{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,10]],"date-time":"2026-10-10T04:49:23Z","timestamp":1791607763032,"version":"4.3.4"},"update-to":[{"DOI":"10.12688\/f1000research.169794.1","type":"new_version","label":"New version","source":"publisher","updated":{"date-parts":[[2026,10,9]],"date-time":"2026-10-09T00:00:00Z","timestamp":1791504000000}},{"DOI":"10.12688\/f1000research.169794.2","type":"new_version","label":"New version","source":"publisher","updated":{"date-parts":[[2026,10,9]],"date-time":"2026-10-09T00:00:00Z","timestamp":1791504000000}},{"DOI":"10.12688\/f1000research.169794.3","type":"new_version","label":"New version","source":"publisher","updated":{"date-parts":[[2026,10,9]],"date-time":"2026-10-09T00:00:00Z","timestamp":1791504000000}}],"reference-count":35,"publisher":"F1000 Research Ltd","license":[{"start":{"date-parts":[[2026,10,9]],"date-time":"2026-10-09T00:00:00Z","timestamp":1791504000000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["f1000research.com"],"crossmark-restriction":false},"short-container-title":["F1000Res"],"abstract":"<ns3:p>Background The adoption of AI for military targeting, defence and engagement operations changes the tempo, magnitude and scope of lethal decision making. Both Iran and Israel have introduced systems that incorporate AI in different ways, technologically and doctrinally, in the Iranian-Israeli confrontation. The issue that is being analysed here is not about which state is technologically more advanced but if meaningful human control of the application of force can be maintained in a high volume, under compressed decision time algorithmic system. Method The individual AI-enabled system or application is the primary unit of analysis, whereas Israel and Iran are higher-level comparative contexts used for interpreting the findings at the system level. Systems are first broken down into five functional categories (AI-supported target generation, automated defence interception, assisted fire control at the weapon, autonomous platforms and swarms, and cyber and information applications) based on decision tempo, the location of the human in the decision cycle, and the legal test that must be used. A six indicator assessment is conducted for each system across four dimensions of the system: technical capability, operational employment and effectiveness, human control, and legal compliance. Five indicators are measured as an ordinal capability band with a within band rubric, while legal compliance is not scored but is measured qualitatively. There is an explicit level of confidence attached to each assessment, differentiating between documented and assessed competence. Sources were identified and sourced from peer-reviewed literature, OSINT, official documents and investigative journalism and then prioritized on a five-tiered scale. Results The two national sets don't diverge quite as evenly as the aggregate comparison implies within the categories that make up functional sets. The undetermined Iranian capability in target generation is in contrast with the documented high level of Israeli capability. In defensive interception, high automation comes from the physics of the threat. With assisted fire control, AI will be added to the fire control system, but it will only reinforce human control \u2013 its authority will be purely negative. For both autonomous platforms and swarms, the capability sets are closest and the control problem is most acute for them. In cyber and information applications the evidentiary asymmetry inverts. Technical accuracy, operational effectiveness, legal compliance and humanitarian outcome are treated as separate constructs throughout. Conclusion However, the limitation on meaningful human control is not the sophistication of the algorithm, but the ratio of the number of engagement options that can be provided by a system in a given time to the time taken to interrogate them. Two paths of degraded human control are at work, one from advanced military forces, one from resource-poor ones, both moving toward the same end point. Governance should thus directly control the tempo of decisions, the speed at which decisions pass through, and the review of decisions, and not leave autonomy as an on\/off switch on the weapon. A set of 17 specific assignable mechanisms are suggested.<\/ns3:p>","DOI":"10.12688\/f1000research.169794.4","type":"journal-article","created":{"date-parts":[[2026,10,9]],"date-time":"2026-10-09T09:34:09Z","timestamp":1791538449000},"page":"1200","update-policy":"https:\/\/doi.org\/10.12688\/f1000research.crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Algorithmic Targeting in the Iranian\u2013Israeli Confrontation: Technical Realities, Legal thresholds, and the Boundaries of Human Control"],"prefix":"10.12688","volume":"14","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-8559-0999","authenticated-orcid":false,"given":"Lamia","family":"faris","sequence":"first","affiliation":[{"name":"College of Dentistry, Basic Science Ramadi Al Anbar Governorate University of Anbar"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"2560","published-online":{"date-parts":[[2026,10,9]]},"reference":[{"key":"ref1","article-title":"\u2018Lavender\u2019: The AI machine directing Israel\u2019s bombing spree in Gaza. +972 Magazine, in partnership with Local Call.","author":"Y Abraham","year":"2024"},{"key":"ref2","article-title":"What is Iron Dome and how does Israel\u2019s missile defence system work?","author":"A Jazeera","year":"n.d.","journal-title":"Al Jazeera News."