Iran Today

The Verification Floor

by Iran Prism 20 min read

The Floor, Published

On April 8, 2026, the day of the ceasefire that ended Operation Epic Fury, the Hengaw Organization for Human Rights published its tenth serial report on the war. The headline figure was 7,650 deaths inside Iran across forty days. Six thousand six hundred twenty were classified as military or government personnel. One thousand thirty were classified as civilians. Among the civilian count, 189 were minors and 215 were women. Documentation extended across 196 cities in 27 of Iran’s 31 provinces.1

That report did not arrive through Iran. It arrived through Iraq’s Kurdish region, through Persian-diaspora networks in Norway and elsewhere, through hospital sources who passed information out during brief connectivity windows, through family members who reached relatives abroad before the regime could stop them. It arrived because Hengaw had spent years building a documentation infrastructure that did not depend on Iranian state cooperation or on Iranian internet connectivity, both of which were absent during the war.

The number Hengaw published is what named-victim methodology produces when applied to a war that occurred under an internet blackout. Each entry is a person identified by name. Each name is paired with documentary anchors: an image, a death certificate, a hospital record, a family identifier, an eyewitness statement. Each entry is reviewed against parallel documentation from HRANA, Airwars, Amnesty International, and the UN Independent International Fact-Finding Mission on Iran. Each report in the serial sequence applies the same definitions. The report is published with explicit acknowledgment of what is verified, what is contested, and what is unknown.

That is what stands between the published number and the alternatives: regime aggregates issued during a blackout, social media counts that include duplicate entries and unverified claims, and AI-generated images circulating as documentary evidence. The methodology is the difference between a casualty count and political theater. This article walks through it.

The Blackout

Iran’s internet connectivity sat at one to four percent of pre-crisis levels for more than fifty-nine consecutive days during Operation Epic Fury. NetBlocks recorded the longest state-imposed shutdown ever logged anywhere, surpassing Myanmar’s Rohingya-period blackout, the Tigray war shutdown, and Iran’s own January 2026 shutdown that preceded the war. Cloudflare Radar measured a ninety-eight percent drop on February 28, the first day of the campaign. The cumulative cost to the Iranian economy ran to at least one point eight to one point nine billion dollars in the first forty-eight days alone.2

The blackout produced three documented effects on casualty documentation, each of which the named-victim methodology had to mitigate.

Verification delay. Incidents that occurred in peripheral provinces surfaced days or weeks after they happened, as documentation crossed borders, reached diaspora networks, or arrived through brief connectivity windows when residents could send a single image before the connection died. The documentation infrastructure was operating in arrears throughout the campaign. Several incidents that Hengaw documented in early April had occurred in early March; the Day-40 report’s serial trajectory captured this lag explicitly.

Under-counting in peripheral provinces. Hengaw’s documentation network is densest in Iran’s Kurdish provinces, where the Iraq-based Kurdish documentation infrastructure can operate with relative independence. Coverage in Sistan and Baluchistan, in eastern Khorasan, and in remote Khuzestan was thinner. The Day-40 report acknowledges this gap: provinces where documentation was densest were also where civilian counts were highest, but the ratio is partly an artifact of what could be verified rather than where deaths actually concentrated.

Regime-controlled narrative monopoly. During the blackout, Iranian-language sourcing inside Iran was overwhelmingly produced by IRNA, Press TV, Tasnim, IRIB, Mehr, and the Iranian Health Ministry. Independent Iranian outlets that had operated through the early 2020s had been progressively shut down. The result: the only Iranian-language casualty figures that crossed the border in real time were regime-produced. Western outlets that quoted “Iranian Ministry of Health figures” were quoting one tier of evidence; Hengaw and HRANA were producing a separate tier through methodology that did not depend on the regime’s information environment.

Iranian authorities arrested more than 1,800 people for internet-related offenses during the campaign, expelled foreign journalists, restricted Red Crescent reporting, and conducted door-to-door seizures of Starlink terminals and satellite dishes. The infrastructure for documentation was actively suppressed at the same time that documentation was most needed.

That the verification floor exists at all under those conditions is itself a methodological achievement. What it took to produce is what the next section walks through.

