Iran Today

The Minab Test: A School Strike Forensic

by Iran Prism 33 min read

The Sound

A staff member at Shajareh Tayyebeh girls’ elementary school in Minab was inside when it happened. She heard what she described as a “horrifying sound.” When she made her way back, she found bodies “lying on classroom benches and in different corners of the school.”1

That was the only firsthand account from inside the building to reach the outside world in the first hours. Everything else came through Iranian state media during a near-total internet blackout.

On February 28, 2026, Iran’s Foreign Minister Abbas Araghchi posted photographs on X. The caption read: “The destroyed building is a primary school for girls in the south of Iran. It was bombed in broad daylight, when packed with young pupils. Dozens of innocent children have been murdered at this site alone.”2

Washington Post journalists verified video showing debris, a partially collapsed wall, and smoke. NBC News independently verified additional footage.3

Within hours, the reported death toll climbed through eight revisions, from 5 to 85.

The Minab school claim became the defining image of Operation Epic Fury. It was cited at the UN Security Council. It dominated international media coverage. And it arrived through the same information environment that has produced four decades of manufactured, inflated, and weaponized atrocity narratives.

Sixty Days Later

The headline shifts can be stated up front.

The strike was a U.S. Navy Tomahawk cruise missile. That attribution is no longer in dispute. It rests on geolocated video, munition fragment analysis with a New York Times-traced contract code, the Pentagon’s own confirmation that Tomahawks were the opening salvo, and the operational division of labor that placed Minab in the U.S. area of responsibility.4

The school was a school. By the U.S. military’s own preliminary admission, the targeting package was built on outdated Defense Intelligence Agency data that still classified the school building as part of the IRGC compound it had been physically walled off from a decade earlier.5

The casualty count, however, did not converge. The regime stabilized at approximately 168. Hengaw, after three weeks of named-victim identification work, confirmed only 58. That gap is now the central remaining factual question, and Hengaw published it explicitly.6

Sixty days later, three things are simultaneously true: a real strike; a real institutional vetting failure on the U.S. side; a real continuation of the regime’s documented pattern of inflating contested casualty events.

What the Regime Claimed

Iranian state media reported that on February 28, 2026, a strike destroyed Shajareh Tayyebeh girls’ elementary school in Minab, Hormozgan Province, during the morning school shift. Saturday is Iran’s first workday.7

The casualty figures escalated through eight revisions in a single day:

SourceReported Dead
Vice Governor Ahmad Nafisi (initial)5
Governor Mohammad Ashouri via IRNA24
IRNA (state news)40
Press TV43
Governor Mohammad Ashouri (updated)51
Ministry of Education spokesperson53
Governor Mohammad Ashouri (third update)64
Iran’s Judiciary via Tasnim (Day 1 final)85

In the days that followed, the Mizan News Agency (judiciary) settled on 110 schoolchildren plus 26 teachers and four parents. President Masoud Pezeshkian publicly cited “more than 170 people killed.” Across regime sources, the figure stabilized in a range of 156 to 175. By approximately fifty days post-strike, IFP News reported the canonical regime figure as 168.8

A separate regime claim of a third strike on the school (the “triple tap” framing) came from the Iranian Ministry of Education. Two Iranian Red Crescent medics and a parent told Middle East Eye there were two strikes (the “double tap” framing).9

What the Coalition Acknowledged

CENTCOM confirmed Operation Epic Fury began at 1:15 AM ET on February 28. Targets included IRGC command and control, air defense capabilities, missile and drone launch sites, and military airfields across multiple Iranian provinces. The Pentagon’s published 10-day fact sheet documented over 5,000 targets struck, with approximately 1,000 strikes in the first 24 hours and 1,250 in the first 48.10

CENTCOM did not name Minab in its initial briefings. The U.S. position then evolved over thirteen days, in three distinct phases.

Phase 1: Initial Denial

On March 1, Captain Tim Hawkins, CENTCOM spokesperson, told CBS News the military “had no statement to share regarding the alleged strike on the school.” The Pentagon told CNN it “had nothing to share at this time.”11

Phase 2: The Tomahawk Concession

On March 2, JCS Chairman Gen. Dan Caine told the Pentagon press corps: “The first weapons launched at sea were Tomahawks fired by the United States Navy.” Two days later, Caine displayed a 100-hour strike-locations map. Minab was on it.12

The same day, Defense Secretary Pete Hegseth said only: “All I can say is that we’re investigating that. We of course never target civilian targets.”13

Phase 3: Trump Turbulence and the AR 15-6 Disclosure

On March 8, with Hegseth standing behind him, President Trump told reporters: “Based on what I’ve seen, that was done by Iran. They are very inaccurate, as you know, with their munitions. They have no accuracy whatsoever. It was done by Iran.” The next day Trump amended this to: “Iran or somebody else. A Tomahawk is very generic. Because I just don’t know enough about it.”14

On March 10, Sen. John Kennedy of Louisiana broke from the Republican non-comment line, telling reporters the U.S. “made a mistake” and the strike “was not intentional.” “I’m sorry,” he added.15

On March 11, the New York Times published the preliminary findings of the U.S. Army’s AR 15-6 administrative inquiry. CENTCOM, the Times reported, “created the target coordinates for the strike using outdated information provided by the Defense Intelligence Agency.”16

On March 13, Hegseth formally announced the AR 15-6 investigation, conducted by a U.S. general officer from outside CENTCOM.17

The Israeli Statement

The Israel Defense Forces issued a separate, brief statement to Human Rights Watch and reporters on March 6: “Not aware of any IDF activity in the area.” Tomahawk cruise missiles are operated by the U.S. Navy, the United Kingdom, Australia, Japan, and the Netherlands. Of those operators, only the U.S. is operationally engaged in the conflict.18

What Independent Investigators Established

Within seven to ten days of the strike, the major open-source investigative organizations converged on a consistent picture.

