The Global AI Race: Why No Country Can Hit the Brakes Alone

Imagine that one country decides to approach artificial intelligence more carefully.

It requires companies to evaluate the effects of large-scale automation before eliminating thousands of jobs. It strengthens transition assistance. It gives schools, workers and public institutions more time to adapt. It asks whether the gains from AI will be shared widely enough to preserve purchasing power and economic participation.

Now imagine that its competitors do none of those things.

The careful country may carry higher costs. Investment may move elsewhere. Domestic companies may lose market share to foreign firms that automate faster. Political leaders may be accused of sacrificing growth, technological leadership and national security.

The policy may be wise for humanity and dangerous for the country that adopts it alone.

It is not enough for a government to recognize a collective danger. It must also be able to respond without handing an advantage to every government that refuses to join it.

The Economy Is Global. Political Authority Is Not.

Companies sell across borders. Capital moves across borders. Supply chains cross borders. Software can be distributed internationally almost at once. The most advanced AI systems depend on networks of chips, data centers, energy, technical talent and investment that no single political jurisdiction fully controls.

But most laws still stop at a border.

A legislature can regulate employers in its own country. It can impose taxes, fund retraining, protect workers and shape its own social insurance system. It cannot require a foreign competitor to accept the same costs. It cannot prevent production from relocating. It cannot guarantee that another country will value human welfare, labor standards or democratic legitimacy in the same way.

That mismatch did not begin with AI. Globalization has exposed it for decades.

AI intensifies it because the technology is unusually portable, scalable and fast. A factory takes time to build. A railroad takes years. A powerful software capability can spread through an industry before a legislature has finished holding hearings about it.

The result is a structural temptation: each country may understand that shared restraint would be beneficial while fearing that unilateral restraint would be costly.

A Country Does Not Have to Be Reckless to Keep Accelerating

National leaders are responsible first to their own citizens. They worry about employment, growth, tax revenue and living standards. They also worry about military power, cybersecurity, scientific leadership and strategic dependence.

Once AI is understood as a source of economic and geopolitical power, slowing down becomes more than an economic decision.

If the United States moves cautiously while China accelerates, American leaders will fear losing strategic ground.

If China moves cautiously while the United States accelerates, Chinese leaders will fear the same thing.

If Europe establishes stronger protections while companies can build and operate elsewhere, European leaders must consider whether capital, talent and market power will migrate.

Smaller countries face a different version of the problem. Many do not control advanced chips, frontier models or the platforms through which AI reaches their economies. Yet their workers, tax systems and institutions may still absorb the consequences.

No government needs to believe that unlimited automation is desirable.

Markets Cannot Decide the Ethical Question

Markets are powerful systems for coordinating production and responding to demand. They are not moral authorities.

A market can tell a company that replacing workers will reduce costs. It cannot decide how much disruption a democratic society should accept, how quickly people should be required to adapt or what obligations institutions have to those whose livelihoods disappear.

Those are ethical and political judgments.

For generations, our economic ground rules have assumed that most adults will participate by selling their labor. Work distributes more than wages. It provides access to housing, healthcare, credit, retirement savings, social status and a sense of contribution. Taxes on work support public institutions and social insurance.

If AI reduces the amount of human labor required, the technology has not broken a law of nature. It has exposed a weakness in rules designed for another world.

Governments therefore cannot limit themselves to encouraging innovation and correcting isolated harms. They must ask whether the system connecting production, income and demand remains viable when machines perform more of the work.

And because the competitive pressure is international, the answer cannot remain exclusively national.

Every Job as It Exists Today Won’t Be Preserved

We should be clear about this.

The objective should not be to freeze the economy in place.

Many jobs will disappear.

Some should disappear.

Dangerous, exhausting and degrading work ought to be automated. AI can relieve people of tasks that damage their bodies, consume their lives or waste human potential.

The problem is not that AI can perform work.

The problem is that our economic system still makes employment the principal means by which most people obtain income, housing, healthcare, security and participation in society.

It does not buy groceries, raise children, replace a roof, visit a restaurant or save for retirement.

AI can add enormous productive capacity without receiving wages or creating household demand. When AI replaces a worker, the economy may gain production while losing part of the market that purchases that production.

That is the Paradox of Automation: decisions that are rational for individual firms can become destructive in the aggregate if the elimination of labor income weakens the demand on which those firms ultimately depend.

This outcome is not certain. The scale of job displacement remains unknown, new work may emerge and many occupations will be transformed rather than eliminated. The International Labour Organization estimated in 2025 that one in four workers worldwide held an occupation with some exposure to generative AI, while emphasizing that job transformation was more likely than full replacement for most occupations.¹

Uncertainty should prevent exaggerated claims.

