Should we really be afraid of artificial intelligence?
My answer is no.
I believe AI is one of the greatest technological opportunities of our time. I don't think we should fear it, avoid it, or try to stop its development.
And yet I also believe there are very good reasons to be afraid of it. We should fear it when we don't understand it, when we passively accept whatever it offers us and, above all, when we stop being the ones who govern it.
Today I see enthusiastic people who hand it any task without questioning the results, and others who reject it because they are convinced it will end up replacing them. Two opposite attitudes with the same root: a poor understanding of the technology we are talking about.
Blindly trusting a tool we don't understand is not very different from having an irrational fear of it.
To understand what is happening, though, it's worth taking a short trip back in time.
Four industrial revolutions, and each time the fear of being replaced
Every major technological revolution has stirred up fears, transformed work, and introduced risks that didn't exist before. And each time the response was not to give up on the technology, but to build rules.
The first industrial revolution, which began in Great Britain in the second half of the 18th century, brought mechanization and steam power. A mechanical loom did the work of many people: some trades went into crisis, and many weavers lost their income and their jobs. But factories, mechanical engineering, and transportation grew, and those who were able to acquire new skills found professions that had not existed before.
Enormous problems emerged as well: exploitation, child labor, terrible conditions. The response was the Factory Act of 1833, which limited child labor and introduced inspections, and the one of 1844, which required protections for dangerous machinery.
The second industrial revolution, from the second half of the 19th century, was the revolution of electricity, oil, steel, and mass production. The automobile caused trades tied to carriages and animal traction to decline, while mechanical engineering, the automotive industry, and fuel distribution grew. Electrification changed work and everyday life.
Industrial accidents and new occupational exposures arrived too. In 1884 Germany introduced compulsory accident insurance; in Italy, Law 80 of 1898 made it mandatory for workers in certain industries.
The third industrial revolution, from the second half of the 20th century, is that of electronics, computing, and automation. Typing, switchboards, filing, and manual bookkeeping have shrunk or disappeared. Programmers, system administrators, security specialists, and entire digital sectors have emerged.
With the Internet came cybercrime, privacy violations, and misuse of data. Europe responded with the 1995 data protection directive and then with the GDPR. In the United States, the Occupational Safety and Health Act of 1970 had already strengthened workplace health protection.
The fourth industrial revolution is the one we are living through: advanced robotics, the Internet of Things and, above all, artificial intelligence. This time the transformation also affects cognitive activities: analyzing documents, writing code, creating content, supporting complex decisions.
The mechanism is familiar, though: some activities get automated, others are transformed, and new skills gain value. And this time too, we are building rules.
The European Union has adopted the AI Act (Regulation 2024/1689), later amended by Regulation 2026/1744. It distinguishes uses based on risk, introduces transparency obligations, requirements for high-risk systems, and bans on certain practices. The rules on AI literacy have applied since February 2025; for several high-risk systems the deadlines are December 2027 and August 2028.
The path is not very different from those we have already traveled. The real difference is that today we know the mistakes of the past and can try not to repeat them.
If machines replace us, what will become of work?
This is probably the most widespread fear, and the one that deserves the most careful attention.
It would not be accurate to claim that nobody lost their job in the industrial revolutions. It happened, sometimes dramatically. But it would be just as wrong to tell the story of innovation only through the trades that disappeared, forgetting those that were born.
A study published in 2024 in the Quarterly Journal of Economics estimates that about 60% of US employment in 2018 was in job specializations that emerged after 1940. Many jobs we consider normal today did not exist a few generations ago.
There is, however, a fact that optimistic narratives tend to leave out: the "long run" can last a lifetime. During the first British industrial revolution, between the late 18th century and the mid-19th century, productivity grew while real wages remained almost flat for decades. The economic historian Robert Allen called it "Engels' pause". An entire generation of workers paid for the transition without seeing its benefits.
And this time there are at least two differences to take seriously. The first is speed: previous revolutions spread over decades, while generative AI has reached hundreds of millions of people in just a few years. The second is that it affects cognitive work, that is, precisely the professions in which many took refuge to escape automation.
So we have no certainty that as many jobs will be created as are eliminated, or that everyone will manage to adapt. But we also have no grounds to claim that AI will inevitably cause permanent mass unemployment.
According to a 2025 study by the International Labour Organization, about one in four workers worldwide holds a job that is exposed to generative AI to some degree. For most of these jobs the most plausible scenario is a transformation of tasks, not the elimination of the position.
In the long run, after all, productivity has increased enormously and living conditions have improved. According to Our World in Data, in several advanced countries annual hours worked have roughly halved compared to the late 19th century. Not only thanks to technology: workers' rights, education, social policies, and medicine matter too. But it's hard to imagine the same results without technological development.