What Hengaw Does

Hengaw is a Norway-based Kurdish human rights organization with field documentation infrastructure across Iran’s Kurdish provinces (Kermanshah, Ilam, West Azerbaijan, Kurdistan) and Iraq’s Kurdistan Region (Erbil, Sulaymaniyah, Duhok). The Iraq-based field network is the operational backbone of Hengaw’s documentation capacity. Iraqi Kurdistan is geographically and culturally close to Iran’s Kurdish regions, hosts a substantial Iranian Kurdish diaspora, and operates outside Iranian internet connectivity. When Iran’s networks went dark on February 28, the Iraq-based infrastructure kept working.3

The methodology has three load-bearing components.

Each death is identified by name. Hengaw does not publish aggregate counts that lack named-individual documentation. The Day-40 figure of 7,650 deaths is built from approximately that many individual files. Each file contains the victim’s name, age, location of death, date of death, classification (military or civilian), and at least one documentary anchor.

Each name is paired with documentary anchors. The anchor categories include images from family members or eyewitnesses, death certificates, hospital records, family identifier statements (a relative confirming the death), and eyewitness testimony. A single anchor is not sufficient for inclusion; a typical entry has at least two corroborating sources. Hengaw publishes this corroboration requirement explicitly and applies it consistently across categories.

Each entry is reviewed for cross-corroboration against parallel documentation efforts. HRANA, Airwars, Amnesty International, and the UN Fact-Finding Mission on Iran all maintain their own databases. Hengaw cross-references against these and excludes entries that other organizations have flagged as contested or duplicated.

What Hengaw refuses to do is equally important. It does not publish casualty figures based on social media counts. It does not include AI-generated images or recycled footage in its documentation. It does not extrapolate from named verified entries to estimate larger unverified totals (HRANA does this in some contexts; Hengaw does not). It does not publish under regime pressure to revise figures upward or downward. Its serial reports apply the same definitions throughout the campaign, allowing the trajectory to be followed and audited.

The result is a number that is conservatively scoped. Hengaw’s 7,650 is a floor, not a ceiling. True deaths inside Iran during Operation Epic Fury are almost certainly higher than this figure. But what Hengaw publishes is what Hengaw can defend by name, with documentary anchors, cross-corroborated against parallel methodologies, applying definitions held constant across the campaign.

What HRANA Does Differently

The Human Rights Activists News Agency runs a parallel documentation effort with a different methodological signature. HRANA’s cumulative count through April 7, 2026, recorded 3,636 documented strike deaths: 1,221 military, 1,701 civilians, and 714 unclassified. The civilian share of total documented deaths in HRANA’s data is approximately forty-seven percent (1,701 of 3,636); the share among classified deaths only (excluding the 714 unclassified) rises to approximately fifty-eight percent. Either framing is far higher than Hengaw’s thirteen point five percent.4

The gap between HRANA’s apparent civilian share and Hengaw’s is the kind of methodological discrepancy that needs to be explained before either number can be used. HRANA explains it explicitly. The same HRANA report contains the following caveat:

“It is believed that military casualties are significantly higher than the figures reported in [its] reports… confirmations depend largely on government data that are obscured… [its] count largely reflects reports of senior officers or military personnel who were present in urban areas.”

Translated: HRANA’s military denominator is biased downward by the regime’s systematic concealment of military KIA. The regime publishes funeral notices for senior officers killed in identifiable urban incidents (a Khademi-class general in central Tehran). It does not publish any equivalent record for the bulk of IRGC and Basij personnel killed at military installations across the country. HRANA’s documentation infrastructure can verify the urban senior-officer cases by name; it cannot verify the broader military pool because the underlying records are not produced.

The result: HRANA’s apparent civilian share, whether computed against total documented or against classified-only, is an artifact of the missing military denominator, not a measurement of the campaign’s true ratio. HRANA itself says this in its report. Any analysis that takes HRANA’s apparent civilian share as the campaign-wide ratio is reading the data against HRANA’s own methodological caveat.

