Governing the Digital Citizen

Featured image: A CCTV camera in London, 5 July 2010 (file photo). Photo: Gammew / Wikimedia Commons, CC0 1.0. Resized.

AI, Surveillance and the Rise of Power Without Consent

RAW TRUTH MEDIA | GLOBAL POWER & INSTITUTIONS

Analysis & Opinion | Artificial Intelligence, Digital Rights and Democratic Accountability

The modern citizen is increasingly governed through systems that are neither elected nor easily visible.

Algorithms determine which information appears on a screen, which job applications receive attention, which financial transactions are considered suspicious and which individuals are classified as high risk. Biometric systems verify identity. Digital platforms moderate speech at a scale no public authority could administer manually. Artificial intelligence assists governments, employers, banks, schools, hospitals and law-enforcement agencies in making decisions that shape everyday life.

Many of these systems provide real benefits. They can detect fraud, improve medical diagnosis, expand access to public services, translate languages, identify security threats and reduce administrative delay. Digital identity can help people prove eligibility for benefits or participate in the formal economy. Automated systems can process information faster and more consistently than overstretched bureaucracies.

But efficiency is not the same as legitimacy.

When decisions affecting rights and opportunities are delegated to technical systems, power can become harder to identify and challenge. A citizen may know that a benefit was refused, a bank account restricted or a social-media post removed without knowing which rule, dataset or institution produced the outcome. The official responsible may claim that the system generated the decision. The company providing the system may invoke commercial secrecy. The regulator may lack the expertise or access required to reconstruct what happened.

Authority remains real even when responsibility becomes difficult to locate.

The central challenge of the digital age is therefore not simply how to make artificial intelligence safe. It is how to prevent technological systems from becoming a form of government without meaningful consent, explanation or appeal.

From Administrative Tool to Governing Infrastructure

Technology has always influenced state power. Census records, identity documents, taxation databases and communications systems long predate artificial intelligence. What has changed is the scale, speed and predictive ambition of digital administration.

Traditional bureaucracy generally acts on information already recorded: age, income, residence, criminal conviction or tax liability. Algorithmic systems can go further. They can infer risk, predict behaviour, rank credibility and identify patterns that no individual official would detect.

These predictions may influence decisions concerning employment, credit, insurance, border control, policing, education and access to welfare.

The shift appears technical, but it is political.

Every model contains assumptions about which outcomes matter, which variables are relevant and what level of error is acceptable. A fraud-detection system must balance the cost of missing fraud against the harm of falsely accusing an innocent person. A hiring algorithm must determine which characteristics predict future performance. A platform’s recommendation system must decide what information deserves attention.

These are not value-free calculations. They distribute opportunity, suspicion and visibility.

The fact that an outcome was produced mathematically does not make it neutral.

The Problem of the Invisible Decision

Human decision-makers can be biased, inconsistent and corrupt. Automation is often introduced precisely because institutions seek greater consistency.

Yet automated decisions create a different accountability problem.

A person denied a public benefit may receive a standard notification without an intelligible explanation. An applicant rejected by an automated recruitment system may never know that software screened the application. A customer flagged for financial risk may be unable to discover whether the system used inaccurate information or an indirect proxy for ethnicity, religion, disability or socioeconomic status.

Even when a human officially approves the decision, the machine’s recommendation may dominate. Officials facing large caseloads and institutional pressure are unlikely to override a system presented as objective and data-driven.

The presence of a person at the end of the process does not automatically amount to meaningful human oversight.

Real oversight requires authority, information and time. The reviewer must understand the system’s recommendation, have access to the relevant evidence and possess genuine power to reach a different conclusion without being penalised for doing so.

Otherwise, “human in the loop” becomes a procedural label attached to an automated decision.

Data Does Not Represent Everyone Equally

Artificial intelligence learns from data generated by existing societies and institutions. If those records reflect inequality, selective enforcement or historical discrimination, a model may reproduce those patterns while presenting them as prediction.

