As artificial intelligence (AI) becomes increasingly widespread and embedded in decisions involving credit, investment, production, financial markets and infrastructure, managing the risks associated with its use is becoming more important than ever.
The debate surrounding artificial intelligence is no longer confined to its ability to write texts, analyse data or help people perform tasks more quickly. The world has entered a new phase in which AI is gradually moving to the heart of economic decision-making. Banks use it to assess credit applications and detect fraud, companies rely on it for pricing and production, financial markets deploy algorithms for trading, and institutions use it to analyse risks and inform their decisions.
The most important economic question, therefore, is no longer simply what artificial intelligence can do, but what happens when the economy itself becomes increasingly dependent on it.
This question has become more pressing as AI capabilities advance rapidly and governments and major companies around the world turn greater attention to safety, oversight and governance.
This does not mean that artificial intelligence has become a threat to be feared. Quite the opposite: it could become one of the most important drivers of productivity growth over the coming decades by reducing costs, improving business efficiency, enhancing data analysis, accelerating decision-making and facilitating the development of new products and services.
However, economic history teaches us that every major productive force also creates new forms of risk. And that is where the real challenge begins.
When efficiency turns into dependence
If a single company uses an AI system and that system fails, the company may face an operational problem. But what happens when thousands of companies, banks and institutions depend on the same limited number of models, technology providers or data centres?
At that point, the problem shifts from an internal institutional issue to a potential disruption affecting a substantial part of the economy.
This is known as concentration risk.
Artificial intelligence may make the economy more efficient, but it could also make it increasingly dependent on a limited number of companies providing AI models, cloud computing services, semiconductors and data centres.
The greater this dependence becomes, the greater the possibility that a problem affecting one of these critical components could spread simultaneously across numerous institutions.
The risk, therefore, does not necessarily lie in artificial intelligence itself, but in the concentration of dependence on it.
One of the most serious questions AI raises for financial markets and banks concerns what happens when numerous institutions use similar models, interpret information in the same way and make similar decisions at the same time.
Each institution may believe its decision is entirely rational. However, if dozens of institutions sell the same financial asset simultaneously or restrict lending to the same economic sector at the same time, individually rational decisions could collectively trigger a crisis.
This represents a new form of herding behaviour.
Traditionally, herding occurred when investors followed one another’s decisions. In the age of artificial intelligence, it could happen automatically, as similar models reach similar conclusions at precisely the same time.
The greatest danger, therefore, may not be that a machine makes a mistake, but that thousands of machines make the same mistake, in the same direction and at the same moment.
That is where systemic risks begin to emerge.
Banks at the heart of the transformation
Banking is among the sectors with the greatest potential to benefit from artificial intelligence.
AI can be used to assess customers, detect fraud, combat money laundering, manage liquidity, analyse markets, measure risks and improve customer services. These applications could significantly enhance banking efficiency.
However, as AI becomes more deeply embedded in banking decisions, another question becomes increasingly important: Who reviews those decisions?
If an automated model rejects credit applications from thousands of customers, can a human reviewer examine the reasoning behind those decisions?
If a system downgrades the risk assessment of an entire economic sector, who determines whether that assessment reflects genuine risks or an error in the model?
And if several banks begin using similar models, could their simultaneous decisions trigger a contraction in credit across an entire sector?
This is why human oversight and the ability to override automated decisions are essential.
The principle is not that humans should replace technology, but that people must always retain the authority and ability to understand, review, challenge and, when necessary, stop automated decisions.
The labour market and the cost of transition
One of the most common claims about artificial intelligence is that it will eliminate jobs. However, the reality is more complicated.
Technology does not simply eliminate work; it also redistributes its economic value.
Some jobs may disappear or become less prevalent, while others will emerge. Skills that were once highly valued may become less relevant, while demand for new skills increases.
The central challenge, therefore, may not be unemployment alone, but skill mismatch: jobs may be available, yet workers may lack the skills required to perform them.
This creates another economic burden known as transition costs.
Education systems, vocational training programmes and labour markets do not always adapt as quickly as technology advances. The wider the gap between the pace of technological development and the speed at which people acquire new skills, the greater the economic and social cost of this transition.
The physical infrastructure behind the AI revolution
Another aspect that receives insufficient attention is that artificial intelligence requires enormous physical infrastructure.
It depends on data centres, electricity, semiconductors, communications networks, cooling systems and, in some cases, substantial quantities of water.
In other words, the digital revolution has a very tangible physical dimension.
As AI adoption expands, demand for energy, investment and data centre capacity will also increase.
This raises an important economic question: Could the rapid expansion of artificial intelligence create new bottlenecks in energy supplies, infrastructure or financing?
The world is witnessing enormous investments in AI, data centres, semiconductors and computing capacity.
These investments could mark the beginning of a major new cycle of productivity growth. However, they also raise a legitimate economic question: Will economic returns grow as quickly as investment spending?
If they do, the economy could experience one of the largest productivity gains of the modern era.
However, if investment continues to outpace returns over an extended period, the risk of capital misallocation could emerge.
This means that vast amounts of capital could flow into activities that fail to generate the expected returns within a reasonable timeframe.
That does not necessarily mean a bubble is forming. It does, however, mean that the scale and direction of investment have become important indicators to monitor.
Cybersecurity: A new source of systemic risk
Artificial intelligence can help institutions defend themselves against cyberattacks, but it can also make those attacks faster and more sophisticated.
This creates a new paradox: the same technology that helps a bank detect fraud can also enable attackers to develop more advanced tools.
As institutions become increasingly dependent on shared digital infrastructure, a single cyber incident could spread across a large number of organisations.
Under certain circumstances, cyber risk could therefore evolve into systemic risk, threatening not just individual institutions but the stability of interconnected economic and financial systems.
Governance and accountability: Who is responsible?
Perhaps the most important challenge in the next phase will be governance.
As artificial intelligence becomes increasingly capable of making decisions, questions of responsibility and accountability become more pressing.
Who authorises the use of an AI model? Who monitors its performance? Who establishes the boundaries it must not cross? Who bears responsibility if a decision results in financial losses? And who has the authority to shut down a system if it begins behaving unexpectedly?
These are not merely technical questions. They are fundamental questions of governance, accountability and institutional responsibility.
The future of artificial intelligence will therefore be determined not only by how intelligent its models become, but also by how effectively the institutions using them are managed and governed.
Artificial intelligence could become one of the greatest drivers of productivity in the modern economy.
However, as it becomes more widespread and increasingly embedded in decisions concerning credit, investment, production, markets and infrastructure, managing the risks it generates becomes ever more important.
The issue, then, is not whether to embrace or reject artificial intelligence. That choice may no longer be available.
The real question is how to harness its potential without allowing our growing dependence on it to become a new source of economic vulnerability.
Mohamed Abdel Aal, banking expert