Hengaw’s military count includes Basij personnel, FARAJA (Law Enforcement Forces), Ministry of Intelligence operatives, judiciary security personnel, and foreign militia members deployed inside Iran. These categories are documented through different sources than the senior IRGC officers HRANA’s caveat references. HRANA’s own caveat acknowledges that its military count “largely reflects reports of senior officers or military personnel who were present in urban areas,” because that is the subset of regime military deaths that produces public funeral notices and identifiable urban incidents the regime cannot fully suppress. The broader pool of Basij, FARAJA, and provincial-base IRGC deaths the regime systematically conceals does not produce the documentary anchors HRANA’s military-track verification requires. The two organizations diverge on military counts in the same direction (Hengaw 6,620; HRANA 1,221 with HRANA’s own under-count caveat), and they diverge on civilian counts in the opposite direction (Hengaw 1,030; HRANA 1,701).

The civilian divergence runs in the opposite direction from the military divergence. HRANA’s civilian count is higher than Hengaw’s; Hengaw’s military count is higher than HRANA’s. The reasons for the civilian gap are not as cleanly attributable as the reasons for the military gap. Two methodological differences are documented in the source material. First, HRANA publishes 714 entries it classifies as “unclassified” pending further verification. Hengaw does not maintain a public unclassified category; entries either meet the named-victim threshold or are not published. HRANA’s threshold therefore admits cases at lower verification confidence than Hengaw’s, weighted toward civilian classification under the IHL default (“when in doubt, classify as civilian”). Second, HRANA aggregates through a four-team verification structure (HRANA news agency, Spreading Justice, Pasdaran Documentation Project, and a Statistics Department); Hengaw aggregates primarily through its Iraq-Kurdistan field network and Persian-diaspora corroboration. Whether the threshold difference and the aggregation-structure difference together fully explain the gap, or whether other factors also contribute, is not isolated by either organization’s published methodology. What both databases agree on is the directional finding: the joint Airwars-CIVIC-HRA report covering only February 28 to March 23 (the first three and a half weeks) documented at least 1,443 civilian deaths, drawing on HRA’s verification network for victim counts plus Airwars’ methodology for incident framing, and exceeding either single-organization count for that partial period. The reading the joint count supports is that true documented civilian deaths are higher than the 1,030 Hengaw publishes; both single-organization counts are floors. The 1,030 is the most conservative single-organization floor and is therefore the defensible anchor for the campaign-wide ratio.

Hengaw’s 6.4:1 is the defensible single-organization anchor because it applies the most conservative named-victim methodology to a documentation infrastructure that operated outside Iranian connectivity throughout the blackout. HRANA’s apparent civilian-heavier ratio does not reconcile numerically to Hengaw’s even after correcting for HRANA’s acknowledged military under-count, because the threshold and aggregation-structure differences push HRANA’s civilian numerator higher in the opposite direction. The two databases cross-check each other methodologically rather than producing identical numerical ratios. What both confirm is that named-verified counts are dramatically below regime-claimed counts, which is the directional finding the campaign-wide ratio actually rests on.

The Twelve-Day War Cross-Reference

The methodology divergence between HRANA and Hengaw is not a feature of the 2026 war alone. The June 2025 Twelve-Day War — the same Israel-Iran conflict, the same two organizations counting Iranian deaths under the same blackout-constrained verification environment — produced ratios that diverged in the same direction. HRANA documented 1,190 deaths through June 28, 2025: 435 military, 436 civilians, and 319 unidentified, a roughly one-to-one ratio among classified deaths. Hengaw documented 1,082 deaths over the same window: 900 military and 182 civilians, a roughly six-to-one ratio. The same classification-rule differences explain the same divergence pattern across both wars: Hengaw counts uniformed Basij at checkpoints and provincial-base IRGC deaths aggressively; HRANA holds ambiguous force-affiliation cases in its unclassified bucket pending further determination. The structural divergence is what the comparative casualty-database literature predicts; identical numerical ratios across two independent named-victim methodologies sampling overlapping populations would be statistically suspicious.5

Contamination Forensics

A separate problem from the documented data is the un-documented data circulating during the war. The blackout did not eliminate Iranian-source content; it concentrated content into channels (Iranian state media, Telegram, X) where the absence of independent verification produced a high rate of contamination.

