Tech Policy and Democratic Accountability

Imagine being denied a tenancy, a job interview, a loan, or access to a basic service by a system no one can properly explain.

No person takes responsibility. No clear reason is given. There is no meaningful appeal. You are simply scored, sorted, and excluded.

That is the practical fear beneath the debate about artificial intelligence. Not just that machines may become powerful, but that they may be deployed inside systems already built to distance decision-makers from consequences.

For years, we have heard warnings about AI – existential threats, job displacement, democratic disruption. Politicians speak about it as if it is an emergency unfolding in real time. Yet when it comes to meaningful regulation, almost nothing happens.

Instead, political attention is channelled into social-media moderation and online harms: visible, emotive issues that leave the deeper structures untouched.

This contradiction is not a mystery. It is a symptom of something older and more uncomfortable:

The guardrails that would have made AI safer were dismantled long before AI arrived.

AI is not the original cause of our vulnerability. It is the mirror showing how vulnerable we already were.

I. The Myth That Technology “Moves Too Fast”

We are often told that AI is difficult to regulate because it evolves too quickly. But the truth is simpler, and far more revealing.

The problem is not simply that AI moves too fast. The deeper problem is that the UK has lost much of the institutional capacity needed to govern powerful technologies in the public interest.

Over decades, the state was hollowed out in ways that often sounded efficient at the time. Expertise was outsourced. Regulators were asked to do more with less. Public institutions became dependent on consultants, contractors, and private-sector systems they did not fully control.

The result is not simply a slow state. It is a state that has become structurally dependent on some of the same interests it is supposed to scrutinise.

In a system like this, everything looks too fast. Not because it is, but because the institutions meant to govern it were deliberately stripped of the ability to do so.

This matters because regulation is not just the act of passing a law. It requires expertise, enforcement, independence, funding, technical understanding, and the confidence to say no to powerful actors.

Without those foundations, even well-intentioned rules become symbolic.

II. AI Entered a System Already Designed for Extraction

AI did not land in a neutral landscape. It entered a political and economic system already shaped by forty years of:

Deregulation
Privatisation
Outsourcing
Financialisation
The weakening of labour, environmental, and consumer protections – including the growth of work models that blur employment status and shift risk onto workers

These were not isolated policy choices. They were part of a coherent ideological project:

To free markets from the constraints that protect people, communities, and the environment.

That project changed not only who owned services, but how problems were understood. Social needs were reframed as markets. Public responsibilities became contracts. Citizens were increasingly treated as customers, users, claimants, data points, or risks.

Labour protections are a clear example. In the gig economy and other insecure forms of work, the issue is not simply that people are paid too little. It is that the relationship itself is often structured to avoid responsibility. Workers may be treated as independent enough to carry the risks of the job, but not independent enough to set prices, negotiate terms, build security, or exercise real control.

This is how employment status becomes a loophole. Costs that should sit with the employer – downtime, equipment, insurance, holiday, sickness, pensions, training, scheduling instability, and the risk of fluctuating demand – are pushed onto the worker.

The language is flexibility. The reality is often underpayment with extra responsibility attached.

AI fits easily into that model because it can manage, rate, allocate, monitor, and discipline workers at scale while keeping formal accountability at a distance.

A person can be controlled by a platform, priced by an algorithm, penalised by a rating system, and still be told they are not quite an employee in the traditional sense.

The result was a system where:

data became a commodity
people became resources
public services became markets
corporate actors shaped policy
accountability became optional

In such a system, any powerful technology becomes risky – not simply because of what it can do, but because of where it lands, who controls it, and whose interests it is asked to serve.

This is why AI cannot be understood only as a technical issue. It is also a governance issue, an economic issue, and a question of power.

III. Why Social Media Gets Regulated Instead

Social-media regulation is politically convenient because it is visible, emotive, and easy to explain. It offers recognisable villains, clear examples of harm, and a public debate that fits neatly into news cycles.

It also focuses on behaviour more than power. It asks what people are allowed to say online, but less often asks who owns the systems, who profits from them, who audits them, or who is harmed when automated decisions spread into housing, work, finance, welfare, policing, health, and education.

That does not make social-media harms unimportant. It means they are easier for politics to confront than the deeper structural reforms that remain politically off-limits.

Meanwhile, the real issues – data governance, algorithmic accountability, labour displacement, surveillance, power concentration – remain untouched.

The same narrowness appears in the labour debate. We talk about productivity, automation, and skills, but less often about who carries the risk when platforms classify workers as flexible contractors while directing their work through software.

This is why the public debate can feel strangely narrow. We argue about harmful posts, but not about automated welfare decisions. We debate online speech, but not the ownership of the data used to classify citizens, workers, tenants, borrowers, and patients.

IV. The Political Class Was Not Selected for Systemic Responsibility

Most politicians do not see the contradiction clearly because the political system rarely selects for that kind of responsibility.

The Westminster pipeline rewards:

communication
loyalty
campaigning
message discipline

It does not reward:

systems thinking
regulatory literacy
long‑term governance
understanding of political economy
understanding of technology

These skills matter in politics, but they are not the same as governing complex systems. Winning power and using power responsibly require different capacities.

This helps explain the shock of office. Leaders may arrive with conviction, but then discover the scale of the machinery around them: private contracts, fragmented responsibilities, legacy systems, institutional inertia, and a political culture designed for message control rather than long-term stewardship.