},{"key":"ref3","doi-asserted-by":"publisher","first-page":"187-194","DOI":"10.1007\/s43154-020-00024-3","article-title":"Autonomous Weapons Systems and Meaningful Human Control: Ethical and Legal Issues.","volume":"1","author":"D Amoroso","year":"2020","journal-title":"Curr. Robot. Rep."},{"key":"ref4","doi-asserted-by":"publisher","first-page":"336-370","DOI":"10.1163\/18781527-bja10119","article-title":"The Use of the \u2018Lavender\u2019 in Gaza and the Law of Targeting: AI-Decision Support Systems and Facial Recognition Technology.","volume":"16","author":"E Andersin","year":"2025","journal-title":"J. Int. Humanit. Leg. Stud."},{"key":"ref5","article-title":"A Defence for Guardian Robots: Are Defensive Autonomous Weapons Systems Justifiable?","author":"R Anugrah","year":"2024","journal-title":"Harv Int\u2019l L. J."},{"key":"ref6","doi-asserted-by":"publisher","first-page":"687-709","DOI":"10.1017\/S1816383112000768","article-title":"On banning autonomous weapon systems: human rights, automation, and the dehumanization of lethal decision-making.","volume":"94","author":"P Asaro","year":"2012","journal-title":"Int. Rev. Red Cross."},{"key":"ref7","doi-asserted-by":"crossref","first-page":"393-413","DOI":"10.1017\/S0260210517000614","article-title":"Autonomous weapons systems and changing norms in international relations.","volume":"44","author":"I Bode","year":"2018","journal-title":"Rev. Int. Stud."},{"key":"ref9","first-page":"1347-1402","article-title":"War Torts: Accountability for Autonomous Weapons.","volume":"164","author":"R Crootof","year":"2016","journal-title":"Univ. Pa. Law Rev."},{"key":"ref11","doi-asserted-by":"crossref","first-page":"343-348","DOI":"10.1111\/1758-5899.12665","article-title":"Moving beyond semantics on autonomous weapons: Meaningful human control in operation.","volume":"10","author":"M Ekelhof","year":"2019","journal-title":"Glob. Policy."},{"key":"ref12","doi-asserted-by":"publisher","first-page":"236-256","DOI":"10.37458\/ssj.6.1.12","article-title":"AI And Military Precision In Drones Attack On The Israeli-Hamas Conflict.","volume":"6","author":"K Emegha","year":"2025","journal-title":"Secur. Sci. J."},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.5281\/zenodo.17385265","author":"L Faris","year":"2025","journal-title":"Supplementary Data for \u201cAlgorithmic Targeting in Iranian-Israeli Confrontation: Technical Realities, Legal Thresholds, and the Boundaries of Human Control\u201d (Version 2.0). [Data set]."},{"key":"ref14","doi-asserted-by":"publisher","first-page":"141-189","DOI":"10.1162\/isec_a_00337","article-title":"Why China Has Not Caught Up Yet: Military-Technological Superiority and the Limits of Imitation, Reverse Engineering, and Cyber Espionage.","volume":"43","author":"A Gilli","year":"2019","journal-title":"Int. Secur."},{"key":"ref15","article-title":"Israel\u2019s artificial intelligence system that decides who to bomb in Gaza.","author":"G Pascual","year":"2024","journal-title":"El Pa\u00eds."},{"key":"ref16","article-title":"\u201cThe machine did it coldly\u201d: Israel used AI to identify 37, 000 Hamas targets.","year":"2024"},{"key":"ref17","doi-asserted-by":"publisher","first-page":"642-650","DOI":"10.31703\/grr.2020(v-i).67","article-title":"Artificial Intelligence as the New Art of War: An Appraisal.","volume":"V","author":"S Gul","year":"2020","journal-title":"Global Regional Review."},{"key":"ref19","doi-asserted-by":"publisher","first-page":"764-788","DOI":"10.1080\/01402390.2019.1621174","article-title":"When speed kills: Lethal autonomous weapon systems, deterrence and stability.","volume":"42","author":"M Horowitz","year":"2019","journal-title":"J. Strateg. Stud."},{"key":"ref24","first-page":"2-5","article-title":"Navigating Cybersecurity and Surveillance: Iran\u2019s Dual Strategy for National Security.","author":"N Khorrami","year":"2024","journal-title":"Policy."},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1093\/jogss\/ogae009","article-title":"Digital Targeting: Artificial Intelligence, Data, and Military Intelligence.","volume":"9","author":"A King","year":"2024","journal-title":"J. Glob. Secur. Stud."},{"key":"ref26","article-title":"Iran\u2019s Bet on Autonomous Weapons.","author":"E Lisman","year":"2021","journal-title":"War on the Rocks"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1136\/bmj.q203","article-title":"Gaza-Israel conflict: Hundreds of medics are killed or arrested after intense attacks on healthcare facilities.","author":"E Mahase","year":"2025","journal-title":"BMJ."},{"key":"ref28","doi-asserted-by":"publisher","first-page":"118-175","DOI":"10.4337\/9781802207675.00010","article-title":"Autonomous weapons systems under international humanitarian law.","volume":"2022","author":"D Mauri","year":"2016","journal-title":"Autonomous Weapons Systems and the Protection of the Human Person."},{"key":"ref29","article-title":"Exploring Artificial Intelligence-Enhanced Cyber and Information Operations Integration.","author":"R McGuire","year":"2025","journal-title":"Mil. Rev."},{"key":"ref31","article-title":"Israeli AI Technology Used to Identify 37, 000 Hamas Targets.","author":"V Patel","year":"2024","journal-title":"Int Bus Times"},{"key":"ref32","article-title":"How Artificial Intelligence is Improving Iron Dome.","author":"G Ramachandra","year":"2023","journal-title":"Medium."