The Price of Algorithmic War: How AI Became the New Dynamite in the Middle East

The Iran conflict has turned frontier AI models into contested weapons of state — and the financial and human fallout is only beginning to register.

In the first eleven days of the U.S.-Israeli offensive against Iran, which began on February 28, 2026, American and Israeli forces executed roughly 5,500 strikes on Iranian targets. That is an operational tempo that would have required months in any previous conflict — made possible, in significant part, by artificial intelligence. In the first eleven days of the conflict, America achieved an astonishing 5,500 strikes, using AI on a large-scale battlefield for the first time at this scale. The National The same week those bombs fell, a legal and commercial crisis erupted in Silicon Valley with consequences that will define the AI industry for years. Both events are part of the same story.

We are living through the moment when AI ceased being a future-war thought experiment and became an operational reality — embedded in targeting pipelines, shaping intelligence assessments, and now at the center of a constitutional showdown between a frontier AI company and the United States government. Alfred Nobel, who invented dynamite and then spent the remainder of his life in tortured ambivalence about it, would have recognized the pattern immediately.

The Kill Chain, Accelerated

The joint U.S. and Israeli offensive on Iran revealed how algorithm-based targeting and data-driven intelligence are reforming the mechanics of warfare. In the first twelve hours alone, U.S. and Israeli forces reportedly carried out nearly 900 strikes on Iranian targets — an operational tempo that would have taken days or even weeks in earlier conflicts. Interesting Engineering

At the technological center of this acceleration sits a system most Americans have never heard of: Project Maven. Anthropic’s Claude has become a crucial component of Palantir’s Maven intelligence analysis program, which was also used in the U.S. operation to capture Venezuelan President Nicolás Maduro. Claude is used to help military analysts sort through intelligence and does not directly provide targeting advice, according to a person with knowledge of Anthropic’s work with the Defense Department. NBC News This is a distinction with genuine moral weight — between decision-support and decision-making — but one that is becoming harder to sustain at the speed at which modern targeting now operates.

Critics warn that this trend could compress decision timelines to levels where human judgment is marginalized, ushering in an era of warfare conducted at what has been described as “faster than the speed of thought.” This shortening interval raises fears that human experts may end up merely approving recommendations generated by algorithms. In an environment dictated by speed and automation, the space for hesitation, dissent, or moral restraint may be shrinking just as quickly. Interesting Engineering

The U.S. military’s posture has been notably sanguine about these concerns. Admiral Brad Cooper, head of U.S. Central Command, confirmed that AI is helping soldiers process troves of data, stressing that humans make final targeting decisions — but critics note the gap between that principle and verifiable practice remains wide. Al Jazeera

The Financial Architecture of AI Warfare

The economic dimensions of this transformation are substantial and largely unreported in their full complexity. Understanding them requires holding three separate financial narratives simultaneously.

The direct contract market is the most visible layer. Over the past year, the U.S. Department of Defense signed agreements worth up to $200 million each with several major AI companies, including Anthropic, OpenAI, and Google. CNBC These are not trivial sums in isolation, but they represent the seed capital of a much larger transformation. The military AI market is projected to reach $28.67 billion by 2030, as the speed of military decision-making begins to surpass human cognitive capacity. Emirates 24|7

The collateral economic disruption is less discussed but potentially far larger. On March 1, Iranian drone strikes took out three Amazon Web Services facilities in the Middle East — two in the UAE and one in Bahrain — in what appear to be the first publicly confirmed military attacks on a hyperscale cloud provider. The strikes devastated cloud availability across the region, affecting banks, online payment platforms, and ride-hailing services, with some effects felt by AWS users worldwide. The Motley Fool The IRGC cited the data centers’ support for U.S. military and intelligence networks as justification. This represents a strategic escalation that no risk-management framework in the technology sector adequately anticipated: cloud infrastructure as a legitimate military target.