It should not prevent preparation.

Do Not Remove the Ladder and Blame People for Falling

The simple answer, for many people, will be that those who struggle have only themselves to blame.

They should have studied something different.

They should have retrained sooner.

They should work harder.

They are lazy.

Sometimes personal choices do matter. Some people avoid responsibility. Any serious system must leave room for effort, accountability and contribution.

But personal failure cannot explain a structural shortage of opportunity affecting millions of people at the same time.

Even before large-scale AI displacement, many qualified and hardworking people describe the job market as brutal. If the number or quality of available jobs deteriorates, effort alone cannot create positions that employers no longer need.

Thomas More identified this failure of reasoning more than five centuries ago. In Utopia, he described people displaced from rural livelihoods by changes in land use. They were willing to work but could find no one to hire them. Society then condemned the begging and theft that followed.

More’s accusation was blunt: “you first make thieves and then punish them.”²

Society has an obligation to provide the education, tools, transition time, security and institutional platform that allow people to thrive.

That does not abolish personal responsibility.

It preserves the conditions in which personal responsibility can still produce a meaningful result.

The Greatest Problem May Be Speed

Economic change is not new. Agriculture displaced forms of labor. Industrial machinery transformed production. Computers eliminated some occupations and created others.

But historical analogy can conceal as much as it reveals.

Schools, laws, tax systems and social institutions change slowly. Workers need time to retrain. Communities need time to attract new industries. Families need time to rebuild savings after a lost income.

AI can move much faster.

A capable system can be copied, improved and deployed across companies and countries with extraordinary speed. Each successful deployment pressures competitors to respond. Each round of competition can shorten the time available for adaptation.

The question is not merely whether economies can eventually create new roles.

It is whether people and institutions can survive the interval between the disappearance of old forms of participation and the arrival of new ones.

That interval is where homes can be lost, savings depleted, skills devalued, communities weakened and political trust destroyed.

International Cooperation Has Already Begun—But the Economic Question Is Still Open

The world is not starting from zero.

The OECD AI Principles call for inclusive growth, sustainable development and well-being. The Bletchley Declaration states that many AI risks are inherently international and best addressed through international cooperation. The Council of Europe has created a legally binding framework centered on human rights, democracy and the rule of law. The United Nations’ Global Digital Compact calls for international governance that is inclusive and grounded in human rights.³⁴⁵⁶

These efforts matter. They establish principles, shared language and channels for cooperation.

But much of the international AI discussion still concentrates on model safety, security, privacy, discrimination, transparency and misuse.

Those issues are essential.

They are not the whole economic problem.

An AI model might behave exactly as designed while helping thousands of firms eliminate labor faster than societies replace income and demand.

No alignment failure is required.

No malicious actor is required.

The damage can emerge from ordinary competition operating under rules that reward each participant for moving faster.

What Cooperation Would Actually Need to Do

International cooperation does not require a single world government. It does not require every country to adopt identical laws. And it should not become an excuse for powerful countries to freeze poorer nations out of technological progress.

It requires enough common ground that protecting people does not automatically make a country uncompetitive.

None of these measures requires knowing the final scale of displacement.

They require acknowledging that uncertainty, speed and competitive pressure form a dangerous combination.

Cooperation Is Not the Same as Stopping Progress

The choice is often presented as innovation or regulation.

That is too simple.

Rules can slow harmful behavior while accelerating beneficial adoption. Governments can encourage AI in medicine, science, education, energy, accessibility and dangerous work while demanding more care when deployment threatens to remove income from entire occupational groups.

The point is not to preserve every job.

The point is to preserve broad purchasing power, social stability and a credible path to participation in the wealth AI creates.

This is also why the question is larger than digital technology. Fusion energy, advanced robotics and future scientific breakthroughs may alter production in ways we cannot yet predict. The governing principle should survive the arrival of any particular tool:

The Race We Actually Need to Win

A company cannot completely isolate itself from its competitors.

A country cannot completely isolate itself from global trade.

And humanity cannot indefinitely treat AI competition as though every participant’s private incentives will automatically produce a desirable collective result.

Perhaps AI will create more work than it destroys.

Perhaps productivity will raise living standards broadly.

Perhaps new institutions will emerge in time.

We should work toward those outcomes.

But hope is not a transition policy.

If AI changes the role of human labor, the adjustment should be shaped through evidence, experimentation and democratic choice.

Perhaps the real race is not to build the most powerful AI first.