Israel’s published targeting standard sets no fixed number of civilian deaths it will accept. The New York Times reported on 26 December 2024, citing anonymous Israeli officers, that after 7 October 2023 officers could approve strikes expected to kill up to 20 civilians each. The IDF, as The Times of Israel relayed, confirmed that its rules of engagement changed when the war began.
Israel’s own legal papers place tolerance in the harm expected before each strike, not in the toll afterward. In the Foreign Ministry’s 2023 words, compliance “is conduct-oriented, not result-oriented.” Neither party’s published death count links a single death to a strike or to the estimate approved for it.
Key Findings
- The State of Israel’s May 2015 report on the 2014 Gaza conflict, a party’s own account, states at paragraph 330 that “Excessiveness, however, is not measured using absolute numbers. It is assessed on a case-by-case basis” (Israel Ministry of Foreign Affairs, 2015).
- The Times of Israel reported on 26 December 2024 that the IDF confirmed its rules of engagement changed when the war broke out on 7 October 2023. The IDF maintained that all practices remained consistent with international law. The per-strike thresholds of 20 and 10 civilians in the same article are The New York Times’s figures from anonymous officers, not the IDF’s (The Times of Israel, 2024).
- Israel’s Foreign Ministry, in a paper updated to 2 November 2023, states its own legal position: a strike causing “more collateral damage than expected” does “not necessarily indicate a violation: compliance is conduct-oriented, not result-oriented” (Israel Ministry of Foreign Affairs, 2023, Key Legal Aspects).
- The IDF Military Advocate General’s Corps reported 74 criminal investigations opened as at 3 August 2024, against over 1,000 incidents referred for factual assessment. These are its own count, not independently verified. They were produced under a threshold in which civilian death “in most cases” does not by itself indicate “a reasonable suspicion of criminal conduct” (Israel Defense Forces, Military Advocate General’s Corps, 2024).
- Gaza’s Ministry of Health recorded 73,670 cumulative deaths in its daily report of 10 September 2026. That is a party’s own count, not independently verified, and the ministry’s published name lists carry no field linking a death to a strike (Ministry of Health, Gaza, 2024 to 2026).
Is there a fixed number of civilian deaths the IDF accepts?
The State of Israel’s May 2015 report on the 2014 Gaza conflict states at paragraph 330: “Excessiveness, however, is not measured using absolute numbers. It is assessed on a case-by-case basis.” That is a party’s own statement of its standard. Its published text sets no numeric ceiling on the civilian deaths a strike may be expected to cause.
The IDF restated the same structure for the current war on its page on data technologies, dated 18 June 2024 and updated 14 September 2026. “In each attack, IDF procedures require conducting an individual assessment of the anticipated military advantage and the incidental damage expected to civilians and civilian objects,” the page states. “Such assessments are to be made in relation to each specific strike, and not categorically.” The procedures, it adds, “do not permit carrying out strikes on a category basis, without conducting individual assessments per strike.”
Who counts as a civilian in that assessment is set by a separate line in the same 2015 report. Paragraph 265 states that Hamas’s military wing “is an organised armed group, and its members are therefore lawful targets of attack at all times and in all places,” except when “hors de combat or entitled to special protection.” Israel’s Foreign Ministry repeated the principle in its paper “Hamas-Israel Conflict 2023: Key Legal Aspects,” updated to 2 November 2023: “the IDF only targets persons who are members of organized armed groups or civilians directly participating in the hostilities.” The expected civilian toll of any strike is therefore a count of people outside that category, as the IDF classifies them.
Three documents, published in 2015, 2023 and 2024 to 2026, describe the same test: an expected-harm estimate weighed against an expected military advantage, strike by strike, with no published number on either side of the scale. All three are a party’s own description of its rules. None of them reports how often the rules were applied, what expected tolls were approved, or how the outcomes compared with the estimates. Under this test, a level of tolerance would show up in the expected-harm figures written into individual approvals. For the war that began in October 2023, no approval of that kind has been published by the IDF.
Who says the IDF accepts high civilian tolls, and since when?
The Times of Israel reported on 26 December 2024 that the IDF had responded to a New York Times investigation. In that response it confirmed that its rules of engagement changed when the war broke out on 7 October 2023, and it maintained that all practices remained consistent with international law. The confirmation is the outlet’s paraphrase of a party’s own response.
