What if we got rid of all politicians? The (not so crazy) case for AI-driven governance

From envelopes under the table to the world’s shadiest power networks, the verdict is always the same: Those who govern us seem to play in a league of their own—the league of impunity. We could talk about the usual backroom deals: ghost representatives, ministers shifting comfortably to the private sector, covered-up conflicts of interest. But the real dizzying realization comes when you look up to the top.

The Panama Papers, followed by the Pandora Papers, exposed heads of state hiding fortunes in tax havens while preaching austerity to their people.

And then there’s the Epstein affair: a network of sexual trafficking involving minors, protected for decades by presidents, princes, and billionaires.

With every scandal, the pattern repeats: outrage, commission, oblivion. The faces change. The system remains.

The numbers are harsh. According to the Edelman Trust Barometer 2025, distrust toward political leaders is breaking historical records across 28 countries. In France, the trust gap between the working classes and the wealthy has gone from 5 to 14 points in twelve years.

The CEVIPOF regularly finds that less than 30% of French citizens trust their elected officials. We could list more statistics, but you get the point: something is broken.

So, faced with this reality, a question increasingly pops up in conversations—sometimes as a joke, sometimes not. What if we replaced all of them with artificial intelligence?

No corruption, no careerism, no promises forgotten the day after the election. A machine processing data, optimizing decisions, and just letting us be. Geeky utopia or a serious idea?

The Tempting Argument: Why AI (In Theory) Would Do a Better Job

Let’s take the idea seriously. Because it’s not as outlandish as it sounds.

A general artificial intelligence (AGI) that is, an AI capable of reasoning on any subject as well as, or better than, a human—offers undeniable theoretical advantages for governance.

First, no personal agenda. A machine doesn’t worry about re-election, nepotism, or landing a cushy private-sector gig. It doesn’t have an offshore account in the Bahamas.

Its data processing capacity dwarfs that of any ministerial cabinet. Imagine: analyzing economic flows, demographic trends, public health data, and environmental indicators in real time.

Synthesizing millions of pages of expert reports. Modeling the impact of a tax reform over twenty years. Where a human might need months to produce an approximate impact study, an AGI could do it in a few hours.

An AI doesn’t make decisions based on the next poll or the next election. It can think long-term.

Add in the absence of classic cognitive biases: no groupthink, no magical thinking, no denial in the face of bad news.

And above all, decisions based on evidence (evidence-based policy), not ideology or electoral calculation.

This isn’t pure theory. Forms of algorithmic governance already exist. Estonia has digitized 99% of its public services and relies heavily on administrative automation. The smart cities of Singapore and Dubai entrust algorithms with managing traffic, energy, and sometimes social assistance.

In the crypto world, DAOs (Decentralized Autonomous Organizations) run on coded rules, with no human leadership.

The Turning Point: Why It’s More Complicated Than It Seems

The idea is appealing. But let’s dig deeper.

Who Programs the AGI?

First problem, and a big one: AI doesn’t just appear out of thin air. It’s designed, trained, and fine-tuned by humans. Humans with their own biases, blind spots, and interests. Corruption doesn’t disappear; it just shifts.

The real-world examples are plentiful. The Dutch system SyRI, used to detect welfare fraud, was ruled illegal in 2020: it discriminated against disadvantaged neighborhoods.

In the US, the COMPAS tool used in criminal justice was found to be twice as likely to falsely classify Black defendants as “high risk” for re-offending. The machine wasn’t neutral; it simply reproduced the biases of its creators and training data.

Over 80% of business leaders believe AI-driven decisions will be more precise than human decisions within 10 years. But less than 30% have implemented formal oversight processes. The gap between confidence and caution is staggering.

Governing Is Not the Same as Optimizing

Second problem: politics isn’t an engineering problem. It’s about balancing conflicting values.

Should we prioritize economic growth or environmental protection? Individual liberty or collective security? Efficiency or fairness? These questions don’t have an “optimal” answer hidden somewhere in the data. They are societal choices, based on different worldviews.

An AGI would need an objective function: maximize what, exactly? GDP? Gross National Happiness (good luck measuring that)? Life expectancy? Income equality? Every choice of metric creates winners and losers. And who decides on this objective function? Silicon Valley engineers? A self-appointed ethics panel?

Replacing politicians with a machine doesn’t remove the question of power; it shifts it to those who control the machine.

Democracy as a Process, Not Just a Result

Third problem, the more philosophical one: reducing democracy to just a machine for producing “good decisions” misses its essence.