The point is not to defend jobs at all costs just as they are today. It is to put people in a position to take on tomorrow's jobs, and to shorten their "Engels' pause" as much as possible.
The most concrete risk, in my view, is not that an AI system performs some of our tasks. It is that we keep working, training, and organizing professions as if this transformation were not happening.
Will AI make us dumber and less creative?
Another frequent criticism is that AI will make us less capable of reasoning and less creative. I think that is, at the very least, a superficial claim.
Remember school research projects done by copying pages by hand from the encyclopedia at home? Then Wikipedia arrived, and the pen turned into copy and paste. Today we ask a chatbot directly to write the report.
But has the problem really changed? Whoever copied from the encyclopedia without understanding didn't learn much. Whoever copied from Wikipedia without digging deeper didn't develop critical thinking. Whoever copies an AI-generated answer today is doing exactly the same thing.
With one important difference: a generative system can invent information and present it convincingly. That's why the ability to verify becomes even more important.
It is not a tool that makes us smart or dumb. It is the way we decide to use it.
Creativity
Anyone who asks AI for a text, an image, or a project and settles for the first result contributes to a mass production of increasingly similar content: the same expressions, the same styles, ideas that look like variations of the same model.
The risk is real. A study published in Science Advances in 2024 showed that AI assistance improves individuals' stories, especially for those starting with less developed creative abilities. But the stories written with AI also tended to resemble each other more. AI can therefore improve the individual result and, at the same time, foster collective standardization.
Now let's look at the same technology from the perspective of a person with a vision of their own and ideas to develop. AI lets them explore alternatives, experiment with different languages, test an intuition, and build prototypes that would previously have required far more time and resources. This doesn't apply only to those who are already very creative: it can also help less experienced people express ideas that used to remain stuck.
But the direction remains ours. Creative value doesn't lie in knowing how to ask a machine to produce something. It lies in having something to express and knowing how to recognize what is worthwhile.
Critical thinking
If I delegate all reasoning to the machine without questioning its answers, I will exercise some of my abilities less and less. A 2025 study by Microsoft and Carnegie Mellon observed that greater trust in AI is associated with less self-reported exercise of critical thinking. It doesn't prove that AI makes us less intelligent, but it flags a concrete risk.
If instead I use AI to dig deeper into what I don't know, compare lines of reasoning, and look for objections to my ideas, I am not giving up thinking. I am expanding the tools available to my thinking.
AI can be a shortcut for those who want to stop reasoning or a multiplier for those who want to keep doing it. The risk is not that it thinks along with us. It is that we get used to letting it think in our place.
Can AI be dangerous for humanity? Certainly. But it's not the first technology that is
I don't consider it irrational to see AI as a possible threat. A technology capable of influencing behavior, automating decisions, and interacting with critical infrastructure can cause very serious harm.
But is it really the first with such potential? How many great discoveries have served to improve people's lives and, at the same time, deliberately destructive purposes?
Chemistry: feeding billions of people or producing weapons. In the early 20th century Fritz Haber developed the synthesis of ammonia from atmospheric nitrogen, and Carl Bosch brought it to industrial scale. The Haber-Bosch process revolutionized fertilizers: according to a famous analysis by Vaclav Smil in Nature, without synthetic nitrogen almost half of the world's population would not have enough food. But the same technology allowed Germany to produce explosives during World War I. And Haber led the development of German chemical weapons, taking part in the use of chlorine at Ypres in 1915.
Electricity: restarting a heart or stopping it. Today it powers defibrillators, ventilators, dialysis, and intensive monitoring: it often literally makes the difference between life and death. But as early as 1890, in the United States, the first execution by electric chair was carried out.
Nuclear power: treating cancer or destroying a city. Radiation physics made radiotherapy and nuclear medicine possible: according to the WHO, over half of cancer patients need radiotherapy. The same fundamental research led to the bombs of Hiroshima and Nagasaki, which by the end of 1945 caused between roughly 110,000 and over 200,000 deaths, according to estimates. Nuclear medicine and the atomic bomb are not the same technology, but they derive from the same knowledge, developed in opposite directions.
Computing: improving healthcare or blackmailing hospitals. Computer systems support diagnosis, medication, monitoring, and the exchange of clinical data. In May 2017 the WannaCry ransomware hit organizations across many sectors, including the British health service. According to the National Audit Office, over 19,000 appointments and operations were canceled, and some emergency rooms had to divert patients. You didn't need to destroy a hospital to compromise its operation.
The same goes for steam, which connected continents but also powered the British ship Nemesis in the First Opium War, and for the combustion engine, which moves ambulances as well as tanks.