},{"key":"ref33","article-title":"Israel\u2019s military technology and intelligence capabilities.","year":"n.d.","journal-title":"RAND Corp."},{"key":"ref34","first-page":"1-6","article-title":"Meaningful Human Control.","volume":"36","author":"H Roff","year":"2015","journal-title":"Artif Intell Auton Weapons"},{"key":"ref35","article-title":"United Nations Diplomatic Conference of Plenipotentiaries"},{"key":"ref36","doi-asserted-by":"publisher","first-page":"15","DOI":"10.3389\/frobt.2018.00015","article-title":"Meaningful Human Control over Autonomous Systems: A Philosophical Account.","volume":"5","author":"F Santoni de Sio","year":"2018","journal-title":"Front. Robot. AI."},{"key":"ref38","first-page":"231-281","article-title":"'Out of the Loop': Autonomous Weapon Systems and the Law of Armed Conflict.","volume":"4","author":"M Schmitt","year":"2013","journal-title":"Harv. Natl. Secur. J."},{"key":"ref39","doi-asserted-by":"publisher","first-page":"76-92","DOI":"10.71085\/sss.04.01.204","article-title":"The role of AI (Artificial Intelligence) in military applications in the Middle East: How Iran-Israel conflict affects Turkey\u2019s national security and regional stability.","volume":"4","author":"M Shabbir","year":"2024","journal-title":"Soc. Sci. Spectr."},{"key":"ref40","doi-asserted-by":"crossref","first-page":"787-799","DOI":"10.1017\/S1816383112000732","article-title":"The evitability of autonomous robot warfare.","volume":"94","author":"N Sharkey","year":"2012","journal-title":"Int. Rev. Red Cross."},{"key":"ref41","article-title":"SMARTSHOOTER to present its SMASH family of fire control systems and demonstrate its effectiveness in air defence.","year":"n.d."},{"key":"ref42","doi-asserted-by":"publisher","first-page":"62-77","DOI":"10.1111\/j.1468-5930.2007.00346.x","article-title":"Killer Robots.","volume":"24","author":"R Sparrow","year":"2007","journal-title":"J. Appl. Philos."},{"key":"ref43","doi-asserted-by":"publisher","first-page":"175-187","DOI":"10.1080\/21624887.2020.1760587","article-title":"Algorithmic warfare and the reinvention of accuracy.","volume":"8","author":"L Suchman","year":"2020","journal-title":"Crit. Stud. Secur."},{"key":"ref44","article-title":"Resolution 78\/241, Lethal autonomous weapons systems. A\/RES\/78\/241","year":"2023"}],"container-title":["F1000Research"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/f1000research.com\/articles\/14-1200\/v4\/xml","content-type":"application\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/f1000research.com\/articles\/14-1200\/v4\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/f1000research.com\/articles\/14-1200\/v4\/iparadigms","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,10,10]],"date-time":"2026-10-10T04:14:08Z","timestamp":1791605648000},"score":1,"resource":{"primary":{"URL":"https:\/\/f1000research.com\/articles\/14-1200\/v4"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,10,9]]},"references-count":35,"URL":"https:\/\/doi.org\/10.12688\/f1000research.169794.4","relation":{"has-version":[{"id-type":"doi","id":"10.12688\/f1000research.169794.3","asserted-by":"subject"},{"id-type":"doi","id":"10.12688\/f1000research.169794.2","asserted-by":"subject"},{"id-type":"doi","id":"10.12688\/f1000research.169794.1","asserted-by":"subject"}],"has-review":[{"id-type":"doi","id":"10.5256\/f1000research.211354.r528512","asserted-by":"subject"},{"id-type":"doi","id":"10.5256\/f1000research.211354.r528512","asserted-by":"object"}]},"ISSN":["2046-1402"],"issn-type":[{"value":"2046-1402","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,10,9]]},"assertion":[{"value":"Indexed","URL":"https:\/\/f1000research.com\/articles\/14-1200\/v4#article-reports","order":0,"name":"referee-status","label":"Referee status","group":{"name":"current-referee-status","label":"Current Referee Status"}},{"value":"10.5256\/f1000research.187170.r436966, Elshan Hashimov, National Defense University, Baku, 26 Dec 2025, version 1, 1 approved with reservations","URL":"https:\/\/f1000research.com\/articles\/14-1200\/v1#referee-response-436966","order":0,"name":"referee-response-436966","label":"Referee Report","group":{"name":"article-reports","label":"Article Reports"}},{"value":"<b>Lamia faris<\/b>; \n<i>Posted: 21 Feb 2026<\/i>; We sincerely thank the reviewers for their thoughtful comments and constructive feedback, which have helped us improve the quality of our manuscript. We carefully considered each point raised and made the following revisions","URL":"https:\/\/f1000research.com\/articles\/14-1200\/v4#referee-comment-15507","order":1,"name":"referee-comment-15507","label":"Referee Comment","group":{"name":"article-reports","label":"Article Reports"}},{"value":"10.5256\/f1000research.197018.r486079, Adebayo Yusuf Balogun, University of Tampa, Tampa, FL, 22 May 2026, version 2, 2 approved with reservations","URL":"https:\/\/f1000research.com\/articles\/14-1200\/v2#referee-response-486079","order":2,"name":"referee-response-486079","label":"Referee Report","group":{"name":"article-reports","label":"Article Reports"}},{"value":"<b>Lamia faris<\/b>; \n<i>Posted: 23 May 2026<\/i>; Dear Reviewer,  Thank you for your thorough, constructive, and highly detailed evaluation of our manuscript. Your comments have been instrumental in strengthening the methodological transparency, evidentiary precision, and analytical balance of the paper. We are grateful for both the recognition of the manuscript\u2019s interdisciplinary scope and comparative value, and for the specific guidance on areas requiring improvement. Below, we provide a detailed point-by-point response to each of your comments, organized according to the six evaluation questions in your report. For each concern, we indicate the specific revisions made. Reviewer comments are shown in \n<b><i>bold italics<\/i><\/b>, and author responses follow. References to manuscript sections correspond to the revised version.  