CNN’s geolocation team established the school’s coordinates and proximity to the IRGC base. BBC Verify journalists Shayan Sardarizadeh, Chris Osieck, and Newsha Tabrizy independently confirmed the coordinates as 27.109787 N, 57.084671 E. The school sits approximately sixty meters (about 200 feet) from the IRGC Naval Forces compound, identified in different reporting as the Sayyid al-Shuhada complex, the Asef Missile Brigade base, or the IRGC Navy coastal defense facility for the Strait of Hormuz approach.19

The school taught girls and boys on separate floors. Ages seven to twelve.20

The Munition

The weapon was identified as a BGM-109 Tomahawk Land Attack Missile, a U.S. Navy precision-guided cruise missile launched from surface ships and submarines.

The identification rests on four convergent strands.

A geolocated video released by Mehr News on March 9 showed a missile striking the compound, filmed from a construction site approximately 250 meters away. The cruciform shape with centrally mounted wings and a tailkit at the back matches the Tomahawk profile. Sam Lair of the James Martin Center for Nonproliferation Studies, John Gilbert of the Center for Arms Control and Non-Proliferation, and N.R. Jenzen-Jones of Armament Research Services independently confirmed the identification.21

Munition fragments photographed by IRIB on March 9 carried a contract code that the New York Times traced to Tomahawk procurement.22

Bellingcat’s March 27 analysis of two videos released by Iran’s Foreign Ministry confirmed the Tomahawk identification independently.23

The damage signature was diagnostic. HRW noted “distinctive pancaking of the roof” consistent with top-down precision impact. Jenzen-Jones characterized the multi-impact pattern across eight to ten distinct compound structures as “consistent with highly accurate, guided weapons,” explicitly ruling out an Iranian air-defense interceptor misfire.24

The Strike Sequence

Bellingcat’s March 27 analysis identified two waves of strikes approximately one hour apart. At least ten missiles impacted the compound over a fifty-second window in the first wave. HRW counted at least eight visible impact sites on March 4 Planet Labs commercial satellite imagery, with ten likely direct hits.25

Three different framings of the strike sequence circulate in coverage. Iran’s Ministry of Education described a “triple tap.” Two Iranian Red Crescent medics and a parent told Middle East Eye it was a “double tap.” Bellingcat’s “two-wave” finding is technically distinct from the textbook double-tap pattern, which targets first responders within minutes. U.S. officials have not endorsed any of the three framings.26

The School Separation Timeline

This is the part of the picture that did the most work in shifting the AR 15-6 finding from “we do not know” to a documented institutional failure.

Satellite imagery analyzed by HRW, CNN, NYT, AP, Amnesty Evidence Lab, and TIME shows the following progression:

In 2013, the school building was part of a single unified compound with the IRGC base.

By September 2016, a wall had been erected separating the school from the base. Three new street entrances were established, none of them with military checkpoints. The compound’s four military security posts had been removed.

By August 2017, an outdoor play area was visible from above.

Between 2022 and 2023, the adjacent Shahid Absalan medical clinic was separately walled off. Its public-facing entrance opened with then-IRGC Commander-in-Chief Hossein Salami cutting the ribbon, captured in Iranian media and reported by NPR.

In December 2025, two months before the strike, commercial satellite imagery showed dozens of people in the school’s courtyard, “apparently playing in what appears to be a court for ballgames” (CNN’s phrasing).

According to Minab’s mayor (cited by HRW), the original IRGC base “had been closed for about 15 years” and “all military personnel had moved out; the school was the only operational facility.” Reuters described the school as having a “vivid website and yearslong online presence.” The Guardian’s reconstruction concluded that “there was no indication that the building served a military purpose.”27

The Vetting Failure

The school was a school. The U.S. military had imagery and public-domain evidence available to confirm it was a school. The strike package was built on data that did not reflect any of this.

That conclusion, in an institutional voice, is the headline of the preliminary AR 15-6 administrative inquiry, as reported by the New York Times on March 11, citing officials briefed on the investigation. CENTCOM “created the target coordinates for the strike using outdated information provided by the Defense Intelligence Agency.”28

The DIA legacy data still classified the school building as part of the IRGC compound it had been walled off from at least a decade earlier. The school’s online presence was not reviewed. Neither were the commercial satellite images that, by HRW’s account, made the building’s civilian function visible to anyone who looked.

The Just Security legal analysis by Joseph N. Orenstein locates the failure in Phase 2 of the U.S. joint targeting cycle (target development), governed by Joint Publication 3-60. Appendix A paragraph 4(b)(7) of that doctrine explicitly requires that target data be current and that targeting officers review all-source intelligence rather than accept inherited target coding from a database. Orenstein’s conclusion: the strike “rises well above the threshold of an unremarkable mistake.”29

Sarah Yager of Human Rights Watch put the same point in a different register on March 12: “The findings of the US military investigation into the Minab school attack show a violation of the laws of war that cannot be boiled down to a blameless mistake.”30

The U.S. military’s no-strike list is the procedural mechanism designed to prevent exactly this kind of error. HRW posed the question directly: “If the school was on a ‘no strike’ list, why was the list not referenced prior to targeting Minab?”31

The AR 15-6 investigation has been “elevated” but not concluded as of May 1, 2026. The investigating officer is from outside CENTCOM. What the inquiry has not yet resolved publicly: whether AI-assisted geospatial tools propagated the legacy DIA classification without flagging the conversion, whether the strike-package timeline allowed the Collateral Damage Estimation review (CDE Level 5, mandatory when civilian casualties are anticipated), what level of authority approved the strike, and whether the failure was deliberate, reckless, or an honest mistake.

