How Artificial Intelligence is deciding targets in armed conflicts
Artificial Intelligence has moved well beyond being just the tool that recommends shows on your streaming app or answers questions on your phone. Today, in real armed conflict scenarios, sophisticated algorithms are being used to generate lists of military targets, calculate the probability that someone is a combatant, and accelerate decisions that historically depended on human judgment. What should have been a technological advancement capable of reducing errors has, in practice, become a mechanism that dilutes accountability when civilians are hit. When an automated system flags a target and the result is the death of innocent people, the question left hanging is direct and uncomfortable: who answers for that?
A detailed investigation published by The Guardian revealed how major tech companies, the ones most people know for everyday products and services, are effectively operating as defense contractors. They provide the computational infrastructure, the machine learning models, and the data platforms that power targeting systems in war zones. And the most unsettling part of this story is not that the system failed at some point. It is that it worked exactly as it was designed to work 😐
From fog at the guard post to fog inside the algorithm
The original article from The Guardian opens with a powerful reference to an Israeli military strategy known as the fog procedure. First used during the second intifada, this unofficial rule instructed soldiers at guard posts to fire bursts of shots into the darkness whenever visibility was low, under the justification that an invisible threat could be approaching. It was violence authorized by blindness. Shoot into the dark and call it deterrence.
With the arrival of Artificial Intelligence-driven warfare, this same logic of deliberate blindness was refined, systematized, and handed off to a machine. The darkness in the watchtower was a condition of the terrain. The darkness inside the algorithm is a condition of the design. In both cases, blindness was chosen because blindness is useful: it creates plausible deniability, makes violence seem inevitable, and shifts the question of who decided from a person to a procedure. The fog did not clear. It was given a probability score and rebranded as intelligence.
The recent war in Gaza was described as the first major AI war, the first conflict in which artificial intelligence systems played a central role in generating the list of suspected Hamas and Islamic Jihad militants to be targeted. Systems that processed billions of data points to classify the probability that any given person in the territory was a combatant. And what happened next did not stay confined to Gaza.
The case of the school in Minab and outdated intelligence
It may have been this same chosen blindness that led, at the start of the conflict between the United States and Israel against Iran, to the strike on the Shajareh Tayyebeh elementary school in the city of Minab, in southern Iran. At least 168 people were killed, most of them children, girls between seven and 12 years old.
The weapons were precise. Munitions experts described the targeting as incredibly accurate, with each building struck individually and nothing off-target. The problem was not the execution. The problem was the intelligence behind the decision. The school had been separated from an adjacent Revolutionary Guard base by a fence and converted to civilian use nearly a decade earlier. At some point in the targeting cycle, that fact apparently was never updated.
The exact role of AI in the Minab strike has not been officially confirmed. What is known is that the targeting infrastructure in which these systems operate does not have a reliable mechanism for flagging when the underlying intelligence is a decade out of date. To hit a thousand targets in the first 24 hours of the Iran campaign, the U.S. military relied on AI systems to generate, prioritize, and rank the target list at a speed no human team could replicate.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
As the original article put it: Gaza was the laboratory. Minab is the market. The result is a world in which the most consequential targeting decisions in modern warfare are made by systems that cannot explain themselves, supplied by companies that answer to no one, in conflicts that generate no accountability. This is not a system failure. This is the system.
The role of Big Tech as defense contractors
For a long time, the defense sector was dominated by traditional companies that manufactured weapons, armored vehicles, and military communication systems. That landscape has changed dramatically in recent years. Companies born in Silicon Valley that built their reputation around innovation and connectivity are now signing billion-dollar contracts with armed forces around the world. The Artificial Intelligence these companies develop is not just optimizing logistics or monitoring borders. It is integrated into systems that process surveillance data, cross-reference intelligence, and generate recommendations about who should be considered a legitimate target in armed conflicts. The transition of these corporations into the role of defense contractors happened relatively quietly, without the public scrutiny that normally accompanies decisions of this magnitude.
The original article names the companies directly. Palantir, founded with initial CIA funding and now one of the primary providers of AI infrastructure for the American military, supplied systems used in the Iran campaign. Those systems partly incorporate Claude, from Anthropic, a large language model whose parent company tried to resist Pentagon pressure to remove ethical restrictions on its use in targeting. The Pentagon’s response was to threaten to cut ties and turn to OpenAI and other providers. The market for killing at scale, as the Guardian put it, does not lack suppliers.
