promise consistency yet can’t show clemency.
This isn’t uniquely American. It’s global, adapted to local histories. In China, the fusion of grid management, phone metadata, and cameras has turned “stability maintenance” into a technical operation. Systems like the Integrated Joint Operations Platform in Xinjiang have flagged the mundane as suspicious. Pandemic health codes governed movement; emergency tools proved stubbornly sticky. Russia’s SORM and Iran’s National Information Network combine telecom control with deep packet inspection and data localisation; Vietnam has refined its own cyber‑policing. Gulf states pair biometric identity with competent surveillance bureaucracies.
In market‑led democracies such as the EU, India, and the United States, algorithmic governance has grown out of compliance rather than ideology: anti‑money‑laundering, credit scoring, insurance pricing, predictive policing, welfare fraud detection, content moderation at scale. The tools are sold as risk reduction and built by experts. Laws like the GDPR, the Digital Services Act, and the EU AI Act try to discipline the field. Yet machinery meant to protect rights can harden into process‑heavy coercion if contestation is weak or explanations remain purely formal.
Elsewhere—especially in capacity‑poor states across Africa, Latin America, and South Asia—the apparatus is more assemblage than system. Core functions—identity, payments, communications—run on foreign clouds, donor‑funded digital IDs, and private messaging platforms. That creates double dependency: on the infrastructure and on the opaque risk tools that sit atop it. Kenya’s Huduma Namba rollout slowed under court‑ordered privacy assessments, hinting at guardrails. Nigeria’s SIM re‑registration and central bank digital currency experiments show how identity and money can be linked to nudge behaviour or police dissent. In humanitarian settings, biometric schemes launched with good intentions—from Bangladesh to Yemen—have raised serious worries about consent and future misuse.
How is inescapability engineered? In familiar, practical ways:
- Identity is tied tightly to accounts and devices. Real‑name policies, SIM registration, and national IDs linked to faces and fingerprints are justified by safety. But once identity becomes the key to everything, exit gets costly. To communicate, pay, travel, get benefits, or even enter a building, you must authenticate.
- Payments become chokepoints. Know Your Customer and anti‑money‑laundering rules knit banks, fintechs, and card networks into a global compliance belt. Accounts can be frozen by algorithm. Central bank digital currencies promise inclusion and efficiency while carrying the latent affordance of programmability. Canada’s 2022 emergency account freezes showed how quickly financial levers can be pulled in disputes that politics should settle. NGOs are routinely debanked by risk models few can interrogate.
- Scoring becomes destiny. Australia’s Robodebt, the Dutch childcare benefits scandal, the UK’s A‑level grading algorithm—statistical shortcuts defended as objective scale into mass harm with no single villain.
- Speech is managed through private protocols. Governments from India to Turkey to Brazil legally press platforms to remove content and hand over data. Platforms translate that pressure into global rules executed with a numb, generalised logic. The forced removal of opposition apps in Russia revealed how a few private chokepoints can reshape political life.
- Work is automated. Gig workers across China, Europe, and the Americas are dispatched, priced, and terminated by models they can’t see, with appeals that vanish into ticketing systems. Warehouse labour runs to the rhythm of scanners.
- Borders are externalised into risk scores. Passenger data is sifted long before a human officer appears; travellers are pre‑cleared—or pre‑denied—by models.
Look also at the institutional ecology. The assembly is no longer solely public. Public–private consortia co‑produce policy as pipeline. The Financial Action Task Force sets standards that reshape banking far from any parliament. The ITU and ISO codify technical norms. App stores, cloud providers, and payments networks exercise private vetoes over speech and commerce at scale. Accountability dissolves into terms of service, vendor NDAs, and “security by obscurity.” No single person owns the harm because everyone owns a process. Consensus becomes a self‑sealing loop, validated by compliance reports and audit trails. The machine applauds itself for doing what the machine requires.
This is not nostalgia for the despot’s occasional “good moments.” Absolutism scales its own horrors, and the twentieth century is warning enough. The point is to recognise a different absolutism—distributed, procedural, data‑driven—and to ask what it would take to rehumanise governance in a digital civilisation, if that is still possible.
There are moments of pushback. Dutch courts struck down SyRI. Australia convened a Royal Commission on Robodebt. The UK reversed its grading algorithm after public outrage. European courts curtailed indiscriminate data retention. These wins reintroduce friction, slow the machine, and make room for testimony.