The March 11 letter from forty-six senators including Minority Leader Chuck Schumer, Sens. Chris Van Hollen, Tim Kaine, Brian Schatz, and Elizabeth Warren pressed Hegseth on those exact questions. The senators demanded answers by March 18; as of his April 29 House testimony — seven weeks past that deadline — the answers remained pending.32

Hengaw and the Verification Gap

Hengaw is a Norway-based Kurdish human rights organization. Its methodology for the Minab strike is the same one it applies to every contested casualty event. Identify each victim by name and identifying information. Work through accessible documentary sources. Refuse to publish a number that has not been built up from named individuals.

After three weeks of identification work, Hengaw confirmed 58 victims of the Minab strike: 48 children and 10 adults.

That number is far below the 156 to 175 range that Iranian state media stabilized at. The mechanism is what The Grief Factory documents across multiple incidents: when the regime controls the information environment, regime-sourced numbers reach global media before independent verification can catch up. Hengaw published the gap explicitly:

“Iranian state media and official authorities previously reported that 167 children were killed in an incident at a school in Minab. However, after 21 days, only 58 victims from this incident have been identified.”33

Three things follow.

First, the verified floor is 58. The ceiling is unresolved. The honest accounting is a tier-1 (independently verified) figure of 58 named victims, a tier-2 (regime-claimed, contested) figure of approximately 168, and a tier-3 (genuinely unknown) gap between them. The final death toll cannot be published with confidence by anyone working from public sources.

Second, even at 58 verified, the Minab strike is the largest civilian casualty incident attributable to U.S. forces in approximately thirty-five years. The previous high-water mark was the Amiriyah shelter strike in Baghdad on February 13, 1991: 408 civilians killed in a single incident. Minab does not approach Amiriyah at the regime’s claimed figure, much less the verified one. But it is the worst U.S.-attributable civilian incident since.34

Third, the gap itself is the pattern. Across the Islamic Republic’s contested casualty events going back to the 1988 prison massacres, every initial regime account has been contradicted by subsequent independent verification. (See The Credibility Ledger.) The ratio between Hengaw’s 58 and the regime’s 168 (roughly threefold) is consistent with that record. The gap does not establish that no children died at the school. It establishes that the regime’s number cannot be taken as the verified one.

Hengaw’s broader campaign accounting (covering the full 40 days through the April 8 ceasefire) put total Iranian casualties at 7,650, of whom 1,030 were civilians. That works out to a 13.5 percent civilian-casualty rate aggregate. The aggregate includes deaths attributable both to coalition strikes and to regime actions during the campaign (internal crackdown deaths, Iranian air-defense interceptor misfires, secondary explosions at IRGC facilities). Hengaw does not disaggregate causation in the headline figure.

A second method converges on approximately the same number. Applying the documented regime inflation pattern (the Minab gap of 168 claimed to 58 verified is a roughly 2.9-fold ratio) as a discount to the Iranian Ministry of Health’s broader war-civilian claims (which span 2,362 to 3,375) implies a true civilian-death range of about 800 to 1,500. Hengaw’s independently derived 1,030 sits inside that range. Two methods working from different starting points arrive at the same approximate floor. The historical context for the ratio is set out in “The Comparative Frame” below.35

The Machine in the Loop

The Maven Smart System was operationally present at the speed and scale that produced Minab. That much is established.

The system is a Palantir-built consolidation of “eight or nine” prior intelligence and targeting systems into a single visualization layer. The Pentagon’s contract scaled from $480 million in 2024 to $1.3 billion by 2025. CENTCOM commander Adm. Brad Cooper publicly credited “AI tools” with helping the U.S. military during Operation Epic Fury.36

At AIPCON, Palantir’s user conference, in mid-March, Chief Digital and AI Officer Cameron Stanley described the system as having consolidated those prior systems and displayed a Maven map of Iran with red icons including the Minab area. Public capability claims attribute to the system the production of “1,000 targeting recommendations per hour.” The campaign tempo at which Minab was struck (approximately 1,000 strikes in the first 24 hours, 5,500 to 6,000 in three weeks) was, by the U.S. military’s own public account, only achievable because of Maven.37

Anthropic’s Claude was operating in the Maven analytic stack until February 27, 2026, the day before the strike, when Defense Secretary Hegseth designated Anthropic a “supply chain risk.” That designation was the first time the formal label had been applied to an American company. (Anthropic sued; a federal preliminary injunction blocked the designation as “classic First Amendment retaliation”; the Pentagon-Anthropic dispute is ongoing.) During Operation Epic Fury, Claude was reportedly continuing to operate during a six-month phase-out.38

A 2024 internal Pentagon test placed the Maven system’s object-recognition accuracy at approximately 60 percent.39

The AR 15-6 inquiry, as of May 1, has not formally concluded that AI-assisted targeting tools caused the Minab error. The New York Times’ preliminary reporting indicated investigators found it “unlikely” that AI tools were the primary cause.40 “Primary cause” is a careful formulation. It leaves the secondary-causation question open.

The most productive framing of the AI question comes from Just Security’s Joseph Orenstein:

“If the Maven Smart System or a similar tool incorporated legacy DIA target codes without a verification layer, the AI did not create the error, but it may have laundered it into the strike package with a false aura of analytical confidence.”41

The Senate’s March 11 letter to Hegseth asked the question directly: “Were AI tools used in planning or executing these strikes?” The answer, as of May 1, remained unprovided.

Two open lines remain in the AI dimension of the investigation. The technical question of whether Maven or a related tool propagated the legacy DIA classification has not been answered. The doctrinal question of whether human review at the speed of “1,000 targets in 24 hours” can satisfy JP 3-60 Phase 2 requirements has not been adjudicated.

The Office That Wasn’t There

The Civilian Harm Mitigation and Response apparatus was created by Congress in the FY2023 National Defense Authorization Act. The trigger was the August 2021 Kabul drone strike on Zemari Ahmadi and his family, an incident with the same institutional pattern that produced Minab: outdated intelligence, pattern-of-life misidentification, post-strike defensive narrative, accountability that did not extend beyond administrative measures.42

CHMR was the Pentagon office designed to prevent the next Minab.