Independent fact-checkers documented the contamination patterns in real time. The New York Times identified at least 110 distinct AI-generated images and videos in the first two weeks of the campaign. NewsGuard documented at least eighteen false war-related Iranian state-media claims. Tehran Times posted an AI-manipulated image of a “destroyed US radar” that was actually a photo from Bahrain. AFP debunked images of “Tel Aviv vehicle fires” that originated in January 2026 Tehran (Kaj Square) anti-government protests. Reuters debunked “new” Iranian strike footage on Tel Aviv as June 2025 video. Recycled video-game footage from Arma 3 simulations of SU-57 air operations was viewed more than seven million times claiming to show Iranian strikes.6

A fake Telegram channel posing as IRGC published a false statement claiming accidental responsibility for Minab. Citizen Lab documented Iran-linked influence-operation infrastructure (the “PRISONBREAK” cluster) pre-staging deepfakes for synchronized release with kinetic events. Bellingcat traced staged-imagery patterns matching the broader operations of the Iranian Cyber Army. Gemini and Grok, the chatbots, falsely captioned a mass-burial bird’s-eye image as Kahramanmaraş 2023 (Gemini) or Jakarta COVID 2021 (Grok); the image was actually from Minab.

The estimated contamination rate of the unverified open-source civilian-casualty pool runs approximately fifteen to twenty-five percent based on debunked-claim ratios. This is not a marginal problem. It is the principal reason that Hengaw, HRANA, Airwars, and Amnesty all exclude uncorroborated open-source material from their named-victim totals. The named-verified figures are lower than the regime-claimed figures and much lower than the social-media-circulated figures precisely because the named-verified methodology refuses to count what cannot be documented.

The Carnegie Endowment’s work on the “liar’s dividend” effect captures the second-order problem. Legitimate Iranian victimhood claims are now skeptically received because Iranian fabrication infrastructure made everything skeptically received. US and Israeli officials gain cover to deflect credible critical reporting because the same reporting environment produces genuinely false claims. Both sides of the documentation problem are real. Both are bounded by what named-victim methodology can verify, which is the methodological floor this article describes.

The Firewall

A separate contamination risk is internal to the documentation system itself. Iran during the 2025-2026 period was producing two kinds of civilian deaths: deaths attributable to coalition strikes during Operation Epic Fury, and deaths attributable to the regime’s own ongoing operations, including the January 2026 crackdown documented in HRANA’s Crimson Winter report and the wartime executions documented by Hengaw. The risk is that deaths in the second category could migrate into the documentation pool for the first.

HRANA and Hengaw both maintain explicit category firewalls between the two pools. HRANA’s Crimson Winter report covers regime-attributable deaths from the January crackdown; HRANA’s separate war-casualty count covers strike-attributable deaths during Operation Epic Fury. The two are tracked through different intake protocols, different cross-referencing methodologies, and different publication cycles. Hengaw maintains the same firewall. The fourteen secret prison executions Hengaw documented between March 17 and April 6, 2026, are in the regime-attributable pool, not the war-civilian pool. Saleh Mohammadi, a teenage wrestler whose name appears in Hengaw’s record because the methodology requires that each death be documented as a specific person rather than as an aggregate, was killed inside a regime prison while the coalition campaign was underway. Amir Hossein Hatami, eighteen, was killed the same way in the same window. They are in Hengaw’s records as named individuals because that is what the methodology produces. They are in the regime pool, not the coalition pool, because that is what the firewall enforces.7

The image-based contamination pathway is documented but bounded. AFP’s investigation of “Tel Aviv vehicle fire” imagery that originated in January 2026 Tehran protests is one example of cross-pool image migration. The named-victim infrastructure prevents image-based contamination from entering the documented totals because images alone are not sufficient anchors; a name and at least one corroborating documentary source are required. The image-based contamination affects the open-source pool but does not enter the named-victim totals.

