An analyst’s missed remarks and U.S. intelligence systems that weren’t connected to one another are among the missteps that investigators have surfaced while probing the cause of a missile strike on an Iranian school that killed an estimated 120 children, people familiar with the matter said.
Years before the U.S. attacked Iran at the end of February, an intelligence analyst examining information about potential future strike targets in Iran noticed changes at a site the U.S. had previously characterized as a naval facility belonging to the elite wing of the Iranian military in Minab city in the southeast of the country. It was, in fact, now an elementary school.
The analyst remarked on changes at the site in a digital intelligence tool, but that tool wasn’t linked up to the official intelligence database that the U.S. uses to develop strike targets and the information was never conveyed to military commanders, according to people familiar with the matter who declined to be named discussing sensitive topics.
On Feb. 28, when President Trump announced the start of major combat operations against Iran, a missile struck the school. The attack killed an estimated 120 children, and nearly 200 people in all, representing the worst incident of civilian harm resulting from U.S. operations in decades.
The analyst’s remarks, which one of the people familiar with the matter said were submitted in 2019, were never heeded, and the same building was reviewed several more times over the following years without anyone updating the targeting database. These discoveries are among the issues explored in a Pentagon investigation into the school strike, the people said. The results of the probe have not been publicly released.
A Pentagon official said the incident remains under investigation and that the agency has no updates to provide. On Wednesday, Trump said it may not ever be possible to determine fault and that he doesn’t think the U.S. was to blame.
The details unearthed as part of the Pentagon investigation underscore long-standing weaknesses in the U.S. military’s targeting system, one that was supposed to be improved years ago. Upgrades have instead been beset by delays, and yet they’ve grown all the more urgent with the spread of AI. Some tout the technology as a possible solution to targeting woes while others worry it could scale and accelerate the harms of war.
The investigation into the school strike was submitted in April but remains under review at U.S. Central Command, the military theater and combatant command known as Centcom that is responsible for carrying out combat operations against Iran, according to one of the people familiar with the matter.
Centcom commander Brad Cooper, a four-star Navy admiral, ordered the investigation and appointed an Air Force general from outside the command with the intention of ensuring a thorough, independent review, the person said.
The analyst’s written remarks about the school, the fact that they were entered into a digital system in 2019 that wasn’t connected to the official intelligence database and the current status of the investigation into the strike have not been previously reported. The New York Times had previously reported that an analyst noticed the building appeared to be a school several years ago and informed one other person. Targeting officials were using imagery that hadn’t been updated in seven years, according to the Times.
There are significant and long-standing gaps in how the Pentagon analyzes potential strike targets, according to former senior intelligence officials and others familiar with the matter. They declined to be named to discuss sensitive matters.
At least two intelligence database systems used for inputting remarks based on imagery, for example, have historically not been connected to the official and authoritative targeting database, people familiar with the platforms said, creating a coordination challenge that continues today.
In some cases during the mid-2010s, targeting data for historically low-priority locations where the U.S. had little historical battle experience, such as Syria, proved to be 10 or 20 years old, according to one of the former senior intelligence officials. Some intelligence staff worked double shifts and weekends at that time to manually update the system.
Starting in 2017, the intelligence enterprise undertook a similar effort to update several thousands of outdated targets in North Korea after relations between Washington and Pyongyang rapidly deteriorated, people familiar with the matter said, calling in satellites and other efforts to capture new, clear imagery as well as other types of intelligence. It took more than a year to update critical targeting information.
A legacy database known as MIDB was created in the 1980s and often relies on manual input. The Pentagon plans to replace MIDB with a machine-assisted version known as MARS that will introduce more automation.
A recently revised Pentagon doctrine outlined the challenges of integrating the many systems used to identify military targets: “The process of targeting occurs on many levels and in many locations simultaneously, yet no single interoperable solution has emerged or been established,” according to the non-public targeting document revised in April and reviewed by Bloomberg. “The entire joint targeting enterprise should seamlessly share well-understood, standardized representations of target intelligence and data and not rely on local databases.”
The MIDB and MARS systems are now both in use, but the effort to shift entirely to MARS is years behind schedule, and authoritative targeting data still relies on MIDB, according to the targeting doctrine.
The U.S. Government Accountability Office in 2020, during Trump’s first term, described MIDB as having “long-standing deficiencies” and said it’s “unable to meet current needs.” And yet six years later, the Pentagon’s targeting doctrine still describes the system as the authoritative, all-source repository of worldwide general military and target intelligence, serving as the national database for all target lists and no-strike lists and a baseline source of intelligence on installations, facilities, military forces and population concentrations.
The characterizations of MIDB in the Pentagon’s latest targeting doctrine haven’t been previously reported.
The hope of some targeting experts is that linking digital systems and more AI will bring down targeting errors in future. An automated check against public sites such as Google Maps, for example, may help flag an anomaly for human review. The Pentagon introduced an agentic AI effort along these lines Thursday.
