Who Governs the Machine When States Compete to Own It?

Whenever the world talks about regulating artificial intelligence, the discussion often seems to assume that the problem lies in the technology itself: machines that learn rapidly, programs that make decisions, and systems whose capabilities may, in some areas, exceed the human ability to fully understand or predict them.

But the real problem may lie somewhere else entirely.

Who has the power to make the rules for a technology that has become part of the economic and strategic power of states?

That is what makes British leader Andy Burnham’s call from the United Nations for international safeguards on artificial intelligence more significant than simply another appeal to regulate technology. Burnham has proposed that the United Nations serve as a platform for international consensus, while the Group of Twenty provides a more practical framework for developing common standards, with individual governments responsible for implementing them domestically.

The idea sounds reasonable. But international politics does not always operate according to what seems reasonable.

Artificial intelligence is no longer simply another emerging technology sector that can be regulated like any other industry. It has become intertwined with the economy, defense, security, scientific research, labor markets, and the international standing of states. Talking about regulating AI therefore means, in one way or another, talking about regulating one of the sources of power in the twenty-first century.

And that is where the dilemma begins.

The World Wants Rules—But It Does Not Want to Lose the Race

The international community may find it relatively easy to agree that artificial intelligence needs rules. Agreement becomes much harder, however, when the questions become: What should those rules be? Who should comply with them? Who should monitor compliance? And what happens to those who do not?

Washington and Beijing offer the clearest example.

While Burnham has called for common rules, U.S. President Donald Trump has rejected international restrictions that he believes could constrain the development of American artificial intelligence. Chinese President Xi Jinping, meanwhile, invoked the concept of the “Thucydides Trap” during his meeting with Trump, referring to the risks of rivalry between an established power and a rising one.

Artificial intelligence may not itself be a direct cause of confrontation between the United States and China, but it has undoubtedly become one of the arenas of competition between them.

That helps explain why reaching strict international rules is so difficult.

A country that imposes tighter restrictions on its companies may find itself competing against firms in other countries operating with greater freedom. A company facing additional compliance costs may lose part of its competitive edge. And a government that slows technological development in the name of caution may fear that it has handed its rivals an advantage that cannot easily be recovered.

The problem, therefore, is not that governments fail to understand the risks of artificial intelligence.

The problem is that they also understand the cost of falling behind it.

An Impossible Equation

Governments want to protect their citizens from job displacement, violations of privacy, manipulation of information, algorithmic errors, and dangerous military and security applications.

At the same time, they want investment.

They want data centers.

They want companies.

They want new jobs.

And they want to be part of the economy that artificial intelligence will shape, rather than becoming merely consumers of technologies produced elsewhere.

Governments thus face an equation that is difficult to solve on both sides: regulate the technology without suffocating innovation, while encouraging innovation without allowing the technology to become a force without oversight.

Even the environment has become part of the equation. Data centers require enormous amounts of electricity and water, while governments are simultaneously under growing pressure to meet energy and resource goals.

Regulating artificial intelligence, therefore, is not simply a battle between “supporters” and “opponents” of technology.

It is a struggle between legitimate but competing interests.

The Problem That Has Yet to Be Solved: Responsibility

But there is a question more important than all these economic calculations:

What happens when artificial intelligence gets it wrong?

If an automated system makes a decision that causes harm, who is responsible?

Is it the company that developed the system? The programmer? The executive who authorized its deployment? The institution that used it? Or the person who relied on its output?

This is not a marginal legal issue.

It lies at the heart of the entire debate.

The most important rule may not be telling the machine, “Do not do this.”

It may be determining in advance: Who has the authority to stop it? And who bears responsibility for failing to stop it?

In traditional business systems, responsibility can often be identified with reasonable clarity. Complex intelligent systems, however, may make a decision the result of a long chain of design, training, testing, and deployment, making it easy, after harm occurs, for everyone involved to shift responsibility elsewhere.

This is where meaningful regulation should begin: not with punishment after the damage has been done, but with defining responsibility before it happens.

Technology Moves Faster Than the Law

There is another problem that makes the United Nations’ task even more difficult.