The defensible reading: the firewall holds at the named-victim level. The methodologies are designed to prevent cross-pool migration, and the documented behavior of Hengaw and HRANA in maintaining the firewall is consistent with their published methodologies. The 1,030 coalition-attributable civilian count and the 6,800 regime-attributable civilian subset of the 7,007-named Crimson Winter total do not overlap. Both numbers are independently sourced. Both can be cited side-by-side without double-counting.

The Discount Factor

The four-decade record of regime casualty reporting establishes a directional pattern that any sound analysis has to apply as a Bayesian prior. The regime systematically under-counts deaths it caused and over-counts deaths attributable to its enemies. This is documented inventory, not speculation. The full ledger lives in The Credibility Ledger.8

The January 2026 crackdown provides the cleanest under-count calibration on a scale comparable to Operation Epic Fury. The Iranian Supreme Council of National Security stabilized at 3,117 deaths. HRANA’s named-verified Crimson Winter report documented 7,007 — a regime-to-verified ratio of approximately 0.45, sitting squarely inside the 0.4-0.6 prior for politically sensitive under-counts. The November 2019 protests produced an even more aggressive under-count of the same kind: regime approximately 230 deaths versus Reuters’ Interior-Ministry-sourced approximately 1,500 — a ratio near 0.15, well below the prior’s lower bound. The 1988 prison massacres case (Khomeini-ordered executions of approximately 2,800 to 5,000 political prisoners across at least 32 cities) is the long-form version of the same pattern: the regime denied the deaths for twenty-eight years until Montazeri’s released audio recording in August 2016 forced partial acknowledgment. Across all three events, the directional bias is consistent; the magnitude varies, with the prior reflecting the typical case rather than the most aggressive one.

The Operation Rising Lion / Twelve-Day War of June 2025 provides the cleanest over-count calibration — and the closest analog to the Operation Epic Fury context this article is testing. Iranian senior officials publicly framed the campaign as overwhelmingly civilian-targeted, 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, not the 90 percent civilian share the regime had claimed. The over-count factor sits inside the 1.5-3.0 prior. IRIB head Peyman Jebelli openly admitted the parallel propaganda fabrications at Shahid Beheshti University in December 2025: state media had aired claims of two to four downed Israeli F-35s and circulated photographs of an alleged captured female Israeli pilot, all of which turned out to be fabricated (the photographs were recycled from a 2024 US military crash combined with AI-generated imagery and video-game screenshots). The IDF confirmed the reality across the entire Twelve-Day War: zero manned Israeli aircraft shot down. Both sides of the regime’s wartime information operation — the under-count of harm caused by the regime, the over-count of harm attributable to the enemy — operated simultaneously and to the same political end.

The Bayesian discount applied to regime casualty reporting: approximately 0.4 to 0.6 for politically sensitive under-counts (deaths the regime caused), and approximately 1.5 to 3.0 for politically advantageous over-counts (deaths attributable to enemies). Applying the inflation factor to the regime-claimed total Iranian civilian deaths from Operation Epic Fury (the Iranian Health Ministry’s 3,375 or the Amnesty-cited 2,362) implies a true civilian-deaths posterior of approximately 800 to 1,500. Hengaw’s independently derived 1,030 sits inside that range. The convergence is what the methodology produces when both pools are anchored to the same kind of named-victim verification.

The Floor IS the Floor

The verification methodology produces a documented number that is conservatively scoped. The number is a floor, not a ceiling. True Iranian KIA across both phases probably runs five to fifteen percent above Hengaw’s documented military total of 6,620 (regime concealment of military deaths). True civilian deaths probably run ten to twenty-five percent above Hengaw’s 1,030 (peripheral-province under-counting under blackout). Both effects partially cancel for the ratio, leaving the 6.4:1 figure more robust than the absolute totals.

The defensible reading is that the documented number is the verifiable floor at the time of publication, and that the floor is approximately what the methodology can sustain under the documented conditions of the campaign. It is not a ceiling. It is not a precise measurement. It is what named-victim verification produces when applied honestly to the documentary anchors that exist.

