WSJ wsj.com By
Daniel Michaels
and
Dov Lieber
March 7, 2026 12:00 pm ET
The U.S. and Israeli attacks on Iran have unfolded at unprecedented speed and precision thanks to months of planning, a massive assemblage of military force and a cutting-edge weapon never before deployed on this scale: artificial intelligence.
AI tools are helping gather intelligence, pick targets, plan bombing missions and assess battle damage at speeds not previously possible. AI helps commanders manage supplies of everything from ammunition to spare parts and lets them choose the best weapon for each objective.
Before Israeli jet fighters launched ballistic missiles that killed Iran’s Supreme Leader Ali Khamenei at his residence a week ago, launching the current regional war, Israeli intelligence services had for years been monitoring hacked Tehran traffic cameras and eavesdropping on senior officials’ communications—increasingly relying on AI to sift through a flood of intercepts.
The use of AI in the campaign against Iran follows years of work by the Pentagon and lessons learned from other militaries. Ukraine—with U.S. help—increasingly relies on AI in its war against Russia. Israel has tapped AI in conflicts at least since the October 2023 Hamas attacks.
Defense Secretary Pete Hegseth has urged accelerated adoption of AI to create “an ‘AI-first’ warfighting force.” At the same time, he is engaging in a public battle with Anthropic, a critical AI supplier, and the Pentagon has contracted with rival OpenAI to use its models in classified settings. President Trump has ordered the government to stop using Anthropic’s products. But U.S. officials say the fight unfolding in Iran is showing the usefulness of Anthropic’s AI agent, Claude.
The U.S. and Israel have declined to discuss exactly how they are employing AI in the widening conflict, but recent comments from military leaders and technical experts provide a window.
Most military AI applications aim to give commanders and planners more complete information, faster than is now possible. That, in turn, should let them make better and quicker decisions than the enemy can, gaining a battlefield advantage.
The U.S. says it has struck more than 3,000 targets in Iran since the attacks began Saturday, using an array of weapons including attack drones launched from ships, F-22 jet fighters taking off from Israel and B-2 stealth bombers flying from the U.S.
While the complexity of managing so many aircraft and weapons is getting a boost from AI, its use remains limited and the cost of badly informed decisions remains high. U.S. military investigators believe American forces likely were responsible for a strike on the war’s first day that killed dozens of children at a girls elementary school in Iran, The Wall Street Journal reported.
Talk of military AI can conjure images of killer robots, but the reality is that its biggest uses now are often off the battlefield, in time-consuming and labor-intensive fields like intelligence, mission planning and logistics.
These noncombat areas are ripe for AI-inspired efficiency because out of every 10 people in the military, at most two face combat. Up to 90% of personnel are in support roles.
The Pentagon’s AI tools are similar to ChatGPT and other mass-market large language models, but limited to warfare and trained to tackle specific tasks using relevant information, seeking to avoid glitches and inaccuracies often besetting AI.
Still, war is among the most chaotic and complex human endeavors—posing unique problems for even the cutting edge of robotic thinking. The Pentagon’s first AI chief, retired Air Force Lt. Gen. Jack Shanahan, said building military AI is tough in part because much of the available data for training is out of date or unclear.
“The Department of Defense was built as a hardware company in the industrial age, and it has struggled to become a digital company in a software-centric era,” said Shanahan, who oversaw an AI-powered project in Iraq, dubbed Maven, almost a decade ago.
Military strikes start with intelligence. Gathering and parsing it can require thousands of analysts grinding for hours over communications intercepts, photographs and radar images as they try to divine the locations of missile launchers, tunnels and other targets.
Human analysts can examine at most 4% of the intelligence material that is typically collected, say U.S. officers who have worked in the field.
“The biggest immediate impact of AI is in intelligence,” said Israeli Col. Yishai Kohn, the defense ministry’s head of planning, economics and IT. “Many potential missions simply never happened because the manpower didn’t exist” to assess vital intelligence, said Kohn.