1. Clarity of Presentation and Current Literature 1.1 Differentiating verified from alleged claims  \n<b>Reviewer: <i>&quot;Several claims are presented too definitively despite relying on investigative journalism, anonymous sources, OSINT material, or unverified operational reporting. Distinctions between confirmed evidence, allegations, analytical interpretation, and speculation are insufficiently clear.&quot;<\/i><\/b> \n<b>Response: <\/b>We fully agree. The following revisions have been made: \n<b>Hedging language: <\/b>A comprehensive language audit has been performed throughout the manuscript. Claims based on investigative journalism or anonymous sources now use cautious language such as \u201caccording to investigative reporting,\u201d \u201caccording to reports,\u201d \u201cbasing their claim on anonymous Israeli intelligence officers,\u201d and \u201caccording to humanitarian affairs reports and local health authorities.\u201d These phrases consistently signal the provenance and reliability level of each claim. \n<b>Source reliability classification: <\/b>A formal four-tier source reliability classification system (T1\u2013T4) has been developed and applied throughout the manuscript. Section 3.4 now explicitly states that this hierarchy serves as a source reliability classification, with claims supported by T1\/T2 sources presented with greater certainty and those based on T3\/T4 sources accompanied by hedging language. The full classification system is documented in the new Supplementary Scoring Justification Table (Sheet 2: Source Reliability), deposited on Zenodo (DOI: 10.5281\/zenodo.17385265). \n<b>Performance metrics: <\/b>The Lavender system\u2019s 90% accuracy claim is now explicitly attributed to \u201cUnit 8200 internal validation studies\u201d (self-reported military figures), and the text states that \u201cindependent validation of the system\u2019s performance remains unavailable due to operational secrecy.\u201d The 3,700 misidentification figure is explicitly flagged as \u201ca hypothetical misidentification figure\u2014an extrapolation that must be treated with caution given that the underlying accuracy claim itself remains unverified through independent assessment.\u201d 1.2 Separating peer-reviewed from media evidence \n<b>Reviewer: <i>&quot;Separate peer-reviewed academic evidence from media investigations and commentary.&quot;<\/i><\/b> \n<b>Response: <\/b>The expanded methodology (Section 3.2) now explicitly categorizes sources into five types: academic databases (Web of Science, Scopus, JSTOR, Google Scholar), OSINT platforms, government\/military documents, investigative journalism outlets, and legal\/humanitarian sources. The source priority hierarchy in Section 3.4 ranks these from (a) independently confirmed technical documentation to (e) anonymous source reporting. In the Supplementary Scoring Justification Table, every score is accompanied by the specific sources used and their tier classification, enabling readers to assess the evidentiary weight of each finding.  1.3 Scholarly engagement \n<b>Reviewer: <i>&quot;Strengthen engagement with scholarly literature on AI accountability, explainability, and military human-machine interaction.&quot;<\/i><\/b> \n<b>Response: <\/b>The revised manuscript engages more deeply with scholarly literature in this area, including expanded discussion of Scharre (2018) on human control spectrums and the human-in-the-loop\/out-of-the-loop taxonomy, Roff and Moyes (2015) on meaningful human control aligned with ICRC and Article 36 frameworks, Ekelhof (2019) on the relationship between AI-speed decision-making and responsibility, Sharkey (2012) on the responsibility gap in lethal autonomous weapons, and Mauri (2022) on autonomous weapons under international humanitarian law. A new Key Terminology Definitions section aligns core concepts with established scholarly definitions, including MHC consistent with ICRC Guidelines No. 36.  2. Study Design and Technical Soundness 2.1 Scoring framework transparency (Must Be Addressed) \n<b>Reviewer: <i>&quot;Explain in detail how the scoring process was conducted. Clarify whether scoring was performed by a single evaluator or multiple reviewers. Describe how uncertainty and conflicting evidence were handled during scoring. Include explicit justification for each major comparative score.