Multiple reports (Politico, ProPublica, NPR) document that the apparatus was reduced by approximately 90 percent in 2025 under Defense Secretary Hegseth. The reduction proceeded against the recorded objections of senior commanders including CENTCOM commander Gen. Erik Kurilla.43

In the same period, Hegseth fired the top three service Judge Advocates General. JAGs are the legal officers who advise commanders during target review. According to subsequent reporting, the remaining JAG corps became “increasingly nervous about vocalizing their independent legal opinions.”44

Hegseth’s public rhetoric framed the change. “No stupid rules of engagement.” “Death and destruction from the sky all day long.” “No quarter.”

Senators Reed, Schumer, Warren, and Van Hollen issued formal warnings that some of this language risked crossing into the prohibited-orders category under customary international humanitarian law.45

The institutional context brackets the AR 15-6 finding. The Pentagon office created to prevent strikes built on outdated targeting data had been gutted in the year before the strike that was built on outdated targeting data. The legal review function had been chilled. The doctrinal infrastructure for high-tempo precision-air operations against an enemy in dense urban environments was operating, by the time of Operation Epic Fury, with about a tenth of the civilian-protection capacity Congress had legislated three years earlier.

The Five Red Flags Revisited

The forensic framework from The Grief Factory identifies diagnostic patterns of atrocity narrative inflation. With sixty days of additional evidence, here is how the original red flags hold up.

1. Sole-source information. This applied (and still applies) to the regime’s casualty figures. Every number originating inside Iran was propagated through state and semi-official media during a near-total internet blackout (Cloudflare measured a 98 percent drop in connectivity). What has changed is that the strike, the munition, the impact pattern, and the school’s civilian status are now established by non-regime sources working at distance: BBC Verify, CNN, NYT, AP, Reuters, Bellingcat, Amnesty Evidence Lab, HRW Digital Investigations Lab. The casualty floor (Hengaw’s 58) is non-regime as well. The casualty ceiling remains regime-only.46

2. Escalating numbers. The eight-revision Day-1 escalation from 5 to 85 was documented in the original article. The regime ceiling continued to rise after that, stabilizing at approximately 168 over the next 48 to 96 hours. Hengaw’s verified count of 58, after three weeks of named-victim work, frames the regime trajectory as something other than gradual refinement of an evolving disaster scene. The roughly threefold gap between what was claimed and what could be verified by name is consistent with the regime’s record across contested events from 1988 through January 2026.

3. Immediate political exploitation. Araghchi’s X post arrived within hours of the strike, before any independent verification was possible. The Iranian Ministry of Foreign Affairs subsequently distributed an AI-generated Lego-styled video of the incident produced by the Revayat-e Fath Institute. The pattern of pre-scripted diplomatic exploitation matches the 2019 SNSC directive (cited in The Grief Factory) and the 2022 Armed Forces order documented by Amnesty International.47

4. No independent access. Partially superseded. No journalist has visited the Minab site as of May 1. No Red Cross or MSF team has assessed casualties on the ground. But the open-source intelligence community, working with commercial satellite imagery (Planet Labs, Maxar), geolocated video, named-source interviews with Iranians outside the country, and document analysis, bridged the gaps that would otherwise be insurmountable. The picture that emerged is the most fully verified contested casualty event in the Islamic Republic’s history, not because the regime allowed access, but because the verification methods evolved.

5. Pattern match. Still valid. During the June 2025 Operation Rising Lion (the Twelve-Day War), Iranian state media framed the campaign as overwhelmingly civilian-targeted, with senior officials claiming approximately 90 percent of Iranian casualties were civilians. HRANA’s named-victim breakdown of 1,190 documented dead from that conflict found 436 civilians, 435 military, and 319 unidentified — a roughly even split between civilian and military, not the 90 percent civilian share the regime had claimed. The Minab framing continues that template, applied to a single incident at greater scale. What is new in the 2026 case is that the U.S. side of the story (the vetting failure) is also documented. Both halves of the picture, the regime inflation and the U.S. institutional failure, are independently sourced.48

What Remains Contested

The original “What Would Settle It” section listed evidence not yet available. Sixty days later, much of that evidence exists. The shorter list of what still does not:

The final death toll between Hengaw’s verified 58 and the regime’s claimed 168.

The role of AI-assisted geospatial tools in propagating the legacy DIA classification, and whether they failed to flag the school’s conversion to civilian use.

Whether the strike-package timeline allowed the CDE Level 5 review that doctrine requires when civilian casualties are anticipated.

The level of authority that approved the strike and any waivers granted to the high-CDE process.

Whether the institutional failure was deliberate, reckless, or an honest mistake. Yale law professor Oona Hathaway told the press that an honest mistake would not necessarily be a war crime, but that “reckless lack of care” might be. The AR 15-6 conclusion, when published, will be read against that distinction.49

The UN Independent International Fact-Finding Mission on Iran (chair Sara Hossein) opened its investigation March 17. Its report is anticipated in September 2026.50

The Comparative Frame

Hengaw’s day-40 aggregate of 1,030 civilian deaths against 7,650 total fatalities works out to a 13.5 percent civilian-casualty rate, or a roughly 6.4 to 1 ratio of military to civilian deaths. Read against the record of stand-off precision air campaigns over the past three decades, that ratio sits at the favorable end of the modern band.