AI-powered machine vision can now quickly find vast numbers of targets—with the ability to single out specific models of aircraft or vehicles. It can listen for and summarize relevant conversations from intercepts.
“Intelligence agencies already have access to tons of video data, and current AI enables them to detect exactly what they need within an ocean of data,” said Matan Goldner, chief executive of Conntour, an Israeli company selling software to its and other countries’ security agencies that allows them to query video databases the same way LLMs are used to find patterns in texts.
Just as with mass-market AI, users can bore into results with queries, such as to identify every missile launcher located near a hospital. They can also set the system to alert when an event happens, such as “Tell me every time someone takes a photo near this military base.”
The U.S. Army’s 18th Airborne Corps, using software from data company Palantir Technologies in a continuing string of exercises dubbed Scarlet Dragon, matched its own record from Iraq as the military’s most efficient targeting operation ever, according to Emelia Probasco, a senior fellow at Georgetown University’s Center for Security and Emerging Technology. Thanks to AI, the corps achieved that with only 20 people, compared with more than 2,000 staffers employed in Iraq, she said.
Militaries in the North Atlantic Treaty Organization are using AI to track Russia’s shadow fleet of tankers, scanning millions of square miles several times a day for vessels that are illegally transferring fuel at sea, said French Adm. Pierre Vandier, NATO’s top officer for digital transformation. Imagery is then linked to ship identities for closer tracking and potential action, he said.
Vandier said AI is turning military intelligence analysis from a task of groping in darkness for targets to one of sifting through piles of them. “The number of targets you can nominate through AI is just skyrocketing,” Vandier said.
To prioritize targets and develop a course of action, the Pentagon is increasingly using AI to run models and digital wargames. In one of many efforts, last year it contracted with Pittsburgh-based Strategy Robot to develop advanced systems that can churn through vast numbers of scenarios despite imperfect information. From potentially millions of iterations, planners can zoom in on actions that are more likely to achieve their objectives.
In the pre-AI world, after rough outlines were agreed on for an operation, commanders and specialists would develop mission plans, compiling paper-stuffed binders in a weekslong exercise. AI can potentially do the same work in days, military leaders say.
Planning any military assault—from the fast, targeted mission in January to seize Venezuelan strongman Nicolás Maduro to the war with Iran—brings together subject-matter specialists including intelligence officers, combat commanders, weapons experts and logistics managers. Sessions can include around 40 people.
“The more people you add into planning, the longer it takes,” said a U.S. Army officer in Europe with experience in the process.
As preparations advance and plans evolve, each specialist revises their own plans, with knock-on effects for the others. If intelligence reports, for example, shift a bombing target to a more-distant objective, commanders may opt to use different aircraft or weapons, which in turn can affect crew rostering, flight planning and fuel consumption.
Until now, updating all those factors was slow and often subjective. Now AI can process complex interactions instantaneously, accounting for how each change ripples through military choreography.
Once a strike occurs, AI can speed assessments of battle damage, via image-processing software like tools helping with initial intelligence. While analysis is limited by the quality of imagery—which can depend on factors as basic as weather and whether a target is above ground—AI’s ability to merge varied inputs is changing the discipline. In a process known as sensor fusion, AI can digest visuals, radar, heat signatures and mass-spectroscopy to synthesize a list of possible conclusions. Fast analysis of where attacks succeeded or failed in turn helps refine lists of subsequent targets.
One thing AI can’t replace is human judgment. Many military officials involved in AI projects warn that the technology’s capabilities risk prompting an overreliance on information it provides—a trend linked with the phrase “The computer said to do this.”
Offloading decisions to AI “is a serious concern,” said Probasco at Georgetown, who held various posts in the Navy. She said that, as with other weapons systems, safeguards must be implemented to limit risks. “That infrastructure is underinvested in now,” she said.