An algorithm does not need to use ethnicity directly to produce unequal outcomes. Postcode, school history, employment gaps, language patterns or purchasing behaviour may function as proxies. A facial-recognition system may perform differently across demographic groups if its training data were unbalanced. A fraud model trained on past investigations may identify more suspicious activity in communities that were previously monitored more intensely.

Data can record institutional behaviour as though it were objective social reality.

The result is a feedback loop. A community receives greater surveillance because historical data classify it as high risk. Increased surveillance produces more recorded incidents. Those records then appear to confirm the original classification.

Technical accuracy must therefore be examined in context. A system can perform well according to an aggregate statistical measure while producing unacceptable errors for a particular group.

The relevant question is not only whether an algorithm is accurate. It is accurate for whom, under what conditions and at what human cost when it fails?

Digital Identity: Inclusion or Infrastructure for Control?

Digital identity systems are often introduced in the language of inclusion. They can help people access banking, healthcare, education, taxation and social protection. In countries where paper records are incomplete or public administration is fragmented, secure digital identification may reduce fraud and make services more accessible.

These benefits are substantial.

But identity infrastructure becomes dangerous when access to essential life depends upon a system the individual cannot realistically refuse.

If one digital credential becomes necessary for banking, travel, healthcare, employment and public benefits, technical failure can produce social exclusion. A person whose biometric data cannot be read, whose record contains an error or whose access is suspended may become unable to participate in ordinary life.

Centralisation also creates the possibility of function creep. Data collected for one legitimate purpose may later be linked to policing, immigration control, taxation, political monitoring or commercial profiling.

This does not mean every digital identity programme is a surveillance project. The danger lies in architecture and governance: how much data are collected, whether databases are linked, who can access them, how long information is retained and whether individuals have practical alternatives.

Consent has little meaning when refusing a system means losing access to essential services.

A rights-respecting digital identity system must therefore collect the minimum data necessary, separate incompatible functions, provide non-digital alternatives and establish rapid procedures for correcting errors.

Convenience should not become compulsory visibility.

The Biometric State

Biometric technologies promise certainty. Faces, fingerprints, irises and voices appear to provide a direct link between a person and an identity.

Unlike a password, however, biometric characteristics cannot simply be replaced after a breach. They are attached to the body.

Remote biometric identification also changes the relationship between citizens and public space. Traditional identification normally requires a specific interaction: an officer asks for a document, or a traveller presents a passport. Remote facial recognition can identify or track individuals without their awareness and without any direct encounter.

The distinction between targeted investigation and continuous population monitoring is fundamental.

Law-enforcement agencies may have legitimate reasons to use biometric tools in narrowly defined circumstances, such as locating a missing child or identifying a dangerous suspect. But a technology justified through exceptional cases can gradually become ordinary infrastructure.

Once cameras, databases and analytical systems are connected, surveillance no longer requires a visible decision to follow a particular person. The system can monitor everyone and search retrospectively when authorities develop an interest.

The European Union’s AI Act recognises some of these dangers. It prohibits particular manipulative and discriminatory practices and places remote biometric identification and other sensitive uses within a high-risk regulatory framework, subject to legal exceptions.

Regulation, however, remains only as strong as its definitions, enforcement and exceptions. A broad public-security exemption can swallow a narrow prohibition. A technically compliant system can still create a society in which anonymity in public becomes exceptional.

Content Moderation and the Privatisation of Speech

Some of the most consequential digital decisions are made not by governments but by private platforms.

Social-media companies decide what content is removed, restricted, labelled, monetised or recommended. Their systems influence which political arguments become visible and which speakers reach an audience. These companies do not merely host public discussion. Their algorithms structure it.

Content moderation is necessary. Platforms must respond to terrorism, child exploitation, fraud, threats, harassment and other unlawful conduct. A refusal to moderate would not create a neutral public square; it would often reward intimidation and manipulation.

The difficulty lies in the expanding category of harmful content.

Illegal content can at least be connected to identifiable law. “Harmful,” “misleading,” “unsafe” or “authoritative” information depends upon definitions that may change across political contexts. During crises, governments and platforms face pressure to act rapidly, often before facts are settled.