CampaignDurationCivilian deaths (independent)Military:civilian ratio
Israel-Hezbollah, Lebanon (2006)34 days1,000+ (HRW)~1:2
NATO Allied Force, Kosovo (1999)78 days489-528 (HRW)1:1 to 3:1
US-led invasion, Iraq (2003 air phase)~42 days3,750-7,400 (PDA, IBC)1.2:1 to 2.5:1
Coalition anti-ISIS, Iraq + Syria (2014-19)~5 years8,000-13,000 (Airwars)~3:1 to 5:1
US-led, Gulf War (1991 air phase)43 days2,278-3,500 (HRW, IBC)3:1 to 8:1
Iran 2025-2026 (Hengaw-anchored)52 days~1,466 named~4.8:1 to 6.4:1
NATO Unified Protector, Libya (2011)222 days72-403 (HRW, Airwars)5:1 to 10:1

Iran 2026 is more favorable than Kosovo 1999, the 2003 Iraq invasion, and the coalition anti-ISIS campaign. It sits roughly at par with Gulf 1991. It is dramatically more favorable than the dense-urban precedent of Lebanon 2006. It is modestly less favorable than Libya 2011, the often-cited modern best case. The Amiriyah shelter strike of February 13, 1991, in which 408 civilians died in a single incident, remains the high-water mark for U.S.-attributable civilian harm in any single strike since the Gulf War. Minab does not approach that figure even at the regime’s claimed count.51

A second test is the expected-ratio counterfactual. Given the documented target mix in Operation Epic Fury (roughly 60 to 70 percent IRGC bases, missile-production complexes, and air defenses; 20 to 25 percent Tehran-embedded leadership, intelligence, and command nodes; 5 to 10 percent dual-use or contested targets), CDE-Level-4-or-better precautions would predict a campaign-wide civilian share in the 9 to 13 percent band. The observed 13.5 percent sits at the top of that predicted band, within the range standard methodology would expect rather than anomalously above it.52

A third test is the reverse counterfactual. If the IRGC had garrisoned bases away from civilian neighborhoods and had not operated underground tunnel complexes beneath Tehran schools, hospitals, and mosques, applying NATO Libya 2011 norms to a 5,000-target campaign of comparable munition mix would predict roughly 200 to 400 Iranian civilian deaths, not the approximately 1,030 Hengaw documents. The 600 to 800 deaths that fall above the no-co-location counterfactual are defender-attributable under Article 58 of Additional Protocol I to the Geneva Conventions. Article 51(8) holds attacker obligations under Articles 51(5)(b) and 57 undiminished regardless. The legal framing and the political framing diverge on the same numbers: under the Schmitt and ICRC consensus, the attacker bears full proportionality liability for shielded civilians; under the Dinstein-leaning framing, the defender absorbs much of the moral culpability for civilian deaths attributable to its own Article 58 violations.53

The campaign-wide picture, in other words, is not a story of indiscriminate harm. The IDF and the U.S. employed precision weapons throughout, the headline ratio sits at the favorable end of the modern historical band, and the observed civilian share is what the documented target mix would predict. The Center for Civilians in Conflict, Human Rights Activists, and Airwars joint report (March 27) identified the factors that pushed the rate to the top rather than the floor of the predicted band: the campaign tempo enabled by Maven, the gutting of CHMR, Hegseth’s “no stupid rules of engagement” rhetoric, IRGC co-location practices, and the targeting of dense urban environments including Tehran (population approximately nine million).54

Within that picture, Minab is the apex case. The single strike accounts for between 6 and 17 percent of all Phase 2 civilian deaths depending on which casualty count is used. Excluding it, the campaign-wide ratio improves to roughly 7 to 1. The strike is structurally distinct from the campaign average: the largest single civilian casualty incident, attributable to a documented institutional vetting failure, occurring in a campaign whose civilian-protection infrastructure had been systematically dismantled in the preceding year. The institutional failure documented at Minab is real and consequential. The campaign-wide ratio it sits inside is not anomalous against the historical record. The forensic framework holds both readings at once.55

The Honest Assessment

Three things can be simultaneously true.

Children died at the Shajareh Tayyebeh school. The verified floor is 58 named individuals. The ceiling is unresolved. The school was operating during school hours when struck. Some number greater than zero and less than or equal to the regime’s claimed 168 children, teachers, and parents was killed in a strike that the U.S. military’s own preliminary investigation has attributed to outdated intelligence data.

The U.S. military, by its own preliminary admission, struck a school it should have known was a school. The school’s separation from the IRGC base was visible in commercial satellite imagery from 2017 onward. The school had a public website. The DIA target data the strike was built on was approximately a decade out of date. The Pentagon office created to prevent exactly this kind of error had been reduced by approximately 90 percent in 2025. None of this requires believing any regime claim.

And the regime that placed a military base sixty meters from an elementary school, that operates the school as part of a 32-school IRGC Navy recruitment network with admission preferences for personnel families, that has a documented four-decade system for inflating, manufacturing, and weaponizing civilian casualty claims, that imposed an internet blackout specifically preventing independent verification, is the sole source for the upper portion of the casualty range.56

These three claims are not in tension. They are all true. The honest framing holds each.

Human rights lawyer Hossein Raisi stated the legal principle: “Any party that knowingly turns a civilian site into a center of military activity during armed conflict commits a war crime.” Political analyst Jamshid Barzegar was more direct: “The Islamic Republic has once again brought inside the country a criminal pattern it has tested for years through its proxy forces in the region: militarizing civilian spaces and using civilians as human shields.”57

Both observations are about the regime side. The U.S. side has its own legal frame. Article 57 of Additional Protocol I to the Geneva Conventions imposes the attacker’s duty to do “everything feasible to verify that the objectives to be attacked are neither civilians nor civilian objects.” Just Security’s analysis: a strike package built on intelligence data years out of date, for a fixed installation in a non-denied access environment, against an object whose civilian conversion was visible in commercial satellite imagery and on the school’s own website, likely fails the feasibility test. Customary IHL Rule 140 holds that the U.S. obligation to comply with these rules does not depend on Iran’s conduct.58

On March 11, with the AR 15-6 inquiry opening, Hegseth told reporters: “Whatever the report shows, I’m willing to live with that report.”59

The report is still pending.

