Artificial intelligence has crossed a line that many people thought was still far from being reached.
On February 28, 2026, the United States carried out Operation Epic Fury against Iran, and within the first 24 hours, more than 1,000 targets were struck. It was not an army of analysts working around the clock for days that made this possible. It was an AI. To put that into perspective, the entire opening phase of the 2003 invasion of Iraq involved less airpower than this single day of operations, and now we are talking about something that happened in real time, guided by algorithms and language models running on servers.
The system behind this feat is the Maven Smart System, developed by Palantir, with the Claude language model from Anthropic integrated into the core of the process. It processes satellite imagery, drone feeds, radar data, and signals intelligence in real time, then delivers prioritized target lists complete with GPS coordinates, weapons recommendations, and even automated legal justifications for each strike. What previously required roughly 2,000 human analysts can now reportedly be operated by approximately 20 people.
Impressive? Absolutely. But that is also where the hardest questions begin, especially when a girls primary school shows up on one of those lists and more than 165 civilians die. Operation Epic Fury is not just a technological milestone. It is the first major real-world test of what happens when life-and-death decisions are accelerated by algorithms, and the world is still trying to figure out what to do about it. 🤔
What CENTCOM said publicly and what was left out
Admiral Brad Cooper, commander of CENTCOM, confirmed the role of artificial intelligence in a video statement released publicly on March 11. In his words, these systems help filter enormous amounts of data in seconds so leaders can make smarter and faster decisions than the enemy can react to. Cooper also made a point of emphasizing that humans always make the final decision about what to strike, what not to strike, and when to strike. But that advanced AI tools transform processes that once took hours and sometimes days into a matter of seconds.
So far, the message sounds reasonable. But what Cooper did not mention also matters, and it matters a lot. He did not identify any specific AI system by name. And the official statement did not address a data point circulating among analysts and experts: the reported accuracy rate of the Maven Smart System sits around 60 percent, while human analysts achieve roughly 84 percent in some comparative assessments. That is not a minor detail. In a scenario where more than a thousand targets are struck in 24 hours, a 40 percent error rate could mean hundreds of misidentified targets, and each one of those errors could cost the lives of innocent civilians.
That gap between what was said and what went unanswered fuels much of the distrust that human rights organizations and members of the U.S. Congress are now expressing. Partial transparency does not resolve the issue. In fact, it makes things even more complicated, because it gives the impression that something is being deliberately withheld.
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.
What the Maven Smart System is and how it works
The Maven Smart System is not exactly a new product, but the scale at which it was used during Operation Epic Fury is something that had never happened before. Project Maven started in 2017, when the U.S. Department of Defense launched a partnership with Google to use artificial intelligence for analyzing drone imagery. After significant internal backlash and pressure from Google employees, the company pulled out of the project. Palantir stepped in and took the system to a completely different level, integrating multiple data sources and adding natural language capability through Anthropic’s Claude model, which allows the system to explain its own decisions in human language, including legal arguments based on the laws of armed conflict.
In practice, what the system does is pull together information from multiple sources simultaneously: images captured by satellites, live video streamed from surveillance drones, radar data, communications intercepts, and behavioral pattern analysis. With all of that processed in seconds, the Maven Smart System generates a prioritized list of targets, already packaged with GPS coordinates, a recommendation for the most appropriate type of weapon for each situation, and an automated legal justification explaining why that target is considered legitimate under the rules of engagement. It is literally a system that thinks, prioritizes, and justifies all at the same time, autonomously, leaving human operators with only the final decision to pull the trigger — at least in theory.
That reduction from 2,000 analysts to roughly 20 operators is not just an impressive number from an operational standpoint. It represents a structural shift in how armed conflicts are planned and executed. The time between identifying a target and acting on it has dropped dramatically, which in military terminology is called decision-cycle compression. The shorter the cycle, the faster the response. But also the less time there is to review, question, or correct an error before it becomes an irreversible tragedy.
When the algorithm gets it wrong: the Minab school incident
One of the most disturbing episodes associated with Operation Epic Fury was the strike on the Shajareh Tayyebeh girls primary school in the city of Minab. The system identified the location as a legitimate target, generated the automated justifications, and the strike was carried out. According to Iranian reports, the result was the deaths of more than 165 civilians. Pentagon officials stated that outdated intelligence contributed to the strike and that a full investigation is underway. More than 120 Democratic members of the House of Representatives have formally demanded answers about the role of AI in this incident.
As war studies scholar Craig Jones noted in an interview with Democracy Now!, AI-driven target selection is reducing a massive human workload of tens of thousands of hours down to seconds and minutes, but at the same time automating human targeting decisions in ways that open up all kinds of problematic legal, ethical, and political questions.
This episode ignited an urgent debate about what it means to trust artificial intelligence to make, or at least recommend, decisions that carry irreversible consequences for human lives. The question is not whether the system worked technically. By all accounts, it performed exactly as it was designed to. The question is whether what it was designed to do is acceptable, and who is responsible when it goes wrong.
From a technical perspective, artificial intelligence systems like the Maven Smart System learn from historical data and patterns. If the training data contains biases, if the programmed rules of engagement leave gaps, or if the real-world environment introduces variables the system has never encountered before, the outcome can be catastrophic. In the case of the school, it is still not entirely clear what happened in the decision chain, but the mere fact that this was possible within a system equipped with automated legal justifications raises a serious question: when an AI produces a legal argument for a strike, who validates that argument? Does a human being still critically analyze it, or has trust in the system grown so deep that human review has become nothing more than a rubber stamp?
