This article was originally published on AI News.
The debate on AI risks losing sight of its true scope. Because artificial intelligence is not limited to large generative systems, and the future of businesses cannot depend on a handful of technology players. The new issue of AI THAT WORKS
In recent months, some of the leading figures in artificial intelligence have made increasingly alarming statements about the risks associated with the evolution of this technology. From Sam Altman to the leadership of Anthropic, the public debate has been filled with predictions of potentially uncontrollable scenarios, even suggesting extremely serious consequences for humanity. Faced with these statements, the most immediate question should be: if artificial intelligence is truly so dangerous, why do the very companies warning about these risks keep developing ever more powerful models?
And there is an even more important question, one that risks remaining in the background: why do we keep identifying the entire artificial intelligence revolution with technologies developed by an extremely small number of companies?
The point is not to establish whether the alarms raised by Big Tech leaders are sincere, the product of a communication strategy, or tied to market-positioning needs. What matters is the effect of this narrative, which ends up reinforcing an equation as widespread as it is misleading: the one between artificial intelligence and large generative models.
Artificial intelligence is not just Big Tech’s
AI represents a technological transformation potentially comparable in scope to the Industrial Revolution. But to truly understand its implications, we must move past the idea that its future coincides exclusively with that of OpenAI, Anthropic, Google or the other large players that currently dominate the generative systems market. Artificial intelligence is a much broader universe, encompassing different technologies, approaches and applications. From machine learning to predictive systems, from industrial automation to robotics, there are many fields in which AI can deliver concrete results without necessarily relying on the most advanced generative models.
Think of energy consumption optimisation, predictive maintenance of plants, production process management, or the analysis of large volumes of industrial data. In these contexts, what matters is not having the most powerful model in absolute terms, but developing the most effective, reliable and controllable solution for a specific problem.
Continuing to portray artificial intelligence almost exclusively through the evolution of large generative models therefore means representing only part of the phenomenon, at the risk of overshadowing precisely those applications capable of generating immediate value for the productive system.
A revolution cannot be in the hands of a few
There is also a second, even deeper risk: technological concentration. If we accept that the entire economic and social transformation driven by AI must necessarily pass through the platforms of a limited number of companies, we are in fact imagining a revolution whose functioning depends on the decisions of a few private actors. Imagine if, during the Industrial Revolution, the first large car manufacturers had forced all other companies to use only their machinery, while also retaining the ability to control it or shut it down. That transformation would hardly have achieved the same level of diffusion, innovation and economic development. A technological revolution truly becomes one when its applications multiply, diversify and become the shared asset of a broad ecosystem of companies and operators. The same must happen for artificial intelligence: it is not just a matter of making existing systems accessible, but of fostering the development of different technologies capable of meeting the specific needs of individual sectors. Otherwise, the risk is that businesses become mere users of tools over which they have insufficient control, delegating an ever more important share of their activities to external parties.
The race for models does not match the needs of businesses
There is a further paradox in the current evolution of artificial intelligence: while the major technology players compete to develop ever more sophisticated systems, a significant part of the productive fabric has yet to fully harness the potential of the technologies already available. The speed of model innovation and the speed of their adoption in businesses move at profoundly different paces.
For many industrial applications, the technological capabilities available today are already sufficient to solve concrete problems. It is not necessarily the arrival of a new generation of models that brings a significant improvement to business processes. The real challenge lies elsewhere: bringing artificial intelligence into businesses, integrating it into existing systems in a secure, governable and efficient way, adapting it to production processes and making it a tool capable of delivering measurable results. The constant emphasis on ever more powerful systems, accompanied by statements about their potential catastrophic risks, instead risks widening the gap between those who develop these technologies and those who should use them. The effect could be paradoxical: just as artificial intelligence offers concrete opportunities for innovation, the way it is portrayed helps make it more distant, complex and difficult to understand.
Control and explainability are first and foremost a matter of effectiveness
For businesses, moreover, the question of control is fundamental. In many industrial contexts, from energy production to predictive maintenance, through to healthcare applications and critical processes, it is not enough to obtain a result from an artificial intelligence system. It is necessary to understand how it works, verify its decisions and intervene when needed. Explainability is often addressed solely as a regulatory requirement or a safety issue. In reality, it is also an essential condition for the effectiveness of applications. A company that does not adequately understand how a system works may struggle to manage it and optimise its performance. Control, therefore, is not only a tool for preventing risks, but a fundamental component of the ability to create value. This means that the future of industrial artificial intelligence need not coincide with ever more general-purpose and complex systems: in many cases, it will be precisely the specialised solutions, designed around the needs of businesses and their data, that offer the greatest development opportunities.
Europe’s challenge is not to lose sight of the real transformation
This reflection takes on particular significance for Europe. While the United States and China compete in developing large generative models, Europe has the opportunity to strengthen a path based on specialised applications of artificial intelligence and their integration into industrial processes. China, for its part, is pairing the race for general-purpose models with a strong interest in AI adoption across its own productive fabric. This deserves attention because it shows that the technological challenge is not limited to building the most advanced systems. For Europe, it would be a mistake to keep interpreting and regulating the entire phenomenon of artificial intelligence as if it coincided exclusively with the large generative platforms. Instead, we need to foster a diversified ecosystem in which businesses can develop, adopt and govern solutions suited to their needs, while retaining control over their data and strategic processes.
The future of AI does not belong only to those who build the most powerful models
The warnings of the leading players in artificial intelligence certainly deserve attention, as do the risks associated with the development of ever more advanced technologies. However, we cannot allow this narrative to end up identifying the entire AI transformation with the commercial fate of a few companies. The most concrete risk is losing sight of the extraordinary variety of possible applications and, above all, slowing down businesses’ ability to use a technology that can already help improve processes, efficiency and competitiveness today. Artificial intelligence must not be a revolution watched from afar, governed exclusively by a few large players. It must become a widespread transformation, in which businesses are protagonists and not mere users of systems developed elsewhere.