The reputational and legal costs of AI’s battlefield role may ultimately dwarf both. Anthropic’s court filings stated that the Pentagon’s supply-chain designation could cut the company’s 2026 revenue by several billion dollars and harm its reputation with enterprise clients. A single partner with a multi-million-dollar contract has already switched from Claude to a competing system, eliminating a potential revenue pipeline worth more than $100 million. Negotiations with financial institutions worth approximately $180 million combined have also been disrupted. Itp

The Anthropic-Pentagon Fracture: A Defining Test

The dispute between Anthropic and the U.S. Department of Defense is not merely a contract negotiation gone wrong. It is the first high-profile case in which a frontier AI company drew a public ethical line — and then watched the government attempt to destroy it for doing so.

The sequence of events is now well-documented. The administration’s decisions capped an acrimonious dispute over whether Anthropic could prohibit its tools from being used in mass surveillance of American citizens or to power autonomous weapon systems, as part of a military contract worth up to $200 million. Anthropic said it had tried in good faith to reach an agreement, making clear it supported all lawful uses of AI for national security aside from two narrow exceptions. NPR

When Anthropic held its position, the response was unprecedented in the annals of U.S. technology policy. Defense Secretary Pete Hegseth declared Anthropic a supply chain risk in a statement so broad that it can only be seen as a power play aimed at destroying the company. Shortly thereafter, OpenAI announced it had reached its own deal with the Pentagon, claiming it had secured all the safety terms that Anthropic sought, plus additional guardrails. Council on Foreign Relations

In an extraordinary move, the Pentagon designated Anthropic a supply chain risk — a label historically only applied to foreign adversaries. The designation would require defense vendors and contractors to certify that they don’t use the company’s models in their work with the Pentagon. CNBC That this was applied to a U.S.-headquartered company, founded by former employees of a U.S. nonprofit, and valued at $380 billion, represents a remarkable inversion of the logic the designation was designed to serve.

Meanwhile, Washington was attacking an American frontier AI leader while Chinese labs were on a tear. In the past month alone, five major Chinese models dropped: Alibaba’s Qwen 3.5, Zhipu AI’s GLM-5, MiniMax’s M2.5, ByteDance’s Doubao 2.0, and Moonshot’s Kimi K2.5. Council on Foreign Relations The geopolitical irony is not subtle: in punishing a safety-focused American AI company, the administration may have handed Beijing its most useful competitive gift of the year.

The Human Cost: Social Ramifications No Algorithm Can Compute

Against the financial ledger, the humanitarian accounting is staggering and still incomplete.

The Iranian Red Crescent Society reported that the U.S.-Israeli bombardment campaign damaged nearly 20,000 civilian buildings and 77 healthcare facilities. Strikes also hit oil depots, several street markets, sports venues, schools, and a water desalination plant, according to Iranian officials. Al Jazeera

The case that has attracted the most scrutiny is the bombing of the Shajareh Tayyebeh elementary school in Minab, southern Iran. A strike on the school in the early hours of February 28 killed more than 170 people, most of them children. More than 120 Democratic members of Congress wrote to Defense Secretary Hegseth demanding answers, citing preliminary findings that outdated intelligence may have been to blame for selecting the target. NBC News

The potential connection to AI decision-support systems is explored with forensic precision by experts at the Bulletin of the Atomic Scientists. One analysis notes that the mistargeting could have stemmed from an AI system with access to old intelligence — satellite data that predated the conversion of an IRGC compound into an active school — and that such temporal reasoning failures are a known weakness of large language models. Even with humans nominally “in the loop,” people frequently defer to algorithmic outputs without careful independent examination. Bulletin of the Atomic Scientists

The social fallout extends well beyond individual atrocities. Israel’s Lavender AI-powered database, used to analyze surveillance data and identify potential targets in Gaza, was wrong at least 10 percent of the time, resulting in thousands of civilian casualties. A recent study found that AI models from OpenAI, Anthropic, and Google opted to use nuclear weapons in simulated war games in 95 percent of cases. Rest of World The simulation result does not predict real-world behavior, but it reveals how strategic reasoning models can default toward extreme outcomes under pressure — a finding that ought to unsettle anyone who imagines that algorithmic warfare is inherently more precise than the human kind.