The numbers in that reporting are the newspaper’s, not the IDF’s. The New York Times, as The Times of Israel summarised it, drew on anonymous officers to describe thresholds of 20 and 10 civilians per strike and a tightening it dated to 5 November 2023. The IDF told the paper that “officers always verified the information provided” by AI systems. It described those tools as no more than a “starting point” of target selection. The Guardian had published an analysis on 3 April 2024 under the headline “Israel’s rules of engagement seem looser than ever.” So the record from 2024 onward holds a claim of loosened thresholds and a party’s confirmation that the rules changed. It holds no IDF figure for what the new rules permit.
The largest figure in circulation comes from the documentary NAZA. An interviewee in a promotional clip, quoted by The Times of Israel on 14 September 2026, was asked for the largest expected civilian toll ever approved for a strike. He replied: “I don’t know how many were killed, because they never check this, but there was once an approval for 500.” His own sentence makes the 500 an approval figure, not a death toll, and states first that the outcome was never checked.
The IDF answered on X on 14 September 2026, Israel time: “This is completely false. The IDF has never planned, approved, or carried out a strike in Gaza in which 500 civilians or anything near that number were anticipated to be killed.” As quoted by Israel Hayom, the statement continued: “Nor was any credible claim raised throughout the war suggesting that an IDF strike had caused anything near that number of fatalities.” Both outlets quote its further line that “nothing about the interviewee, not his service nor his role can be independently checked and verified.” The IDF’s reply is a party’s own statement. The interviewee’s claim rests on one unnamed witness. The documents each side has offered are set against each other in a separate reading of the IDF’s targeting records.
What does Israel say a higher-than-expected civilian toll means legally?
Israel’s Foreign Ministry, in “Hamas-Israel Conflict 2023: Key Legal Aspects,” updated to 2 November 2023, states its own legal position: “If the attack results in more collateral damage than expected, this does not necessarily indicate a violation: compliance is conduct-oriented, not result-oriented.” Under that test, the civilian toll after a strike is not, on its own, the measure of what was accepted.
The same paper names the difficulty that sits inside the test: “Estimating expected civilian harm can be very challenging, especially when operating in a complex and dense urban environment where civilian presence is dynamic and often hard to ascertain.” The 2015 report on the 2014 conflict adds a constraint on the process. Paragraph 247 states that “certain input, such as fact-specific legal advice, will not be available in real time” in time-sensitive combat. In such cases commanders “rely on the training they have received, as well as directives that specify the checks and authorisations required prior to carrying out attacks.”
This is the point at which two readings of the same practice part company. One reading measures tolerance by outcome: a war-level toll, or the deaths in a given strike. Israel’s stated test measures conduct by expectation: what the commander anticipated, on the information available, at the moment of approval. A large gap between the expected and the actual toll is, on Israel’s stated position, not a breach in itself. The same gap can be read as evidence of what was tolerated under the other approach. Both readings describe the same strikes. They differ on which number is the measure.
The test also determines what evidence would settle the question. On Israel’s own terms, the decisive record is the expected-harm figure and the information behind it for each approval. The 2015 report reproduces one such approval, for a 2014 strike. Neither the 2023 paper nor any IDF publication on the current war prints an expected-harm figure for a single strike.
Why does Israel judge each strike’s civilian toll on its own?
Israel’s Foreign Ministry, in its FAQ updated to 8 December 2023, states at question 8 that proportionality is “to be applied to each and every attack independently.” Its illustration: “one attack causing 500 casualties should logically, and legally, be seen differently than 500 attacks against military objectives resulting overall in the same number of casualties.” Both statements are a party’s own legal position.
The 500 in that passage is a hypothetical published in December 2023, almost three years before NAZA, and it is not a reply to the film. It shows how the stated method is built. Each strike carries its own expected harm and its own expected military advantage. A cumulative toll for the war is therefore a sum of many separate judgments, and no war-level tolerance figure exists anywhere in the method.
The IDF’s data-technologies page of 2024, updated in 2026, applies the same rule from the other direction: assessments are made “in relation to each specific strike, and not categorically.” The 2015 report’s paragraph 330 rejects “absolute numbers” on the same ground.
