The IDF’s own targeting presentation sets out 9 stages from intelligence collection to execution, names no software in any of them, and puts a qualified commander’s approval at the eighth. NAZA’s interviewees, The Guardian reports, describe AI-based systems built to identify large numbers of potential low-ranking Hamas targets. The IDF states that strike decisions “are made by human personnel only,” and describes its two named systems as a tool that points analysts toward objects and a database that is not a list of targets. Its records show who approves a strike. They do not show how a target first entered the process, or how often a system’s suggestion became one.
Key Findings
- The IDF presentation, a party’s own document, sets out 9 targeting stages, with legal review seventh and commanders’ approval eighth, and names no system or software at any stage.
- The IDF states that Habsora, in operational use since 2021, “cannot be solely relied upon to identify a target,” and that Lavender “is a general-purpose database,” “not a list of confirmed military operatives eligible to attack.”
- Israel’s May 2015 report, its own account, states that a senior commander may approve, suspend or abort a pre-planned attack after a legal adviser’s binding opinion, and that such advice “will not be available in real time” for time-sensitive targets.
- In 6 of the 18 clips of IDF footage provided to the Institute, the subtitles use a recommendation or an acknowledgment, “I recommend” or “Copy that”; no subtitle in any clip refers to a system or a score.
- The Guardian reports NAZA’s insiders describing AI systems that identified “large numbers of potential low-ranking Hamas targets”; no IDF record examined gives a count of targets suggested, approved or rejected.
What does NAZA allege about AI and targeting?
The Guardian, which produced NAZA, reports that its 24 Israeli military insiders describe “AI-based systems built to identify large numbers of potential low-ranking Hamas targets, who were then bombed in their homes at night, in the knowledge attacks would also kill their families.” The allegation concerns selection: that software proposed whom and where to strike, and in large numbers.
The directors put it more concretely in an interview with the same newspaper. The insiders, The Guardian reports, “tell us how they helped develop an AI mapping tool that trawled for phone signals and recommended apartment buildings in Gaza to bomb.” The claim is about where targets came from, not only about who approved them.
The IDF’s own answer of 11 September 2026 addresses both. It states that “decisions regarding the selection and approval of strikes in the IDF are made by human personnel only, in accordance with IDF guidelines, and not by artificial intelligence.” It adds that the film “relies on anonymous testimonies from individuals whose identity cannot be verified, including whether they served in the IDF, in what capacity, or whether they were involved in the matters about which the allegations are being made.” Of one interviewee, the IDF said three days later that “nothing about the interviewee, not his service nor his role can be independently checked and verified.”
Two questions sit inside the dispute and they are not the same question. The first is whether software generated the candidates. The second is whether a person decided to strike them. The IDF’s denial answers the second directly and names “selection” as well. Its published records, set out below, describe the second in detail and the first in a few sentences. The Institute has not viewed the film; its content enters here through its producer’s reporting.
What does the IDF say Habsora and Lavender do?
The IDF states that its system Habsora was “first utilized in an operational setting in 2021,” its own account, and that it “suggests to the intelligence analyst to focus his or her research on certain physical objects with greater potential of a military affiliation with a terror group.” On the IDF’s description, the system proposes where to look, and a person looks.
The same page, first published on 18 June 2024 and updated on 14 September 2026, sets two conditions on that suggestion. “A suggestion from ‘Habsora’ cannot be solely relied upon to identify a target without being thoroughly explored by the intelligence analyst, and approved by an authorized intelligence officer.” The analyst researches; a second officer approves the result as a target.
Of the second system, the IDF states that “‘Lavender’ is a general-purpose database,” and that “this is not a list of confirmed military operatives eligible to attack.” The page’s questions and answers, added in the 2026 update, ask “Does the IDF use AI to select targets in Gaza?” and answer with one word: “No.”
The two descriptions use different verbs for a first step. The directors’ tool “recommended apartment buildings”; the IDF’s system “suggests” physical objects for research. They part at the step after: the film’s account is of suggestions that became strikes, and the IDF’s is of suggestions that a person researched and another approved.
The IDF page gives no figure for how many suggestions either system produced, how many an analyst reviewed, or how many became targets. Its account is of the rule each suggestion passes through, not of how many passed. The Institute’s report on the IDF’s targeting record sets out the rest of the presentation those rules sit beside.
Where would a system sit in the IDF’s nine stages?
