AI Is a Gift to People Who Want to Leave the System

The most political thing about artificial intelligence may not be what it allows governments to do. It may be what it allows individuals to stop needing.

The most boring way to describe artificial intelligence is to say that it makes people more productive.

It writes faster. It translates faster. It generates code. It edits images. It summarizes documents. It answers customers at three in the morning. It can help a mediocre marketer produce competent copy, a competent programmer enter an unfamiliar stack, or a tiny company imitate some of the administrative surface area of a much larger one.

All of this is true, and all of it misses the more interesting change.

AI is reducing the minimum organizational size required to leave a system.

For most of modern economic history, exit was expensive. Not merely physical exit—the act of buying a plane ticket, obtaining a visa, and crossing a border—but economic exit: reducing the number of things in your life that must pass through the institutions immediately surrounding you.

Imagine a programmer living under an authoritarian government twenty years ago. Suppose he disliked the political system, distrusted domestic companies, wanted his income outside the local economy, and dreamed of building something for customers he had never met in another country.

What did he need?

English, probably. Foreign business knowledge. Designers. Marketing. Customer support. Access to information that might not exist in his own language. Someone who understood contracts. Someone who understood how foreign consumers behaved. Someone who could write sales copy without sounding ridiculous. Perhaps a local partner. Perhaps an employer large enough to absorb all these costs.

None of these barriers was individually insurmountable.

Together, they constituted a wall.

That wall is becoming thinner.

A single person can now sit in a room and converse with software in one language while producing work in another. He can inspect an unfamiliar market, prototype a product, write its interface, generate graphics, draft documentation, answer customer emails, analyze feedback, prepare marketing material, and learn the vocabulary of a foreign legal or commercial environment without assembling a conventional organization first.

The significance of this change is easy to underestimate because every individual task looks trivial.

Translation became cheaper.

Design became cheaper.

Programming became cheaper.

Research became cheaper.

Coordination became cheaper.

But when many small organizational costs collapse simultaneously, the result is not simply cheaper production. The viable size of an organization changes.

And once that happens, the politics changes too.

The First Exit Is Organizational

People tend to imagine leaving a political system geographically.

There is a border. On one side, you are inside. On the other, you are outside.

Real life is much messier.

A person can move abroad while remaining economically dependent on the country he left. His customers may still be there. His social network may still be there. His employer may still be there. His information diet may remain almost entirely domestic. His assets may remain trapped in institutions governed by the same political authorities.

Conversely, someone can remain physically where he was born while gradually moving pieces of his economic life elsewhere.

The first exit is not necessarily geographic.

It is organizational.

You stop requiring a domestic employer.

Then perhaps you stop requiring domestic customers.

Then you discover that you can communicate professionally in a language you barely spoke before.

Then your collaborators are scattered across several countries.

Then the software you use, the people who pay you, the information you consume, the company you work for, and the markets whose judgment determines whether your work succeeds no longer occupy the same jurisdiction.

Nothing cinematic has happened.

Nobody climbed over a wall at midnight.

But the dependency graph has changed.

That phrase matters more than the passport.

Authoritarian systems are often discussed primarily through prohibition: what they censor, what they ban, whom they arrest, which websites they block. Yet durable systems of control depend on something quieter than prohibition. They depend on necessity.

It is much easier to control someone who needs institutions you control.

If employment, customers, payments, professional reputation, information, language, distribution and business administration all converge inside one political environment, formal freedom to leave can coexist with enormous practical difficulty in leaving.

The person is not held by one chain.

He is held by a thousand inconveniences.

AI attacks some of those inconveniences.

Not dramatically. Not heroically. Often not even intentionally.

It simply makes fewer other people necessary.

The One-Person Institution

Silicon Valley has spent years talking about the one-person unicorn, usually with the intellectual depth one expects from a phrase designed to excite venture capitalists.

The valuation is the least interesting part.

The interesting question is what happens when one person acquires capabilities that recently belonged primarily to organizations.

A corporation is partly a way of assembling specialized cognition. One employee understands accounting. Another understands software. Another writes. Another researches. Another speaks to customers. Another turns a rough idea into a presentation that strangers can understand.

The corporation coordinates these people and pays the transaction costs between them.

AI does not eliminate specialization, and the fantasy that everyone will become a perfectly self-sufficient genius is nonsense. But it can compress the lower and middle layers of many specialties into something an individual can access on demand.

That is a different kind of leverage.

The important comparison is not:

one worker with AI versus one worker without AI.

It is:

one individual with AI versus the minimum organization previously required to accomplish the same objective.

That difference explains why the technology may matter disproportionately to people who begin outside the centers of global opportunity.

If you already speak the dominant commercial language, live near capital, possess a prestigious professional network, understand the legal system around you, and can hire specialists whenever you encounter something unfamiliar, an AI assistant is useful.

If you possess none of those advantages, it can be something closer to infrastructure.

Language is the easiest example.

English has long functioned as one of the admission tickets to the international economy. A brilliant engineer with weak English could be far less internationally mobile than a mediocre professional who grew up speaking it. Language ability affected not only communication but confidence, access to documentation, negotiation, hiring, sales and the ability to understand what was happening outside one's own information environment.

Machine translation existed before generative AI, of course. What changed is the scope of mediation.

The software does not merely translate a sentence. It can explain why a sentence sounds strange, rewrite it for a customer, summarize an unfamiliar discussion, help prepare for a meeting, interpret jargon, transform rough thoughts into professional prose and maintain a conversation across a linguistic boundary.