Google, despite significant internal protests from employees, signed Project Nimbus, a cloud computing and AI contract with the Israeli government and military valued at more than one billion dollars. Amazon is a co-signatory of Project Nimbus alongside Google. Microsoft had deep integration with Israeli military systems before partially pulling out under pressure in 2024, at which point the data migrated to Amazon Web Services within days.
Anduril, founded by Palmer Luckey and staffed largely by former U.S. defense officials, builds autonomous weapons systems explicitly designed for lethal targeting. OpenAI, which until recently prohibited military use in its terms of service, quietly removed that restriction in early 2024 and has since been pursuing Pentagon contracts. These are among the most valuable companies in the world, with consumer-facing products used by hundreds of millions of people, research partnerships with universities, and significant political influence in Washington, Brussels, and beyond.
What makes this situation particularly complex is the business model involved. Unlike a missile manufacturer, whose product has an obvious and regulated purpose, tech companies provide cloud computing services, pattern recognition algorithms, and data analytics platforms that can be applied to virtually any purpose. This creates an enormous gray area. The same infrastructure that powers a virtual assistant or a search tool can, with relatively small adjustments, be adapted to process military intelligence data and help define military targets. That versatility, which from a commercial standpoint is an advantage, is an ethical and legal nightmare.
When the algorithm gets it wrong, who takes the blame?
This is perhaps the hardest knot to untangle in this entire discussion. In traditional armed conflicts, there is a relatively clear chain of command. An officer authorizes a strike, an operator executes it, and when something goes wrong, there are mechanisms, however imperfect, for determining accountability. When Artificial Intelligence enters that equation, the chain of command becomes murky in a way that favors impunity.
The original article brings striking data on this. In Gaza, an algorithm processed data on every person in the territory, including phone records, movement patterns, social connections, and behavioral signals, and produced a ranked list of names, each with a probability score indicating the chance of being a combatant. Verification, in this system, meant that a human operator reviewed each name for an average of about 20 seconds, just enough time to confirm the target was male. Then signed off on the approval. A single system produced more than 37,000 targets in the first weeks of the war. Another was capable of generating 100 potential bombing locations per day. The humans in the loop were not exercising judgment. They were managing a queue.
This phenomenon has a name among technology ethics researchers: accountability gap. It is the empty space that forms when decisions with lethal consequences are distributed between humans and machines in such a way that no single party can be individually held responsible for the final outcome. Attribution dissolves along a chain of engineers, commanders, operators, and corporate suppliers, each of whom can point to the other. The reasoning disappears into a probability score that no lawyer can audit and no court can cross-examine. The process collapses into a 20-second rubber stamp of a machine recommendation.
The Guardian article also revisits an earlier case that illustrates how this logic did not start with AI but was amplified by it. In July 2014, four boys from the Bakr family, Ismail, Zakariya, Ahed, and Mohammad, aged between 9 and 11, were killed on a beach in Gaza. No AI was involved. The location had been previously classified as a Hamas naval compound. The boys were flagged as suspects because they ran and then walked, behavior that matched a targeting model for combatants trying not to draw attention. When the first missile struck, the surviving children fled. The drone followed them and fired again. An officer later testified that from a straight-down aerial view, it is very hard to identify children.
A classified Israeli military database, reviewed by the Guardian, +972 Magazine, and Local Call, indicated that of the more than 53,000 deaths recorded in Gaza, identified Hamas and Islamic Jihad combatants accounted for approximately 17%. That suggests the remaining 83% were civilians. The Israel Defense Forces disputed the figures, although they did not specify which ones.
There is also a documented cognitive bias called automation complacency, which describes the human tendency to place excessive trust in automated systems, especially when those systems are presented as technologically superior. In a war scenario, where every second counts, the temptation to accept the machine’s recommendation without questioning it is enormous. And when an entire military structure begins operating under this logic, the role of human judgment becomes increasingly ceremonial.
The lobby behind the curtain and the regulatory vacuum
The original article does not hold back on how the regulatory vacuum around these companies is deliberately maintained. Palantir spent nearly 6 million dollars on lobbying in Washington in 2024, and in one quarter of 2023 outspent Northrop Grumman, one of the largest traditional defense contractors in the world. The company launched a foundation dedicated to shaping the regulatory environment in which it operates.
