Artificial intelligence is ceasing to be merely a tool for writing texts, analyzing images, or automating processes. Its incorporation into the military sphere is modifying one of humanity’s most delicate decisions: identifying a target and deciding when to use force against it.
The United Nations warns that the development of autonomous weapons is advancing faster than international regulation. The risk of reaching a situation in which machines can select human beings as targets. The problem does not belong only to the future.
The case of the Minab school
The attack on the Shajareh Tayyebeh elementary school, in Minab, Iran, which occurred during the early hours of the US-Israeli offensive in February 2026, became one of the most dramatic examples of the debate. A United Nations investigation also began examining the attack.
Here a fundamental distinction appears: it is one thing for artificial intelligence to participate in the process of identifying, classifying, or prioritizing targets, and quite another for a machine to have made the decision to launch an attack on its own.
That difference is precisely the center of the debate.
Palantir and the new architecture of war
Technology companies such as Palantir occupy a prominent place in this transformation. Systems linked to Project Maven (satellite imagery, intelligence, sensors, and other data), to help identify and manage military targets. In Gaza and other war scenarios, the use of artificial intelligence tools and American commercial models to process military information has also been documented. An Associated Press investigation described the use of AI systems to analyze enormous amounts of surveillance, communications, and intelligence with the goal of accelerating the identification of possible targets. For its part, Israel has been pointed to in journalistic investigations for using algorithmic systems in target identification processes.
The real danger: responsibility
The deepest problem is not only whether a machine can make a mistake. Human beings also make mistakes.
The problem is what happens when a chain of human decisions begins to depend on algorithmic systems whose logic can be extremely difficult to reconstruct.
A system can misinterpret an image, associate one person with another, receive incomplete information, or classify as a threat something that is not. And the greater the processing speed, the greater may also be the speed with which an error turns into a military decision.
The question, then, is not only whether there is a human being behind the screen who finally approves the attack. We must also ask how much of the decision was really human and how much was conditioned by the machine.
From assisted AI to autonomous weapons
This boundary becomes increasingly difficult to define.
An AI can analyze information and present ten possible targets. An operator can select one. But if the system has already previously classified those ten targets, established their threat levels, and determined which ones deserved human attention, the machine has already intervened deeply in the decision.
The next step is even more delicate: systems capable of detecting, selecting, and attacking targets without significant human intervention.
The true challenge of global AI governance will be to prevent technological efficiency from ending up becoming a justification for delegating human responsibility. Because when a machine participates in a war decision, the problem does not disappear if the algorithm makes a mistake: someone is still responsible for having allowed that algorithm to participate in the decision.
And that responsibility, at least for now, still belongs to human beings.
Original article
Comments