The consequence for the question is one of category. An aggregate toll and a per-strike threshold answer different questions. The first describes what the war produced. The second describes the rule an officer applied before a single strike. The thresholds of 20 and 10 civilians per strike that The New York Times reported in 2024, from anonymous officers, belong to the second category. The cumulative counts published by Gaza’s Ministry of Health belong to the first. Under Israel’s stated method, a tolerance level would be read from the distribution of expected-harm figures across approvals. No party to the war publishes that distribution.
Does a warning change how much civilian risk Israel accepts?
The State of Israel’s 2015 report states at paragraph 306 that “the provision of a warning never, on its own, affected a proportionality assessment.” The same paragraph states that after a warning “the IDF did not assume that a relevant site or area had been evacuated.” This is a party’s own account of its 2014 practice.
The 2023 Key Legal Aspects paper carries the same rule into the current war: “the IDF does not assume that there are no civilians in areas or sites where advance warnings had been given.” Warnings also have a stated exception. The 2015 report, paragraph 308, says none is required “where the element of surprise is necessary for the success of a military operation (e.g., where a target is a militant who would escape if warned).” The 2023 paper restates it: where a warning is not feasible, “there is no legal requirement to do so.” The 2015 report adds at paragraph 307 that civilians who stayed after a warning were not regarded “as voluntary human shields and thus legitimate targets for attack.”
Israel’s Foreign Ministry FAQ to 8 December 2023 gives its own counts of the warning effort. It reports over 45,000 individual phone calls and over 15 million text messages. It also reports over 12 million pre-recorded calls and over 4.5 million leaflets. Each figure is a floor, cumulative to that date, and a party’s own count. The channels overlap, so the figures are never summed into a number of people warned.
By Israel’s own account, then, a warning works on the inputs to the estimate, not on the threshold. It may reduce the number of people present, and the estimate is supposed to count whoever remains. None of the published counts measures how many people left after any warning. The effect of warnings on expected harm is therefore stated in the record but not quantified.
Where does expected civilian presence enter the IDF’s approval chain?
The IDF’s undated targeting presentation, slide 18, places “Civilians in Target?” among four yes-or-no gates in an eight-step chart. The chart runs from “Intel Based Targets” to “Permission to Strike.” The other three gates are “Area Evacuated?”, “Proximity to Sensitive Sites?” and “Target Coherent RoE?”. The chart is a party’s own diagram of its procedure.
The input to that gate is described on slide 10. A Southern Command team “led by a Brigadier General” developed “a layered map which divides Gaza into 630 sectors. For each sector, the unit made continuous assessments of the population presence that is there.” The deck mixes material from 2014 with material from the current war, and slides 10 to 13 belong to the current war. Israel’s Foreign Ministry FAQ of December 2023 describes a matching “Civilian Harm Mitigation Unit” that “constantly monitors in real time the implementation of these evacuation recommendations.” How that unit’s 630-sector map relates to strikes in populated areas is read separately.
The 2015 report, paragraph 245, describes what a commander may do when the civilian picture is uncertain. A senior commander may “suspend the attack (for instance, because more information about potential collateral damage is necessary), or (iii) abort the attack altogether.” Paragraph 316 adds that “individual pilots exercised their discretion to cancel or delay planned strikes due to the presence of civilians.”
What the deck does not supply is a count. Slide 17 states: “Many strikes were cancelled due to unexpected presence of civilians.” “Many” is the whole quantification, and the only dated example on the slide is from 14 July 2014. The FAQ of December 2023 says “countless attacks have been aborted or avoided completely,” which is also not a number. The party’s own record shows where the civilian-harm judgment sits in the chain. It does not show how often that judgment stopped a strike in this war.
Of over 1,000 reviewed incidents, how many became criminal cases?
The IDF Military Advocate General’s Corps reported 74 criminal investigations opened as at 3 August 2024. Over 1,000 incidents had been referred for factual assessment by that date. These are the IDF’s own count of its own processes, published by a party to the events and not independently verified. They are investigations opened, not findings, indictments or convictions.
The count is a product of the threshold the Corps applies, and the threshold travels with it. In the Corps’ own words, civilian death or injury, “while tragic and regrettable, in most cases do not by themselves indicate a reasonable suspicion of criminal conduct, as they can be explained as reasonable aftermath of lawful operational activity.” This is the after-the-fact form of the conduct-oriented test in Israel’s 2023 legal paper. A civilian toll opens a criminal file only when something beyond the toll suggests criminal conduct.