The result is a politics that can describe crises fluently but struggles to rebuild the institutions needed to prevent them.

V. The Deeper Guardrails: Distance, Centralisation, and Dehumanisation

Beneath the political and economic layers lies a deeper shift: decision-making has moved further away from the people affected by it.

1. Centralisation created distance

When decisions were made locally, decision‑makers lived among those affected. They had to look consequences in the eye.

Centralisation – and later globalisation – changed that.

Now decisions are made:

in London about people in Cornwall
in New York about people in Newcastle
in Singapore about people in Sheffield
by algorithms about people they will never meet

Distance dissolves accountability. It allows decisions to be made without ever encountering the human cost.

That distance does not automatically make people cruel. It makes consequences easier not to see. And what is not seen is easier to ignore.

2. Globalisation hid the extraction

Globalisation dispersed responsibility.

It created a world where:

supply chains are opaque
ownership is labyrinthine
accountability is diffused
harms are exported
profits are centralised

Power became global. Accountability remained local.

3. The digital revolution turned distance into dehumanisation

Digital systems do not see people.

They see:

risk profiles
credit scores
behavioural patterns
demographic segments
optimisation targets

This is why exclusion can now happen instantly, automatically, invisibly, and without meaningful recourse.

A person may experience this as a rejected application, a higher insurance quote, a closed bank account, a fraud flag, or a risk score they are never allowed to inspect.

The language is technical, but the consequence is ordinary: life becomes harder, and no one is accountable.

This is the point at which distance becomes dehumanisation. The person is still there, but the system no longer has to encounter them as a person.

AI did not invent this dehumanisation. It accelerated it.

VI. Finance and Creditworthiness: The Quiet Architecture of Control

The financial system is one of the clearest examples of guardrails removed, because it already decides who can participate fully in society.

credit scoring is privatised
risk modelling is proprietary
trading algorithms operate beyond oversight
access to finance is controlled by private gatekeepers

Creditworthiness has become a quiet tool of social sorting.

For many people, creditworthiness now functions less like a narrow financial measure and more like a passport to ordinary life.

It can determine:

who can rent
who can buy
who can work
who can move
who can access services

Because the system is largely self-policing, people can be excluded without ever fully understanding why. The companies making those judgements can hide behind commercial confidentiality, proprietary risk models, or automated decision-making.

AI supercharges this exclusion – making it faster, more opaque, and more difficult to challenge.

VII. The State Now Subsidises the Extraction It Cannot Control

As life becomes more unaffordable, the state steps in with:

housing benefit
universal credit
tax credits
energy subsidies
childcare subsidies

But these are not only social protections. They are also, indirectly, subsidies for a system that extracts more from people than many can afford to lose.

The state is paying to keep people afloat in an economy designed to drain them.

This is why public spending rises even as public wellbeing falls.

When wages, rents, energy costs, childcare costs, debt, and insecure work pull in the same direction, the state is forced to compensate for the damage while leaving the underlying model intact. In effect, public support can end up cushioning a labour market where too much risk has been transferred from employers to workers.

VIII. The System Has Become Too Embedded to Correct Itself

This is the uncomfortable truth.

The system cannot be corrected by slogans, ethics panels, or narrow technical fixes alone. It is too embedded, too centralised, and too dependent on extraction to repair itself without deeper political choices.

Even many of the technology leaders driving the digital revolution express fear about where this is heading – yet the machinery continues, because the system rewards momentum more than restraint.

We have built a world where:

power is concentrated
accountability is diffused
decisions are automated
consequences are invisible
people are abstracted into data
profit outranks wellbeing

In such a world, AI is not a disruption. It is the logical next step.

That does not mean nothing can be done. It means the solution cannot be limited to regulating individual tools after they have already been deployed.

The deeper task is to rebuild the conditions under which powerful tools can be governed in the public interest.

IX. The Reckoning

The reckoning is uncomfortable because it reveals that today’s risks were not inevitable. They were created by choices – political, economic, and ideological – made over decades.

But discomfort is not despair. It is clarity. And clarity is the first step toward rebuilding the protections we dismantled.

AI forces this recognition because it compresses old failures into visible form. It makes weak accountability faster, opaque decisions broader, and distant power harder to challenge.

X. The Paradigm Shift We Need

We cannot regulate AI – or housing, finance, labour, welfare, education, health, or the environment – within a system that continues to prioritise extraction over wellbeing.

We need a shift from a money-centric model to a people-centric one. That must not mean another abstract slogan. It should be a practical test for every major decision: does this system increase human agency, democratic accountability, and material security, or does it simply make extraction more efficient?

That means rebuilding practical guardrails that people can feel in everyday life:

regulators with the capacity and independence to act
public expertise that is not permanently outsourced
democratic oversight of systems that shape people’s lives
data rights that give people visibility, control, and meaningful protection
financial accountability when credit, risk, or automated systems exclude people
local decision-making where proximity to consequences matters
institutional responsibility that cannot be hidden inside contracts or algorithms
clear rights of appeal when automated systems affect people’s homes, work, money, services, or freedom
employment protections that prevent firms from using status, platforms, or algorithmic management to transfer employer responsibilities onto workers

AI is not the problem. It is the test.

And it is showing us, with painful clarity, that the guardrails we need are the ones we removed long ago. Rebuilding them will require more than better software or better speeches. It will require institutions capable of seeing people again – and strong enough to act when they do.