If your child’s school sat sixty meters from a military base, administered by a regime that had walled off the buildings but kept admission preferences for personnel families, and you learned this only after the bombs fell, the betrayal would be absolute, regardless of who dropped them. And if you learned that the military that dropped them had built its strike package on ten-year-old maps because the office that would have caught the error had been deliberately gutted the year before, that betrayal would be absolute too.

The question is not whether children are worth grieving. Every child is worth grieving. The question is whether a tragic event in which all three claims are true (a real strike, a real institutional vetting failure, a real four-decade pattern of regime inflation) can be reported in a way that holds all three.

That is what the forensic framework does. It documents the verified, cites the regime-claimed as regime-claimed, marks the unknown as unknown, and holds the institution that struck and the institution that propagandized to their separately documented records.

The fuller picture is less comfortable for both sides than the early framing allowed. It does not approach the framing the regime continues to broadcast through state cameras. It does not approach the framing that any “honest mistake” account would prefer either.

This article is part of The Grief Factory. For the five-stage anatomy of regime atrocity laundering, see Kill, Withhold, Coerce, Broadcast. For the historical pattern of regime atrocity claims, see The Same Playbook Since 1988. For the institutional failure pattern of U.S. precision-strike vetting, see How Targeting Actually Works. For the credibility ledger across contested events, see The Credibility Ledger.

This article first ran February 28, 2026 with Day-1 evidence available at the time. It was substantially updated on May 2, 2026 to incorporate the sixty days of subsequent reporting and investigation. The forensic framework is unchanged. The data inside it has substantially evolved.

Footnotes

  1. School staff member eyewitness account via Middle East Eye. The only firsthand account from inside the building to reach international media in the first hours. ↩

  2. FM Abbas Araghchi post on X, February 28, 2026. Full quote cited. ↩

  3. Washington Post verified video showing debris, collapsed wall, and smoke. NBC News independently verified additional footage. ↩

  4. Tomahawk attribution chain: BBC Verify, CNN, NYT, AP, Reuters, Bellingcat, Amnesty Evidence Lab, HRW Digital Investigations Lab. Caine March 2 Pentagon briefing confirmed Tomahawks were the opening salvo. Caine March 4 100-hour strike-locations map included Minab. Sources: CNN March 9, Bellingcat March 27, NYT March 11. ↩

  5. NYT, March 11, 2026, citing officials briefed on the AR 15-6 inquiry. Outdated DIA targeting data classified the school as part of the IRGC compound it had been walled off from at least a decade earlier. ↩

  6. Hengaw 25-day verified count: 58 victims (48 children, 10 adults). Source: Hengaw Organization for Human Rights, official report. ↩

  7. Saturday is Iran’s first workday. School in session during morning shift. Shajareh Tayyebeh school in Minab, Hormozgan Province. ↩

  8. Regime casualty figures: Mizan News (judiciary) 110 schoolchildren + 26 teachers + 4 parents (March 3); President Pezeshkian “more than 170 people killed” (March 3, CBS News); range 156-175 across regime sources; IFP News final regime canonical figure 168 by 50 days post-strike. ↩

  9. “Triple tap” framing per Iran’s Ministry of Education. “Double tap” framing per Middle East Eye, citing two Iranian Red Crescent medics and one parent. ↩

  10. CENTCOM confirmed Operation Epic Fury, 1:15 AM ET, February 28, 2026. Approximately 900 strikes in first 12 hours, exceeding 1,000 in first 24, 1,250 in first 48; over 5,000 targets struck in first 10 days per Pentagon fact sheet at media.defense.gov. ↩

  11. Captain Tim Hawkins, CENTCOM spokesperson, to CBS News, March 1, 2026: “had no statement to share regarding the alleged strike on the school.” Pentagon to CNN: “had nothing to share at this time.” ↩

  12. Caine Pentagon briefing March 2, 2026: “The first weapons launched at sea were Tomahawks fired by the United States Navy.” 100-hour strike-locations map displayed at Pentagon March 4 included Minab. ↩

  13. Hegseth, March 4, 2026: “All I can say is that we’re investigating that. We of course never target civilian targets.” ↩

  14. Trump, March 8, 2026 (with Hegseth standing behind): “Based on what I’ve seen, that was done by Iran. They are very inaccurate, as you know, with their munitions. They have no accuracy whatsoever. It was done by Iran.” Trump, March 9, 2026: “Iran or somebody else. A Tomahawk is very generic. Because I just don’t know enough about it.” ↩

  15. Sen. John Kennedy (R-LA), March 10, 2026, breaking from Republican non-comment line: U.S. “made a mistake,” strike “was not intentional,” “I’m sorry.” ↩

  16. NYT, March 11, 2026, citing officials briefed on AR 15-6 preliminary findings: CENTCOM “created the target coordinates for the strike using outdated information provided by the Defense Intelligence Agency.” ↩

  17. Hegseth press conference, March 13, 2026: formally announced AR 15-6 investigation under U.S. general officer from outside CENTCOM. AR 15-6 is the Army Regulation 15-6 administrative investigation, named for the Army regulation that prescribes commander-directed fact-finding. ↩

  18. IDF to HRW and reporters, March 6, 2026: “Not aware of any IDF activity in the area.” Tomahawk operators: U.S., UK, Australia, Japan, Netherlands; only the U.S. is operationally engaged in this conflict. ↩

  19. CNN geolocation: ~60 m / 200 ft from IRGC base. BBC Verify journalists Sardarizadeh, Osieck, Tabrizy: coordinates 27.109787 N, 57.084671 E. Compound identified as IRGC Naval Forces / Sayyid al-Shuhada / Asef (also Asif) Missile Brigade. ↩

  20. School demographics: ages 7-12, girls and boys on separate floors. Source: Wikipedia, BBC, Iran’s Ministry of Education. ↩

