For years, the debate around artificial intelligence was framed primarily as a matter of innovation. More speed, more computing power, more automation, greater efficiency. Yet there are moments when a seemingly isolated event reveals something far more profound. The presence of Christopher Olah, one of the world’s most influential researchers in the internal architecture of advanced language models, alongside Pope Leo XIV during the presentation of the encyclical Magnifica Humanitas, belongs to that category of signals.
This was not merely an encounter between Silicon Valley and the Vatican. Nor was it a symbolic exercise exploring the relationship between technology and spirituality. What emerges instead is a growing awareness that artificial intelligence is moving beyond the realm of technical innovation and entering the domain of the fundamental questions that define a civilization: power, knowledge, responsibility, work, and trust.
The Industrial Revolution transformed humanity’s relationship with physical force. Artificial intelligence is transforming its relationship with cognition itself. The distinction is profound. Nineteenth-century machines replaced muscle. Contemporary systems intervene in the very processes through which analysis, forecasts, decisions, and even interpretations of reality are produced. This is why the central issue is no longer how powerful these models have become. It is how understandable they remain.
Much of Olah’s work has revolved around a question that, until recently, seemed almost secondary: what actually happens inside a neural network when it generates a response? That question has suddenly become crucial because frontier systems are reaching levels of complexity that render even their decision-making architecture increasingly opaque. Today’s most advanced models are trained on trillions of words, require infrastructures costing tens of billions of dollars, and consume computational resources comparable to those of entire industrial sectors.
In previous technological eras, innovation produced tools that could be observed, dismantled, and rebuilt. Today, organizations are becoming dependent on systems that function more like emerging statistical ecosystems. This is not a semantic distinction. It is a structural transformation in the relationship between control and decision-making.
The consequences are already visible across the corporate world. According to several international analyses, generative artificial intelligence could generate productivity gains worth trillions of dollars in the coming years. Goldman Sachs has estimated that up to 300 million jobs could be affected by the automation of cognitive tasks, while McKinsey believes that a significant share of white-collar work may be fundamentally reshaped within the next decade. Behind these numbers lies more than an employment challenge. What is emerging is a redefinition of the geography of corporate power.
When a global consulting firm uses language models to accelerate strategic research and document production, or when a financial institution integrates generative systems into risk assessment and scenario planning, it is not simply adopting a new technology. It is altering the process through which organizational knowledge is created. Likewise, in the pharmaceutical sector, companies such as Moderna and other biotechnology leaders are deploying AI platforms to accelerate the discovery of promising molecules, compressing research timelines that until recently appeared impossible to shorten. Competitive advantage increasingly stems from the ability to collaborate with systems capable of generating cognitive value.
Yet every act of cognitive delegation introduces a question that balance sheets alone cannot answer: who truly controls the decision-making process?
If a strategic recommendation emerges from the interaction between executives, proprietary data, and opaque models, accountability becomes more difficult to define. The risk is not merely error. It is the gradual erosion of the boundary between human judgment and algorithmic suggestion. In other words, technology does not necessarily replace the decision-maker. It transforms the mental environment within which decisions are made.
This is where the debate opened by the Vatican acquires unexpected relevance for chief executives, investors, and boards of directors. When Pope Leo XIV emphasizes that machines possess neither consciousness, suffering, nor moral responsibility, the issue is not theological. It concerns the preservation of a fundamental principle: responsibility cannot be automated.
The discussion becomes even more significant when viewed through a geopolitical lens. Artificial intelligence has become a strategic infrastructure. The United States and China are investing hundreds of billions of dollars in data centers, advanced semiconductors, computational capacity, and proprietary platforms. Control over the cognitive supply chains of the future has become a national objective, much as control over oil, energy, or telecommunications shaped global power during the twentieth century.
The difference is that today’s critical resource is not merely physical. It is cognitive. Those who control advanced models increasingly influence the production of knowledge itself. Those who own the infrastructure define technological standards. Those who dominate access to data acquire competitive advantages that simultaneously affect economic performance, national security, scientific research, and cultural influence.
Europe observes this transformation from a unique position. It has established itself as a regulatory leader while remaining largely dependent on platforms and infrastructures developed elsewhere. This condition raises questions that are not merely economic but strategic. Digital sovereignty is no longer simply about protecting data. It is about participating in the construction of the cognitive architectures that will organize value creation in the decades ahead.
This is why the meeting between a leading AI interpretability researcher and the head of the Catholic Church carries a significance that extends far beyond the news cycle. Both are observing the same transformation from radically different perspectives. On one side stands the effort to understand what is emerging within artificial systems. On the other stands the effort to understand what is changing within the societies that adopt them.
The real challenge is not determining whether a machine can become human. It is understanding which conception of humanity will ultimately be embedded in the infrastructures that are reshaping work, authority, knowledge, and power. Great technological revolutions change tools. More rarely, they change the environment within which reality itself takes shape.
Artificial intelligence appears to belong to that second category. And when a technology ceases to be merely a tool and becomes the context within which a civilization organizes itself, the debate can no longer belong exclusively to engineers or technology companies. It becomes a question of economics, culture, governance, institutions, and the future itself. Ultimately, it concerns how societies choose to exercise power over what they have created before what they have created begins to redefine the very meaning of power.