In this world Who Gets to Decide What Artificial Intelligence Can Do?
A few years ago, we were still amazed that a computer could write meaningful sentences. Today, we have reached the point where artificial intelligence (AI) selects CVs, creates advertisements, analyses medical reports, and sometimes gives a wrong answer with such confidence that even an experienced politician would envy it.
Technology is developing at an incredible speed, while lawmakers are trying to keep up with it. The problem is that while a new AI model can be launched on the market within a few months, adopting a more serious piece of legislationcan take years.
Moreover, there is not even agreement on exactly what should be regulated. The technology itself? The developers? The users? Or should authorities only intervene when something has already gone wrong?
And this is where the story starts to get really interesting.
Technology Is Global, but the Law Is Not
Imagine that an American company develops an AI-powered recruitment system. The software is purchased by a German company, which uses it to evaluate applications from Hungarian workers. Meanwhile, the data is processed on a server located in another country.
For some reason, the system regularly ranks female applicants lower based on flawed criteria.
Who is responsible?
The American developer? The German employer? Or perhaps the company that processes the data?
And which country's authorities should handle the case?
In such a situation, it is no longer enough to look at the legislation of a single country. Technology easily crosses national borders, while legal systems still operate largely within national frameworks.
This is one of the biggest problems with regulating AI. Companies develop products for international markets, while states create their own rules, often based on completely different ideas.
Europe Prefers to Think Ahead
The European Union decided not to wait until AI starts causing problems on a massive scale. It created the regulation known as the AI Act, which was adopted in 2024.
The main idea behind the regulation is that the greater the risk posed by a system, the stricter the requirements it must meet.
At first, this sounds perfectly logical. A chatbot operating on a restaurant's website obviously does not pose the same danger as an algorithm that makes decisions about people's creditworthiness or employment opportunities.
In the latter cases, a system error can have serious consequences, especially if the system learns from data that already contains biases.
The EU regulation therefore prohibits certain applications, while requiring documentation, monitoring, risk management and human oversight for others.
Of course, there is another side to this.
A multinational technology company has its own legal and compliance teams. A small European startup with only a few employees, however, may easily find itself spending more time and money understanding the rules than developing its own product.
Well-intentioned regulation may therefore end up benefiting large corporations. Not necessarily because lawmakers want this, but because smaller businesses find it harder to meet the same requirements.
America Does Not Want to Over regulate the Market
The United States approaches the issue differently. Traditionally, businesses have been given more room to operate, and AI-related rules also appear across several different areas of law.
If an advertisement created using artificial intelligence misleads consumers, consumer protection rules may come into play. If an algorithm discriminates against someone, anti-discrimination legislation may apply.
Individual states can also introduce their own AI regulations.
This has one obvious advantage: there is no need to create an entirely new regulatory system for every new technological solution.
At the same time, requirements may differ even within the same country. Something that is permitted in one state may already be subject to additional obligations in another.
There is also a very real economic question behind the American debate. If overly strict regulations are introduced, could American technology companies find themselves at a competitive disadvantage?
This is particularly important at a time when the development of artificial intelligence is no longer just a business opportunity, but also a matter of international economic competition.
China Would Rather Keep Things Under Control
In China, the state takes a much more direct role in regulating technology.
Chinese lawmakers have introduced specific rules concerning algorithmic recommendation systems and generative artificial intelligence, for example. Data security and the control of online content also play an important role in the regulatory framework.
This model clearly shows how differently countries think about the same technology.
While Europe places considerable emphasis on protecting fundamental rights, and America focuses on innovation and market competition, China closely connects technological development with the state's strategic objectives.
Of course, these are not exclusive priorities. Europe also wants competitive technology companies, America also deals with fundamental rights, and China is equally interested in innovation.
The difference lies more in which interest takes priority when these objectives conflict.
What If the Whole World Followed the Same Rules?
The idea seems obvious: if artificial intelligence operates worldwide, why shouldn't it be regulated in the same way everywhere?
UNESCO, which is associated with the United Nations, already adopted a recommendation on the ethical use of AI in 2021. The OECD has also developed principles supporting trustworthy and responsibly operated artificial intelligence.
In 2024, the Council of Europe adopted the Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law.
These are important steps, but there is a significant difference between recommendations and binding legal rules.
It is relatively easy to agree in an international recommendation that AI should be safe, transparent and fair. The difficulty begins when it becomes necessary to define exactly what these requirements mean, who checks compliance with them, and what happens if someone violates them.
Some countries consider stricter state supervision necessary, while others fear that this would restrict technological development too much.
Creating a worldwide AI authority is therefore currently more of an interesting idea than an easily achievable plan.
If AI Makes a Mistake, Who Pays the Price?
In the middle of regulatory debates, we tend to forget that mistakes made by artificial intelligence can have very real consequences.
Suppose a bank's AI-based system unjustifiably rejects someone's loan application. Or an online store's advertisement created using artificial intelligence promises certain qualities about a product that it does not actually have.
The person affected is hardly going to feel reassured by the explanation that the mistake was caused by a complicated algorithm.
What matters to them is whom they can turn to, who will investigate what happened, and how they can obtain a legal remedy.
In my opinion, this is where every serious AI regulation should begin.
Just because a decision is made partly or entirely by artificial intelligence, human or corporate responsibility cannot simply disappear behind it.
Developers must be responsible for fulfilling the obligations that apply to them, while businesses using these systems must be responsible for how they use the technology.
Of course, the exact allocation of responsibility depends on the circumstances of each case. One thing, however, is certain: AI must not become a convenient excuse for nobody having to take responsibility for the consequences.
Do We Even Need Someone to Supervise
AI Worldwide?
At first, the idea of an international organisation supervising artificial intelligence around the world sounds good.
But a few uncomfortable questions immediately arise.
Who would appoint its leaders? What powers would it have? Could it impose obligations on American, Chinese or European companies? And what would happen if a country simply refused to accept its decisions?
Technological development has now become an economic and national security issue as well. It is difficult to imagine the major powers voluntarily handing over a significant part of their control to a common institution.
It is more likely that national rules, regional regulations and international agreements will continue to operate alongside one another in the coming years.
This is not necessarily a bad thing. It is not certain that every country needs to apply exactly the same rules.
However, it would be good to agree that people's rights must be protected, serious risks must be addressed, and legal violations must have real consequences.
In the End, It Is Still People Who Matter
In debates about artificial intelligence, we often talk about technology as if it were an independent actor that somehow needs to be kept under control.
Yet these systems are developed by people, sold by companies, and people or institutions decide how they are used.
That is why the question is not simply who regulates AI, but also who takes responsibility for its use.
I do not believe that any single country or international organisation could provide an adequate answer to this on its own. The differences between the European, American and Chinese approaches will probably remain with us for a long time.
However, common rules are necessary, particularly where the use of AI affects fundamental rights or where the authorities of a single country can no longer act effectively.
Technological development can hardly be stopped, and it does not necessarily need to be. The role of the law is rather to draw clear boundaries and to work even when something goes wrong.
Because if someone suffers harm due to a mistake made by artificial intelligence, the most important question will not be how advanced the algorithm was.
It will be whether there is someone who takes responsibility for it.
References