The next number Hengaw publishes will land in a war that is not yet over. The methodology will be the same. The discipline that produced the floor in April will produce whatever number the methodology can sustain in the months that follow. Updating to the new number when it lands is what the methodology requires. The floor was the floor for what was known at the time. New information shifts the floor.

This article is a deep dive sub-article of Six to One. For the historical comparator class against which the documented ratio is measured, see The Historical Band. For the legal framework on defender co-location and the expected-ratio counterfactual, see Article 58 and the Co-Location Tax. For the four-decade pattern of regime casualty inflation, see The Credibility Ledger. For the apex case verified by this methodology, see The Minab Test. For the broader information environment, see The Digital Siege.

Footnotes

  1. Hengaw Day-40 aggregate, published April 8, 2026 (ceasefire date): 7,650 deaths total; 6,620 military or government personnel; 1,030 civilians; 189 minors and 215 women among civilians. Documentation across 196 cities in 27 of 31 provinces. Provincial highest civilian counts: Tehran, Hormozgan (driven heavily by Minab), Lorestan, Alborz, Fars, East Azerbaijan. Source: Hengaw Organization for Human Rights tenth serial report on Operation Epic Fury. ↩

  2. Internet blackout: NetBlocks measured Iran’s connectivity at 1-4 percent of pre-crisis levels for more than 59 consecutive days during Operation Epic Fury (February 28 to April 8, 2026); Cloudflare measured 98 percent drop on opening day; longest state-imposed shutdown ever recorded by NetBlocks (surpassing Myanmar Rohingya-period shutdown, Tigray war shutdown, and Iran’s January 2026 shutdown that preceded the war); cumulative cost ≥$1.8-1.9 billion to Iranian economy in first 48 days; cumulative 2026 connectivity loss approximately 33 percent of the year. Iranian authorities arrested >1,800 people for internet-related offenses, expelled foreign journalists, restricted Red Crescent reporting, conducted door-to-door seizures of Starlink terminals and satellite dishes. Sources: NetBlocks, Cloudflare Radar, NPR April 27 reporting. ↩

  3. Hengaw methodology: Norway-based Kurdish human rights organization with field documentation infrastructure across Iran’s Kurdish provinces (Kermanshah, Ilam, West Azerbaijan, Kurdistan) and Iraq’s Kurdistan Region (Erbil, Sulaymaniyah, Duhok); Iraq-based field network operates outside Iranian connectivity; named-victim documentation paired with image, certificate, family identifier, or eyewitness anchor; cross-corroboration with HRANA, Airwars, Amnesty, and UN FFMI databases; explicit exclusion of unverified group deaths from confirmed totals; serial reports with consistent definitions across the campaign. Refusals: no social-media-based counts, no AI-generated or recycled image inclusion, no extrapolation from verified to unverified totals, no revision under regime pressure. ↩

  4. HRANA cumulative count through April 7, 2026: 3,636 documented strike deaths; 1,221 military, 1,701 civilians, 714 unclassified. Methodological caveat verbatim from HRANA report: “It is believed that military casualties are significantly higher than the figures reported in [its] reports… confirmations depend largely on government data that are obscured… [its] count largely reflects reports of senior officers or military personnel who were present in urban areas.” HRANA verification through Pasdaran Documentation Project, HRANA news agency, Spreading Justice, and Statistics Department. The apparent ~47 percent civilian share is a methodological artifact of the missing military denominator, not a measurement of the campaign’s true ratio. ↩

  5. Twelve-Day War (June 13-24, 2025) cross-reference cumulative counts. HRANA documented 1,190 deaths by June 28, 2025: 435 military, 436 civilians, 319 unidentified. Hengaw 12-Day War final report (post-July 2025): 1,082 killed inside Iran — 182 civilians (68 women, 45 children), 900 military. Same Israel-Iran conflict, same blackout-constrained verification environment, same two named-victim documentation methodologies producing different classification ratios. The same classification-rule differences identified in the 2026 war methodology comparison explain the 2025 divergence: Hengaw’s broader military classification (Basij at checkpoints, provincial-base IRGC, FARAJA, MOIS, judiciary security personnel, foreign militias) vs HRANA’s three-bucket structure (civilian / military / unclassified pending verification). HRDAG’s Multiple Systems Estimation framework (Patrick Ball, Megan Price; applied at OHCHR for Syria casualty reconciliation in the 2022 estimate of 306,887 directly killed civilians) treats database divergence as the expected outcome of independent named-victim methodologies sampling overlapping populations; identical totals across organizations would be statistically suspicious. Sources: HRANA daily reports, June-July 2025; Hengaw 12-Day War final report, July 2025; HRDAG technical reports. ↩