  21. Tomahawk identification: Mehr News construction-site video March 9, 2026 (cruciform shape, centrally mounted wings, tailkit). Munitions experts Sam Lair (James Martin CNS), John Gilbert (Center for Arms Control and Non-Proliferation), N.R. Jenzen-Jones (ARES) confirmed identification. ↩

  22. NYT traced contract code on munition fragment photographed by IRIB March 9 to Tomahawk procurement. ↩

  23. Bellingcat March 27, 2026, analysis of two videos released by Iran’s Foreign Ministry: confirmed Tomahawk identification and identified two waves of strikes ~1 hour apart. ↩

  24. HRW noted “distinctive pancaking of the roof” consistent with top-down precision impact. Jenzen-Jones (ARES): multi-impact pattern across 8-10 distinct compound structures “consistent with highly accurate, guided weapons,” ruling out Iranian air-defense interceptor misfire. ↩

  25. Bellingcat March 27 strike sequence: two waves ~1 hour apart, ~10 missile impacts in 50-second window in first wave. HRW analysis of March 4 Planet Labs commercial satellite imagery: 8+ visible impact sites, 10 likely direct hits. ↩

  26. Three framings of strike sequence: “triple tap” (Iran Ministry of Education), “double tap” (Middle East Eye, citing 2 Red Crescent medics + 1 parent), “two-wave” (Bellingcat). U.S. officials have not endorsed any. ↩

  27. School separation timeline: 2013 unified compound; September 2016 wall + 3 street entrances + 4 military checkpoints removed; August 2017 outdoor play visible; 2022-2023 medical clinic walled off (Salami ribbon-cutting per NPR Jane Arraf and Geoff Brumfiel); December 2025 commercial imagery showing courtyard activity (CNN). Mayor of Minab quoted by HRW: original base “had been closed for about 15 years.” Reuters: “vivid website and yearslong online presence.” Guardian: “no indication that the building served a military purpose.” Sources: HRW, CNN, NYT, AP, Amnesty Evidence Lab, TIME. ↩

  28. NYT preliminary AR 15-6 finding (March 11, 2026), citing officials briefed on the inquiry: outdated DIA targeting data classified the school as part of the IRGC compound it had been walled off from a decade earlier. ↩

  29. Just Security legal analysis by Joseph N. Orenstein. JP 3-60 Phase 2 (target development) failure. Appendix A para. 4(b)(7): currency requirement, all-source review. Conclusion: strike “rises well above the threshold of an unremarkable mistake.” ↩

  30. Sarah Yager, Human Rights Watch Washington director, March 12, 2026. ↩

  31. HRW question on no-strike list status: “If the school was on a ‘no strike’ list, why was the list not referenced prior to targeting Minab?” ↩

  32. Senate Democratic Caucus letter, March 11, 2026, signed by 46 senators (Schumer, Van Hollen, Kaine, Schatz, Warren as leads, plus 41 others including Reed, Shaheen, Bennet, Markey, Sanders, King). Demanded answers on AI tool use, approval authority, intended target, CDE Level 5, distinction/proportionality/precaution as “binding and non-negotiable standards.” Response demanded by March 18, 2026; still pending at Hegseth’s April 29 House testimony — seven weeks past deadline. Sources: Van Hollen, Kaine, Warren, Schatz, Bennet, Shaheen senate.gov press releases, March 11, 2026; The Hill; NationofChange; full letter PDF at warren.senate.gov/imo/media/doc/letter_from_warren_van_hollen_dems_to_secretary_hegseth_on_civilian_harm_war_in_iran.pdf. ↩

  33. Hengaw 25-day named-victim verification: 58 victims confirmed (48 children, 10 adults). Verbatim quote published by Hengaw Organization for Human Rights. Methodology: distance verification through documentary sources and named-source interviews, no on-site access. ↩

  34. Amiriyah shelter strike: February 13, 1991, Baghdad, ~408 civilians killed in single incident. Previous high-water mark for U.S.-attributable civilian harm in single strike since 1991. Source: NPR’s Quil Lawrence reporting on Minab as largest U.S.-attributable civilian incident in approximately 35 years. ↩

  35. Hengaw day-40 aggregate (April 8, 2026 ceasefire): 7,650 total killed; 1,030 civilians (13.5%); 189 minors, 215 women; 6,620 military. ↩

  36. Maven Smart System: Palantir-built consolidation of “eight or nine” prior systems. Pentagon contract scaled from $480M (2024) to $1.3B (2025). CENTCOM commander Adm. Brad Cooper March 11, 2026, credited “AI tools” with helping U.S. military during Operation Epic Fury. ↩

  37. CDAO Cameron Stanley at AIPCON, mid-March 2026: described MSS consolidation, displayed Maven map of Iran with red icons including Minab area. Public capability claim: “1,000 targeting recommendations per hour.” Campaign tempo: ~1,000 strikes in first 24 hours, 5,500-6,000 in three weeks. ↩

  38. Anthropic-DOD timeline: $200M prototype OTA July 2025; February 27, 2026 (one day before strike) Hegseth designated Anthropic “supply chain risk,” first time applied to American company; Anthropic sued; federal preliminary injunction late March blocked designation as “classic First Amendment retaliation”; Claude reportedly continued during 6-month phase-out. Sources: CNBC, Anthropic corporate statements, The Hill. ↩

  39. 2024 internal Pentagon test: Maven object-recognition accuracy approximately 60%. Source: Tech Brew. ↩

  40. NYT preliminary reporting on AR 15-6: AI tools “unlikely” to be the primary cause of the Minab error. ↩

  41. Joseph N. Orenstein, Just Security, on Maven Smart System and legacy DIA target codes. ↩

  42. CHMR origin: FY2023 NDAA. Trigger: August 2021 Kabul drone strike on Zemari Ahmadi family. Pattern: pattern-of-life misidentification, post-strike defensive narrative, administrative-only accountability. ↩