&quot;<\/i><\/b> \n<b>Response: <\/b>This was among the most critical observations, and we have addressed each point comprehensively: \n<b>Scoring criteria: <\/b>A new Table A (\u201cScoring criteria for six-dimension framework\u201d) has been added to Section 3.3, providing explicit definitions for Low (0\u201330), Moderate (31\u201360), and High (61\u2013100) capability levels across all six dimensions. Each cell contains specific criteria that must be met for a given score range. \n<b>Single evaluator: <\/b>The manuscript now explicitly acknowledges that all evaluations were performed by a single researcher, which constitutes a methodological limitation. The text states: \u201cIn the absence of multiple independent raters, subjective bias in scoring cannot be fully excluded. To mitigate this, the author applied the standardized rubric in Table A consistently across all systems and relied on the source priority hierarchy in Section 3.4 to ensure evidence-based scoring.\u201d \n<b>Uncertainty and conflict handling: <\/b>Section 3.4 (Priority Level) describes the five-level hierarchical reconciliation process used when sources conflicted, ranging from (a) independently confirmed technical documentation to (e) anonymous source reporting. Conflict notation procedures require that serious disagreements are stated explicitly. A minimum estimation rule applies when quantitative estimates are contradictory. \n<b>Per-score justification: <\/b>A comprehensive Supplementary Scoring Justification Table has been deposited on Zenodo (DOI: 10.5281\/zenodo.17385265, Version 2.0). This workbook contains 48 individual score justifications (8 systems &times; 6 dimensions), each specifying: the score assigned, capability level, primary sources used, source tier(s), written justification, and a confidence flag (High\/Medium\/Low). This directly provides the \u201cexplicit justification for each major comparative score\u201d the reviewer requested.  2.2 Numerical scoring limitations (Additional Suggestion) \n<b>Reviewer: <i>&quot;Consider reducing reliance on numerical scoring where objective measurement is not possible. Acknowledge the limitations of assessing classified or partially documented military systems.&quot;<\/i><\/b> \n<b>Response: <\/b>We have addressed both points: \n<b>Analytical disclaimer: <\/b>Section 3.3 now includes an explicit statement: \u201cThese scores represent qualitative analytical assessments by the author based on the documented evidence hierarchy, rather than statistically validated measurements. The scoring framework is intended to facilitate structured comparison and visual representation of relative capabilities, not to convey measurement precision.\u201d This language is also reflected in figure captions. \n<b>Classified systems: <\/b>A new paragraph at the end of Section 4.3 acknowledges: \u201cMilitary AI systems are inherently classified, and publicly available information represents only a fraction of actual technical capabilities and operational practices. The assessments presented in this study are necessarily constrained by what has been disclosed, leaked, or independently documented.\u201d A new dedicated Limitations subsection (Section 6.4) further elaborates on data access, methodological constraints, evidence asymmetry, and scope limitations.  3. Methodological Replication 3.1 Methodological appendix (Must Be Addressed) \n<b>Reviewer: <i>&quot;Provide a methodological appendix or supplementary table showing: evidence sources used for each system, score justification for each analytical dimension, reliability assessment of each source.&quot;<\/i><\/b> \n<b>Response: <\/b>This has been fully addressed through the new Supplementary Scoring Justification Table deposited on Zenodo (DOI: 10.5281\/zenodo.17385265, Version 2.0). The workbook contains five sheets: (1) Scoring Justification \u2014 48 rows with system, dimension, score, sources, source tiers, written justification, and confidence flag for every assessment; (2) Source Reliability Classification \u2014 the T1\u2013T4 tiering system with examples and required hedging language; (3) Scoring Rubric \u2014 full Low\/Moderate\/High definitions for all six dimensions; (4) IHL Compliance Assessment \u2014 five IHL principles scored for both Israeli and Iranian systems with article references and justifications; (5) Metadata \u2014 DOI, license, versioning, and analytical disclaimer. This directly provides all three elements the reviewer requested: evidence sources per system, score justification per dimension, and reliability assessment per source.  3.2 Coding framework and contradictory sources \n<b>Reviewer: <i>&quot;Clarify whether any coding framework or standardized rubric was used. Explain how contradictory sources were reconciled in practice.&quot;<\/i><\/b> \n<b>Response: <\/b>Table A in the manuscript now serves as the standardized rubric, defining explicit criteria for Low (0\u201330), Moderate (31\u201360), and High (61\u2013100) across each dimension. Section 3.4 describes the hierarchical reconciliation process applied when sources conflicted, including the five-level priority hierarchy, conflict notation procedures requiring explicit statement of disagreements, and the minimum estimation rule for contradictory quantitative data. New subsections 3.2.2 and 3.2.3 specify four inclusion criteria (technical detail, evidentiary support, source credibility, cross-verification) and four exclusion criteria (unverified information, speculative capabilities, non-attributable sources, internal contradictions).  4. Statistical Analysis \n<b>Reviewer: <i>&quot;The manuscript should explicitly state that the numerical comparisons are analytical assessments rather than statistically validated measurements.