This kind of scenario is exactly what AI ethics researchers have been warning about for years. The automation of decisions in high-stakes contexts creates what some experts call the meaningful human control problem — that is, human control exists formally, but in practice the speed and volume of operations make it impossible for a human to review each decision with the care it requires. The operator presses the button, but the real decision was already made by the algorithm long before that. And when something goes wrong, accountability falls into a dangerous vacuum between the system developer, the military official who approved its use, and the operator who executed the command. 😕
The direct impact on commercial tech infrastructure
An important development in this story extends well beyond the battlefield. Iran has explicitly named Palantir, Google, Microsoft, Amazon, and other American tech companies as legitimate military targets, precisely because of the role their infrastructure plays in the war effort. Iranian strikes have already damaged AWS data centers in the United Arab Emirates and Bahrain, proving that this threat is not just rhetoric.
This has enormous implications for the entire cloud computing and digital services ecosystem in the Persian Gulf and beyond. If the commercial cloud infrastructure that supports services used by millions of people around the world is also the same infrastructure powering military AI systems, then attacks against that infrastructure trigger a cascading effect that goes far beyond any regional armed conflict. We are talking about risks to financial services, communication platforms, healthcare systems, and an entire chain of digital services that depend on those very same data centers.
Analysts have already started calling this conflict the first AI war, and one of the main reasons is precisely this unprecedented fusion of commercial technology and military capability. The lines between a civilian tech company and a defense contractor have blurred in a way that has never happened before at this scale. Every escalation in the conflict reverberates through financial markets within hours, and the AI dimension in target selection adds a new layer of systemic risk that the world has not yet learned to price or mitigate.
Ethics, accountability, and the future of AI-guided warfare
The discussion about ethics in the use of artificial intelligence in military contexts is not new, but Operation Epic Fury catapulted this conversation from a theoretical level to an urgent and concrete one. International organizations, humanitarian law researchers, and technology experts are debating more intensely than ever what it means to deploy autonomous systems in conflict zones. The central point is not about banning the technology — after all, technology itself has no moral intent. The point is to clearly establish who is accountable for the outcomes, how the systems are audited, and what the non-negotiable boundaries are that no algorithm can cross without real and effective human oversight.
Palantir maintains that the Maven Smart System increases precision and reduces collateral damage compared to traditional target selection methods, and that human control is still present at every critical stage. But critics point out that when you compress the decision cycle to the point where 20 people manage more than 1,000 strikes in 24 hours, real human control becomes an operational fiction. The math simply does not add up. You cannot deeply review 50 targets per hour per person, analyze the legal justifications, consider the geopolitical context, and still make an informed decision. Something in that chain will inevitably get waved through on autopilot, and at that point the human in the loop becomes nothing more than a stamp of approval for what the algorithm already decided.
What Operation Epic Fury leaves as its legacy goes far beyond the military outcome. It marks the beginning of an era in which armed conflicts will increasingly be defined by each side’s computational capacity, and in which data processing speed will outpace the human capacity for moral reflection in real time. This is not science fiction. It is what just happened, and the world needs answers that do not yet exist in any satisfactory form: how to regulate military AI systems on an international scale, how to ensure legal accountability when algorithms get it wrong, and how to preserve some form of human dignity on a battlefield where the most critical decisions are already being made by language models and neural networks. 🌐
Palantir’s role in the defense AI ecosystem
Palantir is not a defense company in the traditional sense. It positions itself as a software and data analytics company, but over the past several years it has built an increasingly deep presence within the governments and armed forces of U.S.-allied nations. Founded in 2003 by Peter Thiel, Alex Karp, and others, the company grew on government contracts and is now one of the leading providers of artificial intelligence technology to agencies like the CIA, the FBI, and the U.S. Department of Defense. The Maven Smart System is one of the most visible products of that positioning, but it is far from the only one.
What sets Palantir apart from other defense tech vendors is the depth of integration its systems enable. Instead of standalone tools that do one thing, the company builds platforms that connect multiple data sources and allow human operators — or algorithms — to spot patterns that would be impossible to identify manually. In a military context, that means integrating human intelligence, electronic signals, aerial imagery, and historical data into a single operational environment. Operation Epic Fury was the biggest live test of this approach, and the results — both the impressive ones and the deeply troubling ones — will shape how the company and its competitors develop the next generation of these tools.
There is also an important economic and geopolitical dimension to this story. The operational success of the Maven Smart System, regardless of the ethical controversies, will spark a race among nations to develop or acquire similar capabilities. That puts Palantir in a uniquely strategic position, but it also increases the pressure on the company to demonstrate that its systems have robust safeguards, that errors are investigated seriously, and that ethics is not just a nice-looking slide in an investor deck. Because when the product you sell can result in the deaths of civilians at a school, corporate responsibility stops being a branding issue and becomes a matter of humanity. 💡
What to expect going forward
The conflict between the United States and Iran in 2026 is already being treated by defense and technology analysts as a watershed moment. The way artificial intelligence was deployed in Operation Epic Fury demonstrates that commercial AI and warfare are no longer separate domains. The same language models, the same cloud infrastructures, and the same machine learning frameworks used for customer service, text generation, and corporate data analysis are now also feeding target selection systems in active combat zones.
This radically changes the conversation about AI regulation around the world. When a language model like Claude is capable of generating automated legal justifications for military strikes, the regulatory debate can no longer be limited to issues like misinformation, algorithmic bias, or copyright. The conversation needs to include, as a central and unavoidable element, the use of AI systems in lethal contexts, with clear rules around auditing, transparency, accountability, and above all, operational limits that protect civilian populations.
Operation Epic Fury will be studied for decades, not only as a military operation but as the moment humanity saw in real terms what happens when algorithms operate at the speed of war. What we do with the lessons from this episode will determine whether artificial intelligence is used to protect lives or whether it continues to be deployed in ways that make the distinction between combatant and civilian increasingly thin. That is perhaps the most important question technology has ever posed to us, and the answer remains wide open.