Errors are inevitable at scale. Satire may be treated as misinformation. Documentary evidence may be removed because it depicts violence. Political speech may be restricted because an automated system cannot interpret context, irony or historical reference. Similar statements may receive different treatment depending upon language, geography or public attention.

The European Union’s Digital Services Act attempts to improve accountability by requiring platforms to explain moderation decisions, assess systemic risks and provide greater transparency. Very large platforms — those reaching more than 45 million monthly users in the EU — carry additional obligations because of their capacity to shape opinion and amplify harm at scale.

These protections are important. Yet regulation introduces another question: when governments require platforms to mitigate broadly defined systemic risks, where does legitimate oversight end and indirect control over lawful speech begin?

The answer cannot be that platforms should be unregulated. Nor can it be that every restriction imposed in the name of safety is legitimate.

Users need clear rules, specific reasons for decisions, accessible appeals and information about when automated systems were involved. Governments should not pressure private companies to suppress lawful political speech that they could not lawfully prohibit directly.

The same constitutional and human-rights principles should apply whether censorship is administered by a ministry or outsourced to a platform.

Recommendation Systems and Behavioural Power

Digital power is exercised not only by removing information, but by deciding what people see.

Recommendation systems select news, videos, advertisements and political content according to predicted engagement. They shape attention while remaining largely invisible to the user.

This power is different from traditional censorship. A person may retain the formal right to speak while becoming practically undiscoverable. Another speaker may be amplified to millions because anger, fear or outrage increases interaction.

Platforms often describe these outcomes as the product of user preference. But preferences are measured through systems designed by the platform, and users choose within environments engineered to influence behaviour.

The Digital Services Act gives users of very large platforms the option of non-personalised feeds. This is an important recognition that algorithmic curation is not merely a commercial feature but a matter of individual autonomy and democratic concern.

Still, transparency about recommendation systems remains limited. Publishing general information about an algorithm does not reveal why a particular user repeatedly received a particular narrative or how commercial and political incentives affected its distribution.

The ability to influence attention at scale is a form of power even when no individual message is prohibited.

Public Authority and Private Infrastructure

Modern digital governance blurs the boundary between state and corporation.

Governments purchase analytical systems from private vendors. Police agencies use privately developed facial-recognition tools. Public institutions rely on cloud infrastructure operated by global technology companies. Platforms cooperate with authorities concerning illegal content, cybersecurity and national security threats.

This cooperation may be necessary. It also creates gaps in accountability.

A government may avoid disclosing how a system works because the technology belongs to a contractor. A company may insist that it merely provides a tool and bears no responsibility for how officials use it. Each actor points toward the other while the citizen faces the combined power of both.

Commercial secrecy should not override due process.

If a system influences a public decision affecting rights, the state must remain capable of explaining and defending that decision. Governments should not procure technologies they cannot independently audit. Contracts must provide regulators, courts and authorised researchers with sufficient access to evaluate performance, bias and security.

Public authority cannot outsource its responsibility to the algorithm’s manufacturer.

The Global Digital Governance Race

Digital governance is becoming an international competition over standards.

The European Union has chosen a regulatory approach centred on risk classification, prohibited practices, transparency and fundamental rights. The AI Act is the first comprehensive legal framework of its kind and has already influenced global policy discussions.

The United Nations’ Global Digital Compact commits governments to uphold human rights online and pursue a safe, secure and inclusive digital future. It also calls for international cooperation on artificial intelligence and the digital divide.

These initiatives reflect legitimate concern. But global frameworks can also distance rule-making from democratic scrutiny. Technical standards negotiated among governments, international organisations and major companies may eventually shape national policy without receiving sustained public debate.

Developing countries face a particular dilemma. They need digital infrastructure and access to artificial intelligence but may lack the financial and regulatory capacity to shape the systems they import. Technologies developed elsewhere arrive with embedded assumptions about identity, language, risk and administration.