The breakdown shows where the files sit. Forty-four concern deaths of detainees held by the IDF. Of the five published categories, 6 allege illegal use of force and 3 allege destruction of civilian property without military necessity. No category in the published breakdown is named for a strike’s expected civilian harm. The Corps said hundreds of the referred incidents were still under assessment at the time of writing.
The review body, the Fact-Finding and Assessment Mechanism, was created during the 2014 Gaza conflict on a recommendation of the Turkel Commission. It now operates as “a permanent body which conducts factual assessments.” The Corps states that “The MAG is subject only to the law on professional matters and operates with full independence.” It adds that its decisions can be challenged before Israel’s Attorney General and Supreme Court. Those are the Corps’ own statements about its own process.
Can the published death counts show the harm accepted per strike?
Gaza’s Ministry of Health recorded 73,670 cumulative deaths in its daily report of 10 September 2026. That is its own count as a party to the war, not independently verified. Its published name lists carry no field for the strike, the place of death or combatant status, so no death on them can be tied to an approval or an expected-harm estimate.
The lists are built from identity records. The ministry’s list to 30 June 2024 has seven columns: serial number, name, document number, date of birth, sex, age and source. No list read for this report carries a date of death. On 29 July 2024 the ministry stated that a death was registered either on the family’s report or on the date the body reached hospital. A committee was then formed to audit the actual date. The same ministry reported 28,185 records with complete data and 9,715 whose data was not yet complete as of 30 June 2024. The figures each register counts are compared in a reading of the ministry’s two registry channels.
The IDF’s own figure measures something else. The IDF Spokesperson stated on 29 February 2024 that over 13,000 people the IDF describes as terrorists had been eliminated in Gaza since the war began. That classification is the IDF’s own and is not independently verified. The two figures come from different dates, different parties and different classifications. No ratio between them is computed here.
| Record | What it counts | Strike-level field |
|---|---|---|
| Gaza Ministry of Health daily report, 10 September 2026 | 73,670 cumulative deaths, the party’s own count | None |
| Gaza Ministry of Health name list to 30 June 2024 | 28,185 records with complete data, the party’s own count | None; date of birth, not date of death |
| IDF Spokesperson briefing, 29 February 2024 | Over 13,000 people the IDF describes as terrorists, its own count | None |
Source: Ministry of Health, Gaza, 2024 to 2026; Israel Defense Forces, 2024, press briefing of 29 February 2024. Both are parties to the war.
Each count describes an outcome across the war. Under Israel’s stated test, tolerance is measured per strike, against the estimate made before it. No body publishes a record joining deaths to the strikes that caused them and the expected-harm figures approved for those strikes.
Methodology and limitations
The evidence is drawn from three source types. The first is official and primary records of the parties. These include the State of Israel’s May 2015 report on the 2014 Gaza conflict, chapter VI, cited by paragraph and by report page 144 for the target card. They also include two Israeli Foreign Ministry papers of November and December 2023, the IDF’s data-technologies page (18 June 2024, updated 14 September 2026), and the IDF Military Advocate General’s Corps account of 24 February 2024, updated 3 August 2024. To these are added the IDF Spokesperson’s briefing of 29 February 2024, the IDF’s statement of 14 September 2026, an undated 18-slide IDF targeting presentation provided to the Institute, and Gaza’s Ministry of Health daily reports and name lists. The second type is press relays of named statements: The Times of Israel, Israel Hayom and The Guardian. The documents were read directly on 22 September 2026, and the Gaza ministry documents on 10 September 2026.
Every figure from the IDF, an Israeli ministry or Gaza’s Ministry of Health is a party’s own count or statement. Each describes a stated standard or a self-reported total and is not evidence of compliance or independently verified. The New York Times thresholds of 20 and 10 civilians per strike reach this report through The Times of Israel’s summary. They rest on anonymous officers and are the newspaper’s figures. The NAZA interviewee is unnamed, and his account cannot be checked. The IDF’s statement in reply was read from the post’s public syndication record and, where truncated, from the two outlets’ quotations.