The corrosion of accountability is perhaps the most insidious long-term social effect. “There is no evidence that AI lowers civilian deaths or wrongful targeting decisions — and it may be that the opposite is true,” says Craig Jones, a political geographer at Newcastle University who researches military targeting. Nature Yet the speed and opacity of AI-assisted operations makes it exponentially harder to assign responsibility when things go wrong. Algorithms do not face courts-martial.

Governance: The International Gap

Rapid technological development is outpacing slow international discussions. Academics and legal experts meeting in Geneva in March 2026 to discuss lethal autonomous weapons systems found themselves studying a technology already being used at scale in active conflicts. Nature The gap between the pace of deployment and the pace of governance has never been wider.

The Middle East and North Africa are arguably the most conflict-ridden and militarized regions in the world, with four out of eleven “extreme conflicts” identified in 2024 by the Armed Conflict Location and Event Data organization occurring there. The region has become a testing ground for AI warfare whose lessons — and whose errors — will shape every future conflict. War on the Rocks

The legal framework governing AI in warfare remains, generously described, aspirational. The U.S. military’s stated commitment to keeping “humans in the loop” is a principle that has no internationally binding enforcement mechanism, no agreed definition of what meaningful human control actually entails, and no independent auditing process. One expert observed that the biggest danger with AI is when humans treat it as an all-purpose solution rather than something that can speed up specific processes — and that this habit of over-reliance is particularly lethal in a military context. The National

AI as the New Dynamite: Nobel’s Unresolved Legacy

When Alfred Nobel invented dynamite in 1867, he believed — genuinely — that a weapon so devastatingly efficient would make war unthinkably costly and therefore rare. He was catastrophically wrong. The Franco-Prussian War, the First World War, and the entire industrial-era atrocity that followed proved that more powerful weapons do not deter wars; they escalate them, and they increase civilian mortality relative to combatant casualties.

The parallel to AI is not decorative. The argument for AI in warfare — that algorithmic precision reduces collateral damage, that faster targeting shortens conflicts, that autonomous systems absorb military risk that would otherwise fall on human soldiers — is structurally identical to Nobel’s argument for dynamite. It is the rationalization of a dual-use technology by those with an interest in its proliferation.

Drone technology in the Middle East has already shifted from manual control toward full autonomy, with “kamikaze” drones utilizing computer vision to strike targets independently if communications are severed. As AI becomes more integrated into militaries, the advancements will become even more pronounced with “unpredictable, risky, and lethal consequences,” according to Steve Feldstein, a senior fellow at the Carnegie Endowment for International Peace. Rest of World

The Anthropic dispute, whatever its ultimate legal resolution, has surfaced a question that Silicon Valley has been able to defer until now: can a technology company that builds frontier AI models — systems capable of synthesizing intelligence, generating targeting assessments, and running strategic simulations — genuinely control how those systems are used once deployed by a state? As OpenAI’s own FAQ acknowledged when asked what would happen if the government violated its contract terms: “As with any contract, we could terminate it.” The entire edifice of AI safety in warfare, for now, rests on the contractual leverage of companies that have already agreed to participate. Council on Foreign Relations

Nobel at least had the decency to endow prizes. The AI industry is still working out what it owes.

Policy Recommendations

A minimally adequate governance framework for AI in warfare would need to accomplish several things. Independent verification of “human in the loop” claims — not merely the assertion of it — is the essential starting point. Mandatory after-action reporting on AI involvement in any strike that results in civilian casualties would create accountability where none currently exists. International agreement on a baseline error-rate threshold — above which AI targeting systems may not be used without additional human review — would translate abstract humanitarian law into operational reality.

The technology companies themselves bear responsibility that no contract clause can fully discharge. Researchers from OpenAI, Google DeepMind, and other labs submitted a court filing supporting Anthropic’s position, arguing that restrictions on domestic surveillance and autonomous weapons are reasonable until stronger legal safeguards are established. ColombiaOne That the most capable AI builders in the world believe their own technology is not yet reliable enough for autonomous lethal use is information that should be at the center of every policy debate — not buried in court filings.