It does not make language irrelevant.

It makes imperfect language less fatal.

That is an enormous distinction.

The Authoritarian Paradox of AI

Much has been written about AI as an authoritarian technology, and there are good reasons for the concern.

The same class of technologies can improve surveillance, censorship, propaganda, bureaucratic classification and automated policing. Governments can use machine intelligence to process more information about more people at lower cost.

But technologies rarely have only one political vector.

AI also distributes capabilities that governments and large organizations once possessed because only they could afford enough people to perform them.

This creates a strange race.

The state becomes better at watching.

The individual becomes better at operating.

The state becomes better at processing information.

The individual becomes better at accessing and interpreting information beyond his immediate environment.

The state can automate parts of bureaucracy.

The individual can automate parts of the bureaucracy he previously needed an institution to provide for him.

For open societies, this is mostly an economic transformation.

For closed systems, it can become a problem of dependency.

The Chinese Communist Party offers an unusually clear example because its political order has never depended only on censorship or police power. It also benefits from the concentration of ordinary life inside institutions it can ultimately pressure: domestic platforms, banks, employers, payment networks, universities, media systems and commercial intermediaries.

That does not mean every institution receives daily instructions from the Party. Such a caricature would misunderstand how modern authoritarian systems work.

Control does not require manually operating every node.

It requires retaining sufficient authority over the nodes that matter when pressure becomes necessary.

For an individual, therefore, autonomy is not binary. It accumulates as dependencies become replaceable.

AI cannot give someone a passport. It cannot guarantee a foreign bank account. It cannot repeal capital controls. It cannot erase tax obligations, immigration law, financial regulation or the physical coercive power of a state.

But those are not the only costs of exit.

There are also cognitive costs.

Linguistic costs.

Coordination costs.

Knowledge costs.

Professional costs.

Administrative costs.

And these are precisely the categories in which AI is becoming unusually powerful.

Exit Before Emigration

This is why the most interesting AI story in authoritarian societies may eventually have surprisingly little to do with dissidents.

The important person may simply be someone who wants another life.

Not a revolutionary.

Not an activist.

Not someone interested in standing in a square with a sign.

Perhaps she wants to sell software to Americans.

Perhaps he wants to make a game for Japanese players.

Perhaps they want to run a small design studio whose clients happen to live on four continents.

Perhaps someone is simply tired of discovering that every career path eventually leads back to the same handful of companies, platforms and political constraints.

Historically, people in this position faced an ugly choice.

Stay inside the local system and gain access to its networks, or leave those networks and accept a severe loss of capability.

AI weakens that tradeoff.

A person can begin building the outside before physically arriving there.

That changes migration itself.

The old model often looked like this:

leave → rebuild → integrate → regain professional capacity.

A new model can increasingly look like:

build external capacity → build external income → build external relationships → leave when leaving becomes useful.

In other words, emigration can become the final stage of exit rather than the first.

That is a profound reversal.

And it is especially relevant to people living in countries where the state has spent decades making itself difficult to bypass.

AI Does Not Free You

There is a temptation to turn every new technology into a liberation myth. The internet received this treatment. Social media received it. Cryptocurrency received it. Each was supposed to dissolve old concentrations of power, and each eventually produced new concentrations of power alongside the freedoms it created.

AI will not be different.

Model providers can become centralized gatekeepers. Governments can regulate compute. Platforms can restrict access. Payment systems can still identify users. Cloud services can terminate accounts. AI-generated mediocrity can flood markets faster than talented individuals can escape them.

And the physical world remains stubbornly physical.

Borders exist.

Police exist.

Banks exist.

States exist.

If a government decides that money cannot move, a chatbot cannot negotiate with the central bank on your behalf and magically make the restriction disappear.

So AI is not freedom.

That is exactly why it is useful to define its contribution more narrowly.

AI reduces the organizational cost of autonomy.

It gives individuals access to fragments of capabilities that once required teams, institutions, credentials, linguistic privilege or proximity to professional centers.

Sometimes those fragments will be inadequate.

Sometimes they will be wrong.

But the threshold keeps falling.

And thresholds matter enormously in human behavior.

Millions of people may want something while doing nothing because the first step is too expensive. Reduce the first step enough, and you do not need to change their beliefs. You change what their existing beliefs can produce.

The System Becomes Optional One Dependency at a Time

The twentieth century taught people to imagine political power through giant organizations: states, armies, parties, factories, unions, television networks and multinational corporations.

The early internet weakened some of those monopolies by making information and publication cheap.

AI may push the process somewhere more intimate.

It makes competence cheaper to assemble.

That sounds less revolutionary than a manifesto. It is probably more consequential.

A closed system does not become irrelevant on the day everyone escapes from it. It becomes weaker in the life of a particular person each time that person discovers another function for which the system is no longer indispensable.

First, perhaps, he no longer needs it to tell him what is happening outside.

Then he no longer needs it to translate the outside world.

Then he no longer needs a large domestic company to turn his skill into something sellable.

Then he no longer needs all his customers to live nearby.

Then his professional identity stops being determined exclusively by local institutions.

None of these steps constitutes liberation by itself.

Together they create optionality.

And optionality is dangerous to systems built on the assumption that ordinary people have nowhere practical to go.

That may ultimately be AI's most important gift to people who want to leave.

It does not open the gate for them.

It makes more of what lies behind the gate portable.

And long before a person crosses a border, the system may discover that some part of him has already left.