&quot;<\/i><\/b> \n<b>Response: <\/b>Agreed and implemented. Section 3.3 now contains an explicit disclaimer stating that scores represent qualitative analytical assessments, not statistical measurements. Figure captions have been updated to state the analytical basis of the visualizations. For example, Figure 3 states: \u201cSource: Author\u2019s comparative analysis based on documented system specifications\u201d; Figure 4 states: \u201cSource: Author\u2019s compliance analysis based on IHL framework evaluation.\u201d The Supplementary Scoring Justification Table metadata sheet reiterates this disclaimer.  5. Source Data and Reproducibility 5.1 Verifiable vs. contested evidence \n<b>Reviewer: <i>&quot;Clearly identify which findings are based on independently verifiable evidence and which rely on contested reporting. Include a source reliability classification system.&quot;<\/i><\/b> \n<b>Response: <\/b>The five-level priority hierarchy in Section 3.4 now explicitly serves as a source reliability classification system, and the text states that \u201cclaims supported by Level (a) or (b) sources are presented with greater certainty, while those based primarily on Level (d) or (e) sources are accompanied by appropriate hedging language.\u201d The Supplementary Scoring Justification Table assigns source tier(s) and a confidence flag (High\/Medium\/Low) to every individual score, enabling readers to assess precisely which findings rest on independently verifiable evidence (T1) and which rely on contested or single-source reporting (T3\u2013T4). Of the 48 scores in the table, 6 are rated High confidence, 22 Medium, and 20 Low, reflecting the inherent evidential constraints of military AI research.  5.2 Limits of reproducibility \n<b>Reviewer: <i>&quot;Clarify the limits of reproducibility given the nature of military and intelligence-related research.&quot;<\/i><\/b> \n<b>Response: <\/b>Section 3.2 explicitly discusses this: \u201cInformation on conflict areas is both extremely biased and imprecise due to restricted access to the field, the prevalence of propaganda, and the absence of comprehensive ground-truth verification.\u201d The new dedicated Limitations subsection (Section 6.4) further addresses this with specific discussion of: data access and verification limitations (classified systems, propaganda, no ground-truth); methodological constraints (secondary data only, single evaluator, no primary interviews or technical access); evidence asymmetry between Israeli and Iranian systems; and scope limitations (two states, one region, rapidly evolving technology). The data availability section acknowledges this is a secondary data analysis with no primary data collection.  6. Conclusions and Causal Claims 6.1 Moderating causal claims (Must Be Addressed) \n<b>Reviewer: <i>&quot;Moderate causal claims regarding civilian harm and AI targeting systems. Distinguish clearly between technological capability, operational use, and political decision-making.&quot;<\/i><\/b> \n<b>Response: <\/b>This was a critical observation that we have addressed thoroughly: \n<b>Multiple hypotheses: <\/b>The humanitarian impact section (Section 6.2) has been revised to present the disconnect between claimed precision and observed outcomes as multiple intersecting factors rather than a single causal claim. The revised text states: \u201cThese include potential limitations in AI targeting accuracy, inadequate human oversight mechanisms, broader operational and command decisions that may prioritize military objectives over civilian protection, and political decision-making processes that determine rules of engagement.\u201d This explicitly distinguishes technological capability from operational use from political decision-making, as the reviewer requested. \n<b>Attribution caveat: <\/b>The revised text includes an explicit methodological caveat: \u201cIt is important to note that civilian casualties in armed conflict result from complex chains of technological, operational, and political decisions, and attributing specific outcomes solely to AI system performance would oversimplify this relationship.\u201d  6.2 Reframing as analytical interpretations (Must Be Addressed) \n<b>Reviewer: <i>&quot;Reframe some conclusions as informed concerns or analytical interpretations rather than established facts.&quot;<\/i><\/b> \n<b>Response: <\/b>The conclusions section (Section 7) has been revised to use explicitly interpretive framing. The paragraph discussing Gaza operations now reads: \u201cThese reports, while subject to the verification limitations inherent in active conflict zones, suggest a concerning pattern that warrants further independent investigation.\u201d It concludes: \u201cThese observations should be understood as informed analytical concerns based on available evidence rather than definitive causal conclusions.\u201d The phrase \u201capparent tension between stated precision objectives and documented civilian harm\u2014whether attributable to technological limitations, operational tempo, command decisions, or their interaction\u201d explicitly presents the finding as an analytical concern with multiple possible explanations. Additional: Dedicated Limitations Subsection  \n<b>Reviewer: <i>&quot;A dedicated limitations subsection would improve the manuscript.