The Defense Intelligence Agency, an agency responsible for both MIDB and MARS didn’t directly address a request from Bloomberg for comment on MIDB’s deficiencies, delays in the MARS transition or the mislabeled school site. An agency spokesperson said its foundational military intelligence analysts conduct comprehensive analysis of infrastructure and the operational environment, drawing on all intelligence sources to produce expert intelligence analysis and produce and maintain foundational military intelligence.
Such sources can span not only satellite pictures and other imagery analysis, but also signals intelligence, human intelligence and more, the spokesperson said. Combatant commands rely on expert analytic support from these all-source analysts for operational planning and execution, including intelligence for targeting, the spokesperson said.
“DIA works in close coordination with combatant commands and Intelligence Community partners to ensure decisionmakers have the best available intelligence for our national security,” the spokesperson said in a written comment.
Under the latest U.S. targeting doctrine, military commanders are responsible for the decision to prioritize and strike a target. Along with planners, commanders are also required to distinguish between military objectives and civilian ones that are not lawful military objectives for lethal targeting.
A combatant command should establish guidance to mitigate civilian injuries and consider criteria for positive identification of a target, according to an updated section of the Pentagon’s targeting doctrine. A spokesperson for the Joint Staff, the Pentagon’s senior military staff, described that section as a “key update.”
Once a combatant command such as Centcom has assembled a target list, the joint-force commander may also initiate an additional “optional process” called target vetting to assess the accuracy of the intelligence behind the targeting, according to joint targeting doctrine reviewed by Bloomberg. As part of this process, officials would review any potential disagreements about the characterization of a target and any new imagery, the former senior intelligence officials familiar with the process said.
It would be “unthinkable” for a commander not to undertake this target vetting process for attacks planned on the opening day of a new military campaign, one of the former senior intelligence officials said. Centcom vetted targets leading up to the operations against Iran, according to the person familiar with the matter. It wasn’t clear, however, whether Centcom initiated the optional vetting process that would’ve required coordination across intelligence community agencies and a recheck of the underlying information and possibly any new imagery.
Centcom didn’t respond to Bloomberg’s request for comment on the target vetting. A spokesperson for the Joint Staff declined to comment, citing the ongoing investigation.
Jack Shanahan, a former Pentagon director for defense intelligence and retired three-star Air Force general, said there is no excuse for a combatant command to not review and validate the accuracy of information provided for every targeting package. Combatant commanders have the ultimate responsibility for validating the accuracy of targets, he said.
Shanahan described targeting in an interview as a “moribund career field” that had atrophied over two decades while the U.S. military focused on counterterrorism and counterinsurgency in the wake of the Sept. 11, 2001, attacks instead of traditional combat operations. In 2017, he said, he struggled to recruit and fill targeting roles. “We knew there was a dangerous shortage in the number of trained and experienced targeting personnel and weapons effects experts,” he said. “We also knew this would become a major problem in future conventional operations.”
In the days following the Iran school strike, Trump accused Iran of conducting the attack, though he has offered no evidence. Last week, Trump said “mistakes are made and war is nasty” when asked about the strike, committed to releasing the findings of the investigation and added that he’ll accept the results.
Defense Secretary Pete Hegseth said in mid-March that the investigation “will take as long as necessary to address all the matters surrounding the incident” and that his department would “share it when we have it, absolutely.”
Dozens of members of Congress have since demanded answers about what happened. The group Human Rights Activists in Iran said it’s documented the killings of more than 1,700 civilians in the first month of the war.
Emily Tripp, director of the nonprofit group Airwars, a watchdog that logs civilian harm in conflict zones, said that her group had tracked 300 incidents of civilian harm in Iran but that it was difficult to untangle whether the U.S. or Israel was responsible for them. Trump’s own claims on social media about the U.S. being behind some attacks has made it easier for Airwars to pursue accountability, she said.
Tripp said her group refers each incident to Centcom for review. The Defense Department is behind on “every single one of their commitments when it comes to civilian protection,” she said. The Pentagon did not respond to a request for comment on this specific allegation.
Bob Ashley, former director of the Defense Intelligence Agency during the first Trump administration, is among those calling on the Pentagon to publish the results of the investigation.
“Americans know that over 100 children were killed in this strike. We need to talk to them about what happened, because their trust and confidence in us, as the Department of Defense, and as an intelligence community, matters,” Ashley said in an interview.
In a military career spanning 36 years, Ashley helped train generals, was a former commander and senior intelligence officer at the Joint Special Operations Command and Central Command and currently sits on several advisory boards for companies focused on national security.
“We have an obligation to explain the targeting process, how we apply the criteria of the laws of armed conflict and review targets to be transparent to sustain that level of trust and understanding with the American people,” Ashley said.
He said the intelligence community needs to look at what happened, scrutinize their process and ask itself: “What can we do better? What did we miss?”
Manson writes for Bloomberg.