The idea that agentic Artificial Intelligence—an autonomous system capable of planning, making decisions, and executing complex tasks with minimal human supervision—could lead to human extinction has become one of the most debated topics of our time.
The evolution toward a superintelligent system whose goals are not perfectly aligned with humanity's preservation and well-being could, at least in theory, lead Artificial Intelligence to make catastrophic decisions, treating humanity as an obstacle or a mere secondary resource.
Thus, when the leaders of the largest companies in the sector in the United States, such as Anthropic or OpenAI, warn publicly that agentic Artificial Intelligence technology is evolving faster than the human capacity to supervise it, alarm bells should sound and be clearly understood everywhere.
Reason, and above all common sense—which Descartes defined as the most universally shared quality—would demand global alignment to effectively regulate the sector and prevent eventual disastrous consequences identified by some as existential to the human species itself.
However, the geopolitical reality is far more complex. In a multipolar and fragmented world with various military conflicts and fierce competition, these warnings are not only downplayed, but risk falling on deaf ears.
Let us be pragmatic in our analysis: without a coordinated strategy among the three major global blocs—the United States and China leading the race, and the European Union, which is more advanced in regulation—humanity seems poised to fall easily into the Game Theory trap known as the Prisoner's Dilemma.
Effectively, any country or company that decides to slow down the development of agentic Artificial Intelligence out of ethical or safety cautions risks being irreversibly overtaken by its rivals. This could result in the worst collective outcome, considering the warnings from Dario Amodei and Sam Altman mentioned above.
The fear of losing the technological race and the potential advantages that come with it prevents unilateral prudence out of dread of the opponent. Therefore, without an effective agreement with real enforcement and mutual commitments, any self-restraint will be interpreted not as prudence, but as a geopolitical surrender to be exploited without hesitation by rivals.
In short, the Artificial Intelligence race and the accelerated development of agentic models—increasingly operating without human supervision—are turning into a dangerous zero-sum game, where the urgency not to fall behind overrides the necessary prudence.
Overcoming this impasse will require world leaders to understand that the true risk is not losing technological leadership to a rival, but rather forfeiting control over a technology so advanced and complex that it may not forgive a lack of global cooperation.
