  6. Contamination forensics. New York Times: ≥110 distinct AI-generated images/videos in first two weeks of campaign. NewsGuard: ≥18 false war-related Iranian state-media claims. Tehran Times: AI-manipulated “destroyed US radar” image (actually Bahrain). AFP-debunked images of “Tel Aviv vehicle fires” that originated in January 2026 Tehran (Kaj Square) anti-government protests. Reuters debunking “new” Iranian strike footage on Tel Aviv as June 2025 video. Recycled video-game footage (Arma 3 SU-57 simulations) viewed >7 million times claiming to show Iranian strikes. Fake “Radio Gilan” Telegram channel falsely claimed to be IRGC announcing accidental responsibility for Minab. Citizen Lab “PRISONBREAK” influence-operation infrastructure pre-staging deepfakes for synchronized release with kinetic events. Gemini falsely captioned mass-burial bird’s-eye image as Kahramanmaraş 2023; Grok variously identified the same Minab burial as Jakarta COVID 2021. Estimated contamination rate of unverified open-source civilian-casualty pool: 15-25 percent based on debunked-claim ratios. ↩

  7. Methodology firewall. HRANA Crimson Winter report (February 23, 2026): 7,007 named-verified deaths from January 2026 crackdown, broken down as 6,488 adult protesters, 236 minors, 76 non-participants, and 207 security force members. Civilian subset (excluding security forces): 6,800. HRANA war-casualty count: separate intake, separate cross-referencing, separate publication cycle. Hengaw firewall maintained through same separation. Hengaw documented 14 secret prison executions of political prisoners between March 17 and April 6, 2026, including teenage wrestler Saleh Mohammadi and 18-year-old Amir Hossein Hatami; these are in regime-attributable pool, not war-civilian pool. The 1,030 coalition-attributable civilian count and the 6,800 regime-attributable civilian subset of Crimson Winter do not overlap. Image-based contamination (e.g., Kahrizak morgue imagery from January re-circulated as Phase-2 strike imagery) documented by AFP but does not enter named-victim totals because images alone are not sufficient anchors. ↩

  8. Regime casualty inflation pattern. Under-count anchors: January 2026 (regime SCNS 3,117 vs. HRANA Crimson Winter 7,007 named verified, ratio ~0.45 — inside prior); November 2019 protests (regime claim ~230 vs. Reuters interior-ministry sources ~1,500, ratio ~0.15 — more aggressive than prior); 1988 prison massacres (Khomeini fatwa, Death Commissions across 32+ cities, 28-year denial; verification through Montazeri audio August 2016 — Amnesty Blood-Soaked Secrets 2018). Over-count anchors: Operation Rising Lion / Twelve-Day War June 2025 (regime ~90% civilian claim vs. HRANA breakdown 436 civilian / 435 military / 319 unidentified of 1,190 documented — UNDP / UN Security Council briefing reporting and “Casualties of the Iran-Israel War” tracking); IRIB head Peyman Jebelli admission at Shahid Beheshti University December 2025 (state media F-35 kill claims and captured-pilot photos all fabricated; IDF confirmed zero manned aircraft losses across Twelve-Day War). Bayesian discount factor 0.4-0.6 for regime under-counts (typical case; specific events can be more aggressive); inflation factor 1.5-3.0 for regime over-counts. Application to Iranian Ministry of Health 3,375 or Amnesty-cited 2,362 implies true civilian-deaths posterior of approximately 800-1,500; Hengaw’s independently derived 1,030 sits inside that range. Full inventory of regime casualty events including PS752 (denial-of-attribution) and Mahsa Amini (denial-of-cause): The Credibility Ledger. ↩