Laws move slowly.

Technology moves at an entirely different speed.

The nuclear warheads and missiles that major powers sought to regulate during the Cold War were physical objects that could be counted, located, and inspected.

An AI model is not a facility that can simply be marked on a map.

It can be developed in one country, trained on data from another, operated online in a third, and updated weeks after the law governing it was enacted.

Its capabilities may change before governments have even finished agreeing on how to define it.

This creates a problem that traditional treaties cannot easily solve:

How do you write a rule for a technology that changes faster than the rule itself?

Frankenstein Is Not the Problem

More than two centuries ago, Mary Shelley wrote Frankenstein; or, The Modern Prometheus.

Since then, the story of a scientist who creates a being and ultimately loses control over it has become a recurring metaphor for humanity’s fear of science.

But perhaps the more accurate reading of the novel is different.

The problem was not that Frankenstein wanted to know.

The problem was that he failed to think sufficiently about his responsibility for what would happen after he acquired that knowledge.

That is precisely what makes the novel relevant to the age of artificial intelligence.

The question is not whether we should fear the machine, nor whether machines will rebel against humanity.

The more realistic question is:

Will we establish a system of responsibility before we give technology greater authority to make decisions?

If the answer is no, the problem will not be artificial intelligence itself, but the institutions that allowed it to operate without clearly defining who oversees it and who is accountable for its consequences.

There Is Another Danger: Protection Can Become Surveillance

The world must also pay attention to the other side of regulation.

Rules designed to protect people can, if poorly designed, become tools for monitoring them.

The story may begin with protecting society from harmful content and end with determining what citizens are allowed to search for.

It may begin with protecting privacy and end with broader surveillance of data.

It may begin with regulating algorithms and end with granting governments greater authority to determine what people are allowed to know.

That is why the debate over regulating artificial intelligence must include another question of equal importance:

Who regulates those who make the rules of regulation?

It is not enough for technology to be subject to the law. The law itself must also be subject to clear principles that protect rights and freedoms.

Can the United Nations Do It?

The United Nations can provide something no single country can provide: international legitimacy, a platform for dialogue, and a framework within which states with different interests can meet.

But it cannot, by itself, resolve the fundamental contradiction.

International rules will have limited value if major powers do not agree to implement them, if governments cannot translate them into domestic law, or if companies can simply move to jurisdictions with less restrictive regulatory environments.

This is why the Group of Twenty may have a more practical role to play in the initial stages. Its members control a large share of the capital, technology, and markets that will shape the future of artificial intelligence.

Yet even the G20 will face the same question: Can governments accept restrictions that might weaken their ability to compete?

That is where the future of the initiative will be determined.

Human First—but What Does That Mean?

The slogan of “keeping humans in control” may be compelling, but it requires a more precise definition.

Does it mean that a human must review every decision made by an AI system?

That is impossible in many applications.

Or does it mean that decisions affecting people’s lives and fundamental rights must remain subject to meaningful human oversight?

If that is what it means, then the next step should be to define those decisions clearly, determine the level of oversight required, give individuals the right to challenge decisions, and establish clear legal responsibility for those who deploy intelligent systems.

Only then can the phrase “keeping humans in control” become more than an ethical slogan. It can become a political and legal principle.

The Race Will Not Stop

The race for artificial intelligence will not stop because the United Nations adopts a resolution. Nor will states abandon its development because some of its applications carry risks.

And that may be the most important lesson.

The objective should not be to stop the future, but to prevent the future from arriving before we have rules capable of governing it.

The world does not need to choose between innovation and regulation.

It needs a model that allows both.

Companies should continue to innovate, but with clear accountability.

States should benefit from the technology without turning it into an instrument of unchecked control.

Citizens should have access to smarter services without losing the right to know who made the decision that affected their lives.

In the end, the most important question may not be:

Can the United Nations regulate artificial intelligence?

It may instead be:

Can the states competing to control artificial intelligence agree that they must not lose control of it?

The real danger is not that machines will become more intelligent than humans.

It is that humans may become better at creating machines than they are at taking responsibility for what those machines create.