The lesson is that the existence of a technology does not guarantee its responsible use. That's why we have gradually built rules and institutions: international bans on chemical and biological weapons, nuclear non-proliferation treaties, plant safety regulations, data protection requirements. We have often intervened late, after the damage was done. But we haven't given up on progress: we have tried to distinguish what brings benefits from what must be controlled or banned.
But is AI really just a tool?
Here I have to deal with the most serious objection. All the previous examples concern neutral tools: there is always a human being who decides the purpose. Those who consider AI a risk of a different nature argue precisely that it is not entirely a neutral tool.
There are three arguments. Agentic systems can act with a certain degree of autonomy. They can pursue a goal in ways their designers did not anticipate: this is the so-called alignment problem. And their inner workings remain largely opaque even to those who build them.
I believe this objection should be taken seriously. But it doesn't lead to conclusions different from mine: if anything, it makes knowledge, verification, and human oversight even more urgent. A system that can fail in unforeseen ways requires more expertise from those who use it, not less, and stricter rules for those who develop it and put it into production.
Who governs those who govern it?
There is also a difference that concerns power. The most advanced models, the computing power, and the data needed to train them are concentrated in the hands of a few players. This did not happen, to the same extent, with electricity or computing.
Governing AI therefore also means asking who decides how it is developed, with what priorities, and with what controls. It is not merely a technical question: it is a democratic one.
The fear we should really have
An AI can give a wrong answer with apparently absolute confidence. It can invent references, misunderstand information, or propose solutions that are plausible but inappropriate.
If we know at least its principles, we can understand its limits, verify its answers, and decide when to trust it. If we know nothing about it, we risk mistaking a convincing answer for a correct one.
The problem grows now that we are moving from systems that generate content to agentic systems, capable of using tools and taking actions on applications and services. Receiving a wrong suggestion and being able to correct it is one thing. Letting a system carry out operations without adequate controls is another.
I'm not saying everyone should become a machine learning expert. I'm saying that those who use AI should know its capabilities, limits, and risks in relation to what they entrust to it. And those who develop it or introduce it into professional contexts have even greater responsibilities.
In practice, for those who use it every day, this means at least four habits:
- Verify sources, numbers, and quotes before reusing them, especially when the answer seems perfect.
- Don't share personal, confidential, or client data without knowing where it ends up and who can read it.
- Don't delegate judgment on decisions with major consequences for people: AI can prepare, but the decision remains human.
- Use it to challenge yourself, asking for objections and weak points in your own ideas, not just confirmation.
Competence isn't only for making better use of a technology. It's for not becoming dependent on it without noticing. And it applies to a professional as much as to a business or a public administration.
But it's not enough for the individual citizen to learn how to use AI. The responsibility also lies with those who design it, distribute it, and decide where to deploy it. We need training, oversight, transparency, and enforceable rules.
It's no coincidence that the AI Act devotes an article to AI literacy. It's a significant choice: knowledge is not just an individual opportunity, but a component of technology governance.
The future is not to be feared. It is to be governed
We don't know which professions will exist in twenty years, nor how far AI will be able to automate activities we now consider exclusively human. Nor can we rule out risks more serious than those we are able to foresee today.
But we do know that great innovations have created problems, sometimes enormous ones, and at the same time progress that we now consider indispensable. And that we have learned, often too slowly, to build rules, skills, and protections.
That is what we must keep doing with artificial intelligence, trying this time not to always arrive afterward.
I don't believe the way forward is to stop progress because we fear its consequences. I believe it is to understand it, study it, discuss it critically, and take responsibility for steering it. This doesn't make AI automatically safe and doesn't guarantee everyone a better working future. But it offers far more concrete tools than fear does.
We shouldn't be afraid of artificial intelligence because it is powerful. We should be afraid when we don't know what we are doing with that power.
When we give up on understanding, verifying, choosing. When we use it to replace our creativity instead of expanding it. When we delegate not just a task, but our judgment as well.
The real danger is not necessarily that artificial intelligence becomes smarter than us.
It is that we stop using our own intelligence to govern it.
Sources and further reading
Industrial history: UK Parliament – Factory Act of 1833, Factory Act of 1844, Robert C. Allen – Engels' pause (Explorations in Economic History, 2009), Royal Museums Greenwich – The ship Nemesis in 1841, Library of Congress – History of the electric chair.
Chemistry, medicine, and computing: Science History Institute – Fritz Haber, Nature – Haber-Bosch and food production, WHO – Radiotherapy and cancer treatment, National Audit Office – WannaCry and the British health service.
Employment, creativity, and critical thinking: MIT – New Frontiers: The Origins and Content of New Work, ILO – Generative AI and Jobs, 2025, Our World in Data – Working hours, Science Advances – Creativity and AI, Microsoft Research – AI and critical thinking.
Legislation: AI Act, EU Regulation 2024/1689, amending EU Regulation 2026/1744, European Commission – Implementation timeline.