Democracy is the right to disagree. The right to collectively change our minds. The right, sometimes, to inefficiency because consensus takes time, because minorities deserve to be heard, because we don’t want a society that “optimizes” people as variables.

The twentieth century is full of failed “government by experts” experiments. The technocracy of the 1930s promised scientific management of society which often resulted in authoritarianism.

Soviet planning embodied the idea that a central bureau, armed with enough data, could run an entire economy. The result: chronic shortages, falsified statistics, and a system that collapsed from within.

The Real Problem Behind the Fantasy

Let’s put the question differently. Why does the idea of “firing all politicians” resonate so much?

Not because people actually want to be ruled by machines. But because they feel the current system no longer works.

That elected officials don’t represent them. That decisions are made elsewhere—down the corridors of Brussels, in boardrooms, in meetings they’ll never attend.

This desire points to a crisis of representation, not a need for AI. And that distinction is key.

“Politicians are useless” and “the system produces bad results” are two different diagnoses, which require different remedies.

Alternatives That Don’t Throw the Baby Out With the Bathwater

If the problem is representation, solutions exist and they don’t mean abolishing human politics.

Citizens’ assemblies allow randomly selected panels of citizens to deliberate on complex topics.

France’s Climate Citizens’ Assembly showed that “ordinary” people, when well informed, can produce sophisticated, ambitious recommendations.

The problem wasn’t the quality of their proposals, but the lack of political will to implement them.

Random selection for some positions (as in ancient Greece) could renew representation and break the monopoly of career politicians.

Algorithmic transparency in public decisions: Explaining how benefits are calculated, how resources are distributed, would address the sense of opacity.

And AI in all this? It can be a powerful decision-support tool. Simulating the effects of a policy before implementing it.

Finding correlations invisible to the human eye. Processing administrative paperwork faster. But as a tool—not as a decider. That difference isn’t cosmetic; it’s fundamental.

The Real Question

The appeal of AGI-led governance acts as a mirror of our democratic exhaustion. It says something about our fatigue with scandals, broken promises, and the feeling of powerlessness. And this fatigue is legitimate.

But the answer to this exhaustion cannot be to hand over our collective decision-making to a machine.

Because self-government isn’t the problem it’s the project. An admittedly difficult, frustrating, imperfect project. But it’s ours.

FAQ

Can an AI really be totally neutral and unbiased?

No. Every AI is trained on data created by humans and designed by teams making choices. The biases of creators and data inevitably show up in the outcomes. That’s why human oversight remains essential.

Are there any countries already using algorithms to govern?

Yes, but only in limited ways. Estonia relies heavily on digital technology in its public services. Singapore and other smart cities entrust some decisions (traffic, energy) to algorithms. No country has replaced its elected officials with AI.

What exactly is an AGI?

AGI (Artificial General Intelligence) refers to an artificial intelligence capable of reasoning about any subject at (or beyond) human level. It does not yet exist—today’s AIs are “specialized” (playing chess, generating text, etc.).

Why do experts compare AI government to Soviet planning?

Because both rely on the idea that a central entity, given enough data, can make better decisions than decentralized processes. The Soviet failure shows the limits of this approach when confronted with real-world complexity.

What’s a DAO and how does it work?

A DAO (Decentralized Autonomous Organization) is an organization whose rules are coded on a blockchain. Decisions are made by token holders’ votes—with no traditional human leadership. It’s an experimental form of automated governance.

What was the SyRI case in the Netherlands?

SyRI was an algorithm used to detect social welfare fraud by cross-referencing administrative data. In 2020, a court ruled it illegal because it discriminated against poor neighborhoods and violated privacy rights.

What are the arguments for using sortition in politics?

Sortition (random selection) would mean representatives are closer to the actual population, break the monopoly of career politicians, and reduce clientelism. It was the main method in Athenian democracy.

Could an AI at least help politicians make better decisions?

Absolutely. AI can simulate public policy effects, analyze huge amounts of data, and identify trends. Used as a decision-support tool (not as the decision-maker), it has huge potential.

Why do we say politics is about value trade-offs and not a technical problem?

Because major political questions (freedom vs. security, growth vs. environment) involve societal choices based on what we value. There’s no “objectively correct” answer for data to reveal.

Is distrust of politicians really that strong?

Yes. The Edelman Trust Barometer 2025 describes a “slide into resentment” in 28 countries studied. In France, CEVIPOF surveys consistently show that less than a third of citizens trust their national representatives.