&quot; (Implied across multiple evaluation questions)<\/i><\/b> \n<b>Response: <\/b>A new Section 6.4 (\u201cStudy Limitations\u201d) has been added before the Conclusions, consolidating all limitations into a single dedicated subsection. It clearly distinguishes four categories: (1) Data access and verification limitations (classified systems, propaganda, no ground-truth verification, source reliability constraints); (2) Methodological constraints (secondary data only, single evaluator, no inter-rater reliability, no primary interviews or technical access); (3) Asymmetry in evidence (Israeli systems more documented than Iranian, reflecting transparency differences not analytical effort, Iranian assessments framed as \u201cemerging indications\u201d); (4) Scope limitations (two states, one region, rapidly evolving technology). This consolidation addresses concerns raised in evaluation questions 1, 2, 3, 5, and 6.  Additional: Supplementary Data Update \n<b>Response: <\/b>The Zenodo repository (DOI: 10.5281\/zenodo.17385265) has been updated to Version 2.0 with the addition of the Supplementary Scoring Justification Table. This Excel workbook contains five sheets: (1) Scoring Justification with 48 per-system, per-dimension entries including sources, tiers, justifications, and confidence flags; (2) Source Reliability Classification (T1\u2013T4); (3) Full Scoring Rubric; (4) IHL Compliance Assessment; (5) Metadata with analytical disclaimer. The Data Availability section in the manuscript now explicitly maps which figures and tables are derived from this dataset. All supplementary data remain released under CC-BY 4.0.  Summary of Revisions \n<i>The following table summarizes how each \u201cMust Be Addressed\u201d item has been resolved:<\/i> \n<b>1. Evidentiary basis for claims <\/b>\u2192 Language audit + T1\u2013T4 source classification + hedging language throughout (Sections 3.2, 3.4, throughout) \n<b>2. Scoring framework transparency <\/b>\u2192 Table A + single evaluator acknowledgment + analytical disclaimer + Supplementary Table on Zenodo (Section 3.3, Zenodo) \n<b>3. Verified vs. alleged evidence <\/b>\u2192 Five-level priority hierarchy as reliability classification + per-score confidence flags (Section 3.4, Supplementary Table) \n<b>4. Causal conclusions moderated <\/b>\u2192 Multiple hypotheses framing + tech\/operational\/political distinction + \u201cinformed analytical concerns\u201d language (Sections 6.2, 7) \n<b>5. Methodological replicability <\/b>\u2192 Inclusion\/exclusion criteria + scoring rubric + supplementary justification table + dedicated Limitations subsection (Sections 3.2.2, 3.2.3, 3.3, 6.4, Zenodo)   We sincerely thank the reviewer for the exceptionally detailed and constructive evaluation. The feedback has materially improved the methodological transparency, evidentiary precision, and analytical balance of the manuscript. We believe the revisions fully address all \u201cMust Be Addressed\u201d points and the majority of additional suggestions, and we welcome any further feedback.  Respectfully, \n<b>Lamia Faris<\/b> College of Dentistry, Basic Science University of Anbar, Iraq den.lamia.faris@uoanbar.edu.iq","URL":"https:\/\/f1000research.com\/articles\/14-1200\/v4#referee-comment-16279","order":3,"name":"referee-comment-16279","label":"Referee Comment","group":{"name":"article-reports","label":"Article Reports"}},{"value":"10.5256\/f1000research.202374.r494413, Ananya Pandey, Christ University, Bangalore, Karnataka, 03 Jul 2026, version 3, 5 approved with reservations","URL":"https:\/\/f1000research.com\/articles\/14-1200\/v3#referee-response-494413","order":4,"name":"referee-response-494413","label":"Referee Report","group":{"name":"article-reports","label":"Article Reports"}},{"value":"10.5256\/f1000research.202374.r507415, Drazen Smiljanic, Dr.Franjo Tu\u0111an Defense and Security University, Zagreb, 08 Aug 2026, version 3, 5 approved with reservations","URL":"https:\/\/f1000research.com\/articles\/14-1200\/v3#referee-response-507415","order":5,"name":"referee-response-507415","label":"Referee Report","group":{"name":"article-reports","label":"Article Reports"}},{"value":"<b>Lamia faris<\/b>; \n<i>Posted: 24 Sep 2026<\/i>; I am grateful to Dr Smiljanic for a review that engages with the analytical architecture rather than the surface of the manuscript. His diagnosis \u2014 that the weakness lay in the architecture rather than in the evidence \u2014 shaped this revision, and his recommended chain (aim \u2192 questions \u2192 unit of analysis \u2192 taxonomy \u2192 dimensions \u2192 evidence \u2192 comparison \u2192 national findings \u2192 conclusions) is now the organising structure of the article. Each point is addressed below. \n<b>1. Research aim.<\/b> Accepted. The aim in &sect;1.1 now states what is compared, across which dimensions, and for what purpose, and says explicitly that the study extends beyond narrowly defined &quot;targeting systems&quot; because the control problem spans all five functions. The stated purpose is no longer to rank two militaries but to isolate the conditions under which meaningful human control is preserved, attenuated or lost. \n<b>2. Unit of analysis.<\/b> Accepted in full, in the form suggested. The individual AI-enabled system or application is now the primary unit of analysis, with Israel and Iran as higher-level comparative contexts within which system-level findings are interpreted. A rule of attribution states that every conclusion is marked with the level at which it is warranted, and that no national-level claim is presented as if derived directly from system-level assessment. \n<b>3. Functional taxonomy.