The targeting presentation is undated and mixes periods. Slide 15 is the same 2014 target card Israel published in May 2015, and slide 17’s only dated example is from July 2014. Slide 12 dates to February 2024, and slides 10 to 13 are current-war material. The deck publishes no count of warnings, approvals or cancellations. Gaza’s Ministry of Health lists change schema between rounds, so each field statement refers to a named list. Every count in the record is cumulative or a floor, and none is broken down by strike.
Conclusion
So does the IDF have a high tolerance for civilian casualties? The answer depends on which number is taken as the measure, and the record holds only one side of the comparison. Israel’s own published standard contains no fixed number. Its 2015 report rejects “absolute numbers,” and its 2023 legal paper calls compliance “conduct-oriented, not result-oriented.” The New York Times reported thresholds of 20 and 10 civilians per strike from anonymous officers. The IDF, by The Times of Israel’s account, confirmed its rules of engagement changed on 7 October 2023.
The portrait the documents draw is consistent. The test is applied per strike and by expectation. A warning changes the estimate but never the threshold, by Israel’s own 2015 account. Civilian presence is a gate in the IDF’s own approval diagram, and the deck counts its cancellations only as “many.” The MAG Corps, by its own count, opened 74 criminal investigations from over 1,000 referred incidents. It did so under a rule that treats most civilian deaths as the aftermath of lawful activity. Gaza’s Ministry of Health counts 73,670 dead, a party’s own unverified total, on lists with no field for any strike.
Every party has published either a standard or a toll, and none has published the expected-harm figure for a single approval in this war. A tolerance can be stated in a rule or inferred from a toll, but it can only be measured in the space between the two. If the record that would measure it exists only inside the institution whose tolerance is in question, how far can any public judgment of that tolerance rest on evidence?
Frequently Asked Questions
Does the IDF use AI to select targets in Gaza?
The IDF’s data-technologies page, in its 14 September 2026 update, answers “No.” It describes Lavender as “a general-purpose database” and states that a suggestion from Habsora, first used operationally in 2021, “cannot be solely relied upon to identify a target.” These are the IDF’s own descriptions of its tools.
Who decides to strike once a target is identified?
The same IDF page states that “intelligence officers who identify targets are not the ones who also decide to strike those targets,” and that “operational commanders are those who approve strikes.” The 2015 report, paragraph 245, adds that the legal adviser’s opinion on an attack is binding. Both are a party’s own account of its procedure.
Are IDF legal advisers independent of the commanders they advise?
Israel’s Foreign Ministry stated in November 2023 that IDF legal advisers “are not subject to the IDF chain of command on professional matters, but report directly to the Military Advocate General.” The 2015 report, paragraph 238, states that the Military Advocate General “is appointed by the Minister of Defence (a civilian authority), not by the IDF Chief of General Staff.” Both are Israel’s own description of the structure.
How many strikes did the IDF report in the 2014 Gaza conflict?
The State of Israel’s 2015 report states at paragraph 272 that “the IDF conducted over 6,000 aerial strikes against military objectives.” At paragraph 304 it reports “specific warnings prior to more than a thousand attacks against individual targets.” Both figures are floors in a party’s own account.
What does a published IDF legal review of a strike look like?
The one reproduced in Israel’s 2015 report, on report page 144, covers a strike of 8 July 2014 on an arms cache in a residential building. Its legal review is “Conditional” on “the provision of an effective advance warning.” It sets a night-time strike and “Number of bombs: One,” and the IDF states it is “unaware of any claim of fatality or harm.” That last statement is the party’s own.
What did the IDF report doing before the 2014 Shuja’iyeh ground operation?
The 2015 report states at paragraph 302 that the IDF dropped “over 150,000 leaflets” directing residents to central Gaza City. At paragraph 309 it states that the IDF “delayed its ground manoeuvre in the area of Shuja’iyeh for more than 24 hours.” Both are a party’s own account of its 2014 conduct.
Who is left out of the Gaza Ministry of Health’s count?
The ministry stated on 7 January 2024 that its count excludes the missing and those under rubble, and the unidentified who reached hospitals. It also excludes unidentified bodies Israel handed over, burials outside hospitals, and deaths in Gaza and North Gaza after its information system stopped in November 2023. That is the issuing body’s own scope statement for its own count.
