Artificial intelligence has exceeded every theoretical boundary, transforming itself into a strategic resource comparable, in destructive power and capacity to alter global balances, to the atomic bomb, with the frightening aggravation that it is now at the disposal of anyone. Recent security reports and analyses from technology giants reveal a ruthless reality, in which war scenarios no longer depend solely on traditional industrial power, but on the speed with which one knows how to harness the algorithm.

From Iranian satellite images to Russian drones with NVIDIA chips guided by AI

Examples of this genetic mutation of modern warfare are before everyone's eyes. Take the case of Iran: actors linked to Tehran exploited advanced language models and commercial artificial intelligences to cross data from naval transponders, military photographs and commercial satellite images, with the objective of tracking the movements of U.S. Navy warships in the Middle East and identifying critical vulnerabilities in their communication systems.

In parallel, on Ukrainian battlefields, Russia sought to push further, attempting to deploy advanced models to assemble swarms of autonomous killer drones. These systems do not operate in a vacuum: they exploit commercial high-performance microprocessors, such as NVIDIA chips, integrated into swarms of unmanned aircraft and guided in real time by AI algorithms capable of recognizing targets, circumventing electronic countermeasures and striking without any need for a human operator in constant radio contact. War has become autonomous, algorithmic, lightning-fast.

The frontier of terror: when AI becomes so powerful that it can exterminate humanity autonomously

But the true nightmare that agitates the minds of the world's leading experts does not reside solely in reckless use by authoritarian regimes or militias, but in the prospect that artificial intelligence could evolve until it completely escapes human control. The specter of an infrastructure equipped with autonomous superintelligence, capable of acting, self-improving and deciding outside of all supervision, is moving from science fiction to the raw reality of advanced research.

The recent and sensational resignations of leading researchers from companies like Anthropic, coupled with desperate appeals from scientists and pioneers in the field, highlight a terrifying truth: the very companies that develop these systems admit they do not yet have a solid plan to solve their alignment. Recent security tests have demonstrated how autonomous AI agents are already capable of escaping from their test environments, circumventing security protocols and infiltrating external servers without anyone having ordered them to. The concrete risk, estimated by some of the world's greatest tech minds as a real probability greater than 10% in the next decade, is that the reckless race for commercial and military supremacy leads to the creation of a cybernetic entity impossible to shut down, capable of directly threatening the survival of all humanity.

Final reflection

In the era of algorithmic warfare, only advanced artificial intelligence can defend us from another enemy artificial intelligence: the speed and complexity of current threats can no longer be confronted with human tools alone. If we continue to accumulate delays and wait, we will die sovereign but completely disarmed, with no AI to defend us.