The consortium formed by Palantir, Anduril, OpenAI, SpaceX, and Scale AI was described by its own participants as a project to provide a new generation of defense contractors to the U.S. government. The venture capital firms that fund these companies, such as Andreessen Horowitz and Founders Fund, have cultivated influence through proximity to power: former senior officials on their advisory boards, partners rotating through government positions, and direct access to the policymakers who determine how much the Pentagon spends and on what.
In the United States, the AI provisions in the 2025 National Defense Authorization Act do not regulate military AI. They direct agencies to adopt more AI. Pete Hegseth’s AI strategy, issued in January 2026, frames the issue entirely as a race, directing the Pentagon to move at the speed of war. The regulatory culture did not fail to keep up with the technology. It deliberately decided not to try.
What international law says about this and why it is not enough
International humanitarian law, built over decades from the Geneva Conventions, establishes principles such as distinction between combatants and civilians, proportionality in the use of force, and precaution in attacks. These norms were designed for a world in which combat decisions were made by human beings who could be judged for their actions. The introduction of Artificial Intelligence as a tool for selecting military targets does not explicitly violate any of these norms, and that is precisely why it is so problematic.
The European Union AI Act, the most ambitious attempt so far to govern artificial intelligence, explicitly exempts military and national security applications, with the stated justification that international humanitarian law is the more appropriate framework. It is a remarkable act of circularity: the only body of law being systematically hollowed out by these systems is designated as their regulator, while the regulators who could actually constrain them look the other way.
Organizations like the International Committee of the Red Cross and research groups like Human Rights Watch have pushed for specific regulations on the use of autonomous and semi-autonomous systems in armed conflicts. There are treaty proposals that would require meaningful, not merely nominal, human oversight in any decision involving the use of lethal force. But these negotiations move slowly. Many countries that invest heavily in military technology resist any regulation that might limit their strategic advantage.
The original article, however, points out that pressure points do exist and are real. The International Court of Justice advisory opinion on Palestinian rights created a framework under which companies supplying systems used in unlawful attacks face potential legal exposure in jurisdictions that take international law seriously. And AI companies need governments, not just as customers, but as providers of the computing power, energy, and physical infrastructure that frontier AI demands and that no company can sustain on commercial revenue alone. That dependence gives willing states real leverage over companies that would rather not be regulated.
What effective regulation could include
The original text argues that what regulation should include is relatively straightforward, even if it is hard to enforce. AI systems used in targeting need to be explainable, not through probability scores, but through reasoning that a lawyer can audit. The cumulative civilian cost of AI-assisted campaigns needs to be assessed as a whole, not strike by strike. And the accountability that stops at the operator needs to extend up the chain to the companies that knowingly built and sold opaque systems for use in armed conflicts.
These are not new demands. They are the minimum conditions for the laws of war to mean anything in the age of algorithmic targeting. The United Kingdom has already committed more than one billion pounds to a new AI-integrated targeting system, linking sensors and strike capabilities across all domains. France’s leading AI company has partnered with a German defense startup to build autonomous weapons platforms. Germany is deploying AI-guided attack drones in Ukraine. The train has already left the station.
The challenge is not only legal but also conceptual. The existing legal framework assumes it is possible to identify someone responsible for each action in a conflict. When that action is the result of a chain involving data collected by satellites, processed by machine learning algorithms, refined by language models, and presented to an analyst on an interface designed to facilitate quick decisions, the traditional notion of accountability simply does not hold up. This gap between technological capability and the regulatory framework is, at its core, the root of the problem. And as long as it persists, the trend is for more lethal decisions to be made by machines, with fewer people answering for the consequences.
The fog did not clear, it just got better hardware
The original article ends with an observation that hits like a punch to the gut. The fog procedure remains operational and is defining the future of warfare. But the soldiers who fired into the darkness were at least present in it. The companies that built what replaced them do it from Palo Alto, with no personal risk, no legal exposure, and every incentive to do it again.
The question hanging over all of this is not whether Artificial Intelligence will continue to be used in armed conflicts. That is already a reality. The real question is whether any government with the tools to act will decide, before the next Minab, that the cost of inaction has become too high. The darkness is still there. It just has better hardware now 🫠