Sources
- Israel Defense Forces, 2024, updated 2026. The IDF’s Use of Data Technologies in Intelligence Processing. Israel Defense Forces, published 18 June 2024, updated 14 September 2026. https://www.idf.il/en/mini-sites/idf-press-releases-israel-at-war/june-24-pr/the-idfs-use-of-data-technologies-in-intelligence-processing-published-june-18-2024/. Official or primary record (a party to the conflict).
- Israel Defense Forces, 2024. Press Briefing by IDF Spokesperson, Rear Admiral Daniel Hagari, February 29th, 2024. Israel Defense Forces. https://www.idf.il/en/mini-sites/israel-at-war/briefings-by-idf-spokesperson-bg-effie-defrin/february-24-press-briefings/press-briefing-by-idf-spokesperson-rear-admiral-daniel-hagari-february-29th-2024/. Official or primary record (a party to the conflict).
- Israel Defense Forces, 2026. Statement on the “500” claim in NAZA, posted on the IDF’s X account, 14 September 2026 (Israel time), as relayed by The Times of Israel liveblog, https://www.timesofisrael.com/liveblog_entry/idf-denies-claim-in-new-naza-documentary-that-it-planned-gaza-strike-with-500-civilian-deaths/, and Israel Hayom, https://www.israelhayom.com/2026/09/14/idf-denies-naza-documentary-claims/. Official statement by a party, relayed by named press.
- Israel Defense Forces, undated. Addressing Challenges Through LOAC | IHL (targeting process presentation), 18 slides. Published by the Hanover Institute at /hanover/media/naza/idf-targeting-process-presentation.pdf. Official or primary record (a party’s own presentation).
- Israel Defense Forces, Military Advocate General’s Corps, 2024. Addressing Alleged Misconduct in the Context of the War in Gaza. Israel Defense Forces, published 24 February 2024, updated 3 August 2024. https://www.idf.il/en/mini-sites/military-advocate-generals-corps/addressing-alleged-misconduct-in-the-context-of-the-war-in-gaza-published-february-24-2024/. Official or primary record (a party’s own counts).
- Israel Ministry of Foreign Affairs, 2015. The 2014 Gaza Conflict: Factual and Legal Aspects, Chapter VI: IDF Conduct of Operations During the 2014 Gaza Conflict. State of Israel, May 2015. https://www.gov.il/BlobFolder/generalpage/idf-conduct-of-operations-during-the-2014-gaza-conflict/en/English_Documents_IDF1-Conduct-2014-Gaza-Conflict.pdf. Official or primary record (a party to the conflict).
- Israel Ministry of Foreign Affairs, 2023. Hamas-Israel Conflict 2023: Frequently Asked Questions. Updated to 8 December 2023. https://www.gov.il/en/pages/swords-of-iron-faq-6-dec-2023. Official or primary record (a party to the conflict).
- Israel Ministry of Foreign Affairs, 2023. Hamas-Israel Conflict 2023: Key Legal Aspects. Updated to 2 November 2023. https://www.gov.il/BlobFolder/generalpage/swords-of-iron-faq-6-dec-2023/en/English_Documents_Hamas-Israel%20Conflict%202023%20-%20Some%20Factual%20and%20Legal%20Aspects%20-%20Israel%20Ministry%20of%20Foreign%20Affairs%20(2%20NOV%202023).pdf. Official or primary record (a party to the conflict).
- Ministry of Health, Gaza, 2024 to 2026. Published martyr lists, daily statistical reports and completeness statements. Ministry of Health media office channel, https://t.me/MOHMediaGaza, messages 4739, 5651, 5663 and 7607. Official or primary record (a party to the war; self-reported, not independently verified).
- The Guardian, 2024. “Israel’s rules of engagement seem looser than ever” (headline, 3 April 2024). The Guardian. https://www.theguardian.com/world/2024/apr/03/israels-rules-of-engagement-seem-looser-than-ever-if-they-are-followed-at-all. News analysis.
- The Times of Israel, 2024. IDF loosened rules of engagement after Oct. 7, allowing more civilian deaths (summary of a New York Times investigation). The Times of Israel, 26 December 2024. https://www.timesofisrael.com/idf-loosened-rules-of-engagement-after-oct-7-allowing-more-civilian-deaths-ny-times/. News report.