<\/b> Accepted. &sect;3.4 introduces five functional categories \u2014 AI-supported target generation, automated defensive interception, assisted fire control, autonomous platforms and swarms, and cyber and information applications \u2014 classified by decision tempo, the position of the human in the decision cycle, and the principal legal test, which is now a column in Table 1. The reviewer's own examples drove this: Lavender's oversight concerns review and authorisation of a proposed target, whereas Iron Dome's automation is intrinsic to a threat measured in seconds. &sect;4.4.2 now carries an explicit interpretive rule that Iron Dome's Low human-oversight rating is not a governance deficiency comparable to Lavender's, since the two are alike in label and opposite in substance. \n<b>4. Dimensions and indicators.<\/b> Accepted, and implemented exactly as proposed. The six indicators are now grouped under four higher-order dimensions: technical capability (data integration, algorithmic architecture), operational employment and effectiveness (deployment scale, operational accuracy), human control (human oversight), and legal compliance. A &quot;Final analytical model&quot; paragraph states this once, and the same terminology is used in the methods, tables, figures and conclusions. \n<b>5. Comparability.<\/b> Accepted. &sect;3.6 establishes that comparison is made within functional categories first, with national patterns considered only afterwards. Table 6 is read down a column within a category, never across categories, and where direct system-to-system comparison is not possible this is stated rather than implied by a common score. \n<b>6. The 0\u2013100 scoring.<\/b> Accepted in full. The numerical scoring has been withdrawn. Capability is reported in ordinal Low \/ Moderate \/ High bands with a within-band rubric governing placement, and no composite score is computed. This follows the reviewer's conclusion directly: the distinction between 75 and 85 could not be consistently justified from the underlying evidence, so the ordinal categories alone now represent it. The earlier numerical values are retained in the repository, labelled as superseded analytical inputs. \n<b>7. From system level to national level.<\/b> Accepted. &sect;4.5 makes this step explicit and conditional. A pattern is advanced only where it appears across at least two functional categories, is consistent with independently evidenced structural conditions, and is not contradicted by an assessed system. The three resulting propositions are labelled tentative and exploratory, and two statements are marked as \n<i>not supported<\/i>: relative overall capability, and relative legal compliance. &sect;3.8 gives the system-selection rule and the identified-but-unassessed systems, so the reader can judge how far selection shapes the characterisation. A sensitivity note identifies the FC1 asymmetry as the study's principal open evidentiary question. \n<b>8. Evidentiary asymmetry as an analytical issue.<\/b> Accepted, using the reviewer's three-way distinction. The manuscript now separates documented capability, assessed capability, and evidentiary confidence, with a confidence rating attached to every assessment. Where no Iranian system met the inclusion criteria, the result is recorded as \n<i>undetermined<\/i>, defined as evidentiary absence rather than low capability. The point is stated plainly: a score placing Iran low in FC1 would record the absence of journalism, not the absence of a system. \n<b>9. Technical accuracy, operational effectiveness, legal compliance, humanitarian outcome.<\/b> Accepted and applied consistently. These are treated as four distinct constructs throughout. The earlier extrapolation from a reported error rate to a number of misidentified people has been withdrawn, and the manuscript now states that civilian casualties do not by themselves establish algorithmic inaccuracy, nor does correct identification establish proportionality or that feasible precautions were taken. \n<b>10. The analytical chain and the contribution.<\/b> Accepted. Figure 1 sets out the chain from aim to conclusions, and &sect;6.5 states three propositions framed to travel beyond the two cases: the volume\u2013time ratio as the binding constraint on meaningful human control, nominal authorisation as its failure mode under high throughput, and the distinction between negative and positive algorithmic authority.","URL":"https:\/\/f1000research.com\/articles\/14-1200\/v4#referee-comment-17252","order":6,"name":"referee-comment-17252","label":"Referee Comment","group":{"name":"article-reports","label":"Article Reports"}},{"value":"10.5256\/f1000research.211354.r528512, Burak Oktenli, Georgetown University, Washington, DC, 10 Oct 2026, version 4, indexed","URL":"https:\/\/f1000research.com\/articles\/14-1200\/v4#referee-response-528512","order":7,"name":"referee-response-528512","label":"Referee Report","group":{"name":"article-reports","label":"Article Reports"}},{"value":"The author(s) declared that no grants were involved in supporting this work","order":8,"name":"grant-information","label":"Grant Information"},{"value":"This is an open access article distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.","order":0,"name":"copyright-info","label":"Copyright"}]}}