Some democratic experiments hint at another way. Taiwan’s vTaiwan process, using tools like Polis, broadens deliberation without collapsing into social‑media noise. Barcelona’s Decidim treats participation as a continuous, transparent practice, not a rare ritual. Participatory budgeting in parts of Brazil shows how money can teach civic habits instead of disciplining behaviour. In markets, too, we see gestures toward dignity: interoperable messaging, end‑to‑end encryption, and decentralised identity give people real exit and voice. Brazil’s Pix and India’s UPI show how open payment rails can cut private chokepoints, even if state power remains. None of this is sufficient. All of it teaches.
What should a civilisational design response look like?
- Constitutionalise automation. Any system that can deny a job, a loan, a benefit, or a voice should face standards akin to a search or seizure: clear authority, necessity, proportionality, and built‑in safeguards.
- Reassert named human accountability for harm. Not oversight theatre, but a real signature—an identified official with training and personal and institutional liability for negligence.
- Guarantee universal rights to notice, explanation, and contestation. People should know when automation is used, understand in plain language the decisive factors, and have timely access to an independent appeal with power to overturn.
- Require binding algorithmic impact assessments before deployment, with public summaries, engagement with affected communities, and commitments to post‑deployment audits.
- Open black boxes to qualified scrutiny through independent audits and adversarial testing—protecting privacy and legitimate IP, but not allowing secrecy to stand in for justification.
- Design for safe‑to‑fail with sunset clauses and kill switches. Time‑limit automated systems, require periodic reauthorisation, and keep off‑ramps for emergencies and systemic error.
- Relearn restraint. Make data minimisation and purpose limitation defaults, not slogans. Ban promiscuous data fusion without explicit mandate and robust oversight.
- Preserve the right to an analogue and pseudonymous life. Access to cash and offline services for essentials isn’t nostalgia; it’s a guardrail for freedom. Protect pseudonymity where reprisals are likely.
- Extend due process to payments. Financial deplatforming should follow clear standards and be appealable. Build human‑rights guardrails into anti‑money‑laundering and sanctions regimes, not only compliance incentives.
- Reflect dignity in procurement. Refuse generalised social scoring, real‑time face recognition in public spaces, and spyware without stringent judicial controls. Align export controls with human rights, not just geopolitics.
- Make oversight polycentric. Create algorithmic ombudspersons, cross‑agency ethics boards with teeth, legislative committees with independent expertise, and citizen juries to review contentious deployments.
- Invest in verifiable, privacy‑preserving compliance where the state has legitimate aims. Use zero‑knowledge proofs, secure enclaves, and differential privacy to reconcile competing goods—by choice, not as afterthought.
- Build global norms. Ratify human‑rights‑centred AI instruments, bake algorithmic due process into trade and data‑transfer agreements, and coordinate to curb the sale of mercenary surveillance.
This is not a laundry list so much as an ethical stance. Liberal democracy’s genius was to institutionalise friction—to bake human fallibility into the system so no single error proved fatal. The platform economy and digital state have quietly optimised much of that friction away. We feel the result not as orders barked from on high, but as a kind of suffocation. Our task is to put friction back where the stakes are high, design for graceful failure, and insist that every consequential system can still be corrected. In complex systems terms, we need requisite variety to match the diversity of lives we claim to serve, and safe‑to‑fail probes that learn in public rather than grand projects that fail in secret.
The real choice isn’t between emperor and assembly. It’s between architectures that preserve human failure points—mercy, discretion, the capacity to be moved by a story—and architectures that erase them in the name of consistency. The twentieth century left us statecraft built for the industrial age: rational, hierarchical, standardised. The twenty‑first offers something else: polycentric, transparent, participatory, with protections hard‑coded not only in statutes but in protocols. Getting there needs more than reform. It needs a fuller civic imagination—the courage to slow certain processes so responsibility can be taken, the humility to collect less because we cannot protect more, and the wisdom to keep open analogue channels through which compassion can still travel.
If an inescapable code is the new tyranny, then deprogramming the machine is a civilisational imperative. That does not mean rejecting technology. It means designing for human flourishing rather than institutional convenience; trusting citizens assembled as communities, not just datasets; building with an ethic of care. We will know we are making progress when the person turned away by a model can tell their story to someone with the authority to change the outcome—when the system’s asocial indifference is interrupted by a named human who can be persuaded. Until then, we should refuse the fatalism that says, “this is just how things are now.” The law’s assembly must remain a meeting of accountable humans, not an operating system that confuses compliance with justice. We can choose architectures of mercy. We can still insist that code serves life.