The digital divide is therefore not only about who has internet access. It is about who creates the systems, owns the infrastructure, controls the data and writes the standards governing everyone else.

A country can be digitally connected while remaining technologically dependent.

Regulation Can Protect Freedom — and Threaten It

The debate is often presented as a choice between innovation and regulation. That framing is inadequate.

Poor regulation can entrench the largest companies by imposing compliance costs that smaller competitors cannot bear. It can encourage excessive content removal because platforms fear penalties. It can freeze technical categories that rapidly become outdated. It can also give governments legal language for expanding surveillance or influencing political speech.

But the absence of regulation does not preserve freedom. It transfers governing power to private corporations whose decisions are shaped by commercial incentives and internal rules.

The real choice is between different systems of accountability.

Effective regulation should focus on outcomes and rights rather than attempting to control every technical detail. It should impose the strongest safeguards where systems affect liberty, employment, credit, education, healthcare, essential services and democratic participation.

It should also preserve space for experimentation in low-risk uses.

Not every algorithm requires state supervision. But every system capable of seriously affecting a person’s rights should be subject to an identifiable chain of responsibility.

Rights for the Digital Citizen

A democratic digital order requires more than privacy policies and consent boxes.

Citizens should have the right to know when an automated system materially influenced a decision about them. They should receive an explanation expressed in language an ordinary person can understand — not only a technical description of the model.

They should be able to correct inaccurate data, challenge consequential decisions and obtain meaningful human review. Institutions should document system limitations and disclose error rates across relevant demographic groups.

High-risk systems should undergo independent audits before deployment and regular evaluation afterward. Regulators should be able to inspect training data, decision logs and performance without forcing companies to publish genuinely sensitive intellectual property to the entire world.

Data collected for one purpose should not automatically migrate into unrelated systems. Biometric surveillance should require strict necessity, proportionality and legal authorisation. Essential public services should preserve accessible alternatives for people who cannot or do not wish to use a particular digital credential.

Governments should publish registers of the high-risk automated systems they use, the vendors supplying them and the legal authority supporting deployment.

Finally, emergency powers and exceptional surveillance programmes should contain expiry dates. Technology introduced during a crisis has a tendency to remain after the crisis has passed.

The burden should rest on the institution exercising digital power — not on the citizen attempting to discover that the power exists.

Conclusion: Who Governs the Algorithm?

Artificial intelligence will become increasingly integrated into government, commerce and communication. Attempts to stop that development entirely are neither realistic nor desirable.

The question is whether technological progress will strengthen human agency or gradually replace it with systems that classify, predict and direct citizens without meaningful accountability.

A democratic society cannot accept a structure in which an institution denies a right, a company protects the system as a trade secret and an official explains that the algorithm made the decision.

Algorithms do not govern independently. People select their objectives, approve their deployment and decide how much authority to give them. When responsibility becomes invisible, that is a failure of institutional design—not an unavoidable property of technology.

Digital governance should begin from a simple principle: no person should lose access to liberty, livelihood, essential services or public participation because of a system they cannot understand, challenge or escape.

Safety matters. Privacy matters. Innovation matters. So do freedom of expression, due process and democratic control.

The future digital citizen should not be asked to surrender one set of rights in order to protect another.

Because a society is not genuinely free when every decision can be automated, every movement can be recorded and every opinion can be filtered — while no one can identify who is ultimately accountable.

Sources and Further Reading

• European Commission, Artificial Intelligence Act

• European Commission AI Act Service Desk, prohibited AI practices under Article 5

• European Commission AI Act Service Desk, high-risk systems under Annex III

• European Commission, Digital Services Act

• European Commission, very large online platforms and search engines

• European Commission, transparency under the Digital Services Act

• United Nations, Global Digital Compact

• United Nations, full text of the Global Digital Compact

• European Union, General Data Protection Regulation

Raw Truth Media | Global Power & Institutions

Editorial note: This article examines institutional risks created by artificial intelligence, biometric technologies, digital identity and platform governance. It does not argue that all digital systems are coercive or that regulation is inherently hostile to innovation.