  43. CHMR reduction: ~90% in 2025 under Hegseth, against recorded objections of CENTCOM commander Gen. Erik Kurilla. Sources: Politico, ProPublica, NPR. ↩

  44. Hegseth fired top 3 service Judge Advocates General in 2025. Subsequent reporting: remaining JAG corps “increasingly nervous about vocalizing their independent legal opinions.” ↩

  45. Hegseth public rhetoric: “no stupid rules of engagement,” “death and destruction from the sky all day long,” “no quarter.” Senators Reed, Schumer, Warren, Van Hollen formal warnings: language risks crossing into prohibited-orders category under customary IHL. ↩

  46. Cloudflare Radar measured 98% drop in Iranian internet connectivity from February 28, 2026 onward. Sources: Cloudflare, NetBlocks. ↩

  47. Iranian MFA distributed AI-generated Lego-styled video of incident produced by Revayat-e Fath Institute. Pattern of pre-scripted exploitation matches 2019 SNSC directive (obtained by Edalat-e Ali, provided to VOA Persian) and 2022 Armed Forces order obtained by Amnesty International. ↩

  48. June 2025 (Operation Rising Lion / Twelve-Day War). Iranian state media’s “approximately 90 percent civilian casualties” framing contradicted by HRANA’s documented breakdown of 1,190 named dead: 436 civilians, 435 military, 319 unidentified. Source: HRANA documentation via UNDP / UN Security Council briefing reporting and “Casualties of the Iran-Israel War” tracking. ↩

  49. Yale law professor Oona Hathaway: honest mistake would not necessarily be a war crime, but “reckless lack of care” might be. AR 15-6 conclusion will be read against this distinction. ↩

  50. UN Independent International Fact-Finding Mission on Iran (chair Sara Hossein) opened Minab investigation March 17, 2026. Report anticipated September 2026. ↩

  51. Precision-air campaign benchmarks. Kosovo (NATO Operation Allied Force, 1999): 38,000 sorties, 78 days, ~500 documented civilian deaths in 90 incidents (HRW). Libya (NATO Operation Unified Protector, 2011): 72 documented civilian deaths in 7 incidents over 7 months (HRW). Iraq invasion 2003: ~3,200-4,300 documented civilian deaths over 6 weeks (Project on Defense Alternatives, Iraq Body Count). Amiriyah shelter (Baghdad, February 13, 1991): 408 civilians killed in single incident. ↩

  52. Expected-ratio counterfactual applies CDE Methodology (Chairman of the Joint Chiefs of Staff Instruction 3160.01, Levels 1-5) to the documented campaign target mix. Target catalogue composition derived from CENTCOM, IDF, ACLED, Hudson Institute, CSIS, and TWZ campaign reporting through April 2026. Predicted civilian share calculated by weighting target categories against historical civilian-incidence rates: ~3-6 percent for IRGC bases / missile production / air defense in semi-rural settings; ~10-18 percent for Tehran-embedded leadership and command nodes; 20-40 percent for dual-use or contested targets even with maximum precaution. ↩

  53. Reverse counterfactual applies NATO Libya 2011 civilian-share norms (~4-8 percent civilian share per HRW and Airwars retrospective) to the Iran 2026 munition-mix and target-count reported by CENTCOM. AP I Article 51(8): defender violation does not relieve attacker of obligations under Articles 57 (precaution) and 51(5)(b) (proportionality). Schmitt-ICRC consensus per Michael N. Schmitt, International Review of the Red Cross 872; ICRC humanitarian law commentary, “Human Shields and IHL.” Dinstein-leaning framing per Yoram Dinstein, The Conduct of Hostilities under the Law of International Armed Conflict. ↩

  54. Joint report by Center for Civilians in Conflict (CIVIC), Human Rights Activists (HRA), and Airwars, March 27, 2026: identified drivers as Maven-enabled tempo, CHMR gutting, Hegseth rhetoric, IRGC co-location, urban targeting including Tehran (~9 million population). 1,443 civilians killed Feb 28-March 23 including 217 children; 130 distinct civilian-harm incidents; 60 hospitals, 44 schools, 129 residential buildings damaged. ↩

  55. Historical comparison band sourced from Human Rights Watch field investigations (Kosovo, Lebanon 2006, Iraq 2003); HRW Needless Deaths report and Iraq Body Count (Gulf 1991 air phase); Project on Defense Alternatives (2003 invasion); Airwars 2021 retrospective (Libya); Airwars cumulative tracking (Coalition anti-ISIS). Combatant classification follows continuous-combat-function logic per AP I Article 43 and ICRC commentary. Minab share of Phase 2 civilian deaths: 58 (Hengaw-verified) to 168 (regime-claimed) of approximately 1,030 documented Phase-2 civilian deaths in the Hengaw day-40 aggregate. ↩

  56. 32-school Shajareh Tayebeh network operated by IRGC Navy across Khuzestan, Bushehr, Hormozgan, Sistan-Baluchistan. Recruitment courses for ages 13+. Minab Shajareh Tayyebeh is a primary school (ages 7-12), structurally distinct from the recruitment-stream secondary courses. Al Jazeera investigation: admission registration on Baleh messaging app prioritizes children of IRGC Navy personnel. Source: IranWire pre-war documentation, Al Jazeera, FactNameh. ↩

  57. Hossein Raisi, human rights lawyer, on legal principle. Jamshid Barzegar, political analyst, on pattern. ↩

  58. Article 57, Additional Protocol I to the Geneva Conventions: attacker’s duty of feasibility. Just Security analysis: Minab strike package likely fails feasibility test. Customary IHL Rule 140: U.S. obligation to comply does not depend on Iran’s conduct. ↩

  59. Hegseth, March 11, 2026: “Whatever the report shows, I’m willing to live with that report.” ↩