When U.S. forces bombed the Shajareh Tayyebeh elementary school in Minab, southern Iran, on February 28, 2026, at least 160 people, most of them children, died. The attack caused international outrage and brought the question of responsibility in modern warfare into sharp focus. This is because AI in war creates a new excuse: those responsible can hide behind systems to which they ultimately assign blame themselves.
Those in charge hide behind technology.
But instead of taking responsibility, the familiar struggle for the Primacy of interpretation. Donald Trump initially blamed Iran itself. An investigation of the incident followed, during which the available video material was apparently examined more closely. Footage that clearly shows the impact of a Patriot missile in slow motion.
When there was nothing left to relativize and the evidence became overwhelming, data errors, artificial intelligence, and systems like Palantir "Technologies" are brought onto the stage. That sounded modern, technical, almost neutral. And that's exactly the problem. Because while we're talking about AI in war, those responsible are increasingly hiding behind these systems – those who keep lists, those who build AI, those who approve attacks.
The error is not in the AI, but in the system
In Minab, it was never about an AI suddenly going out of control. It’s more about a military system that speeds up decisions, standardizes checkpoints, and distributes responsibility across databases, interfaces, and process chains. Palantir Maven Smart System, or MSS, is built precisely for this form of warfare: a platform that combines data from satellite imagery, sensors, and other intelligence sources to accelerate target identification and military decision-making.
Not one Language model Therefore, the school in Minab has also declared its objective. Rather, the research suggests that a longest-standing entry never properly corrected in a military database, although the building has been designated as School was recognizable. So the actual disaster didn't begin in an „intelligent“ system that erred. It began in a system optimized for speed, and in which a inaccurate data situation can become deadly. Those who speak of an AI error here are no longer talking about those who build and profit from such systems, about those who feed them data, and about those who ultimately authorize the attack.
No AI slip-up, but a predictable system error
The truly disturbing thing about the case isn't that a system knew too much. It's that it knew too little. corrected was. Because the building was in a database of Defense Intelligence Agency, also the US military intelligence service, which among other things provides targeting and reconnaissance data for such systems, is still listed as a military facility. Although it was separated from the neighboring Revolutionary Guard complex at the latest since 2016 and used as a school. The indications were publicly accessible and clear. Photos, Directory entries, Cards and visible markings On the school grounds, it was clear what this building had long been: a school. So there was no shortage of information. One just had to take it seriously.
This is precisely why the talk of an „AI fallacy“ is so misleading. Not because no software was involved. But because it was the Crucial concealed. The error apparently did not result from a sudden loss of control of the aircraft, but from an old, incorrect classification that could no longer be stopped in an accelerated military system. From an inaccurate data situation, no Question mark, but a Target.
You will find more infographics at Statista
And that's precisely where it gets political. Because the more faster the more complex such systems become, the smaller the window of opportunity for appeal becomes. Maven is designed to, up to 1,000 target decisions to process per hour. This is more than a technical metric. It is, rather, the tempo of murder. Where speed becomes the benchmark, control shrinks. Where processes are meant to run ever smoother, doubts quickly appear as obstacles. Incidents like the one in Minab are therefore the logical consequence of a system in which flawed data remains harmless for a long time, until it suddenly becomes deadly.
AI in war follows an old logic of killing
Because the idea of making wars more predictable through technology is much older as any current debate about artificial intelligence. Long before software targeted goals, armies were already trying to Unpredictable the war into procedures, standards, and technical processes. How dangerous this logic is is shown by looking back. In the Vietnam War, the US military deployed thousands of acoustic and seismic sensors, the data from which was relayed via aircraft to an evaluation center and there processed by IBM 360 computers were processed to detect movements on the Ho Chi Minh Trail and to control air strikes. The system could register signals, but could not reliably classify what it was actually detecting. Thus, technology did not create clarity, but often only the appearance of it.
In the Kosovo War, the same logic manifested itself in other ways: faulty target assignment, inadequate maps, and a bureaucratic apparatus that did not stop false assumptions in time, ultimately led to the bombing of Chinese Embassy in Belgrade by the U.S. Air Force. At the time, however, apologies were still made for any errors.
The common thread runs all the way to Minab. It's not the machine that replaces humans. Rather, institutions build systems in which incorrect data, standardized procedures, and time pressure reinforce each other. The myth of precision persists, even though reality has long been fragile. That's why talk of „AI failure“ falls short. It dramatizes the machine and relieves the human.
Responsibility doesn't disappear, it just becomes invisible.
The crucial question, therefore, is not whether a machine can be morally culpable. It is: Who bears responsibility when states and militaries deploy systems that select targets ever faster and make attacks appear ever more technical? In international humanitarian law, this question is answered more clearly than some current debates suggest. The principle of distinction obliges parties to a conflict to only military objectives to attack and civilian objects to protect. Responsibility therefore lies not with the AI, but with the acting actors, the military chains of command and the states that carry out such attacks.
the international debate about Autonomous weapon systems revolves precisely around this point. The International Committee of the Red Cross has warned for years that systems that select targets and apply force without sufficient human control raise fundamental legal, humanitarian, and ethical problems. What matters is not whether you label a technology with the tag AI. What matters is whether humans are still involved in a way that real control, Judgment and Accountability enables. Where this control wanes, it is not the neutrality of technology that grows, but the danger of organized irresponsibility.
Support journalism
Independent journalism needs strong supporters. If you liked this article, we welcome a direct contribution to our editorial team.