The IDF presentation, its own document, lists 9 stages and repeats the list on four of its 18 slides: intelligence collection, analysis, identifying military objectives, link to mission, validation, planning, legal review, commanders’ approval, and execution with battle damage assessment and lessons learned. None of the nine names a system, a database or any software.
The presentation does not place Habsora or Lavender anywhere. The IDF’s separate description does place Habsora’s output with “the intelligence analyst,” and on the presentation’s own list the analyst’s work is the first three stages: collection, analysis and the identification of military objectives. The IDF’s page adds an authorized intelligence officer who approves the result as a target, a step the presentation’s list does not name separately.
Israel’s May 2015 report on the 2014 war, its own account, describes the same sequence in more words. The procedures “call for the collection of intelligence about the nature of a potential target,” which “is relied upon to confirm that the potential target constitutes a military objective,” and the input from intelligence and planning “undergoes timely re-evaluation and re-validation according to IDF procedures.” That matches the presentation’s fifth stage, validation.
The pattern across the three records is the same. The later stages, from validation to approval, are described in some detail, with named roles, conditions and outcomes. The first three, where the allegation as The Guardian reports it places the software, are named in a word or two each and described nowhere at length.
That is not a finding that anything happened in those stages, or that nothing did. It is the shape of what the IDF has published: its records say most about the later stages and least about the first ones, which is where the allegation sits.
Who approves a strike, in the IDF’s own account?
The IDF presentation puts approval at the eighth of its stages and describes the approving commander as “an officer who has been professionally and operationally qualified for this role and possesses the authority to authorize strikes against designated targets.” It gives no rank, and no count of commanders who hold that authority.
The IDF’s page on its data systems states the separation in two sentences: “intelligence officers who identify targets are not the ones who also decide to strike those targets,” and “operational commanders are those who approve strikes.” On the IDF’s account, then, the officer who approves a suggestion as a target and the commander who approves the strike are different people, one in intelligence and one in operations.
Israel’s May 2015 report describes what the approving commander can do. After receiving the professional opinions, including a legal adviser’s, “a senior commander may (i) approve the attack (where appropriate, subject to certain conditions), (ii) suspend the attack (for instance, because more information about potential collateral damage is necessary), or (iii) abort the attack altogether.” The report states that “commanders may add additional conditions for carrying out the attack” beyond the legal ones.
Put together, the IDF’s own records name four roles between a system’s suggestion and a strike: the analyst who researches it, the intelligence officer who approves it as a target, the legal adviser who reviews it, and the commander who approves, suspends or aborts.
None of the records gives a figure for how long each role took over a target, how often a target was sent back, or how many were aborted at approval. The Institute’s report on the question of targeting civilians sets out what the IDF’s rules say the approving commander weighs.
What part do the IDF’s lawyers play?
Israel’s May 2015 report, its own account, states that IDF legal advisers “are not subordinate to the commanders they advise,” and that the Military Advocate General “is appointed by the Minister of Defence (a civilian authority), not by the IDF Chief of General Staff.” It adds that “Legal opinions of the MAG Corps are binding upon the IDF, including with regard to the legality of individual attacks.”
The Foreign Ministry repeated the point in November 2023, stating 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 Military Advocate General’s Corps states of its own head that “The MAG is subject only to the law on professional matters and operates with full independence.”
The legal review is the presentation’s seventh stage, and the one review Israel has published shows what it produces. The target card for a strike on 8 July 2014 is marked “Conditional,” on “the provision of an effective advance warning.” The Institute’s report on the one target card Israel has published sets out its contents.
The same 2015 report states where the lawyer is absent. In time-sensitive situations, “particularly in dynamic, fast-moving ground combat,” “certain input, such as fact-specific legal advice, will not be available in real time,” and commanders “rely on the training they have received, as well as directives that specify the checks and authorisations required prior to carrying out attacks.”
On Israel’s own account, then, a legal adviser reviews pre-planned attacks and may not review the fastest ones. None of the records says what share of strikes in the current war were pre-planned and what share time-sensitive, so none says how many passed a legal review before they were carried out.
Does membership alone make someone a target, in Israel’s view?
The IDF presentation defines a military target in 2 parts, and the second is people: “A Member of an enemy armed force, a member of an organize [sic] armed group, or a civilian directly participating in hostilities (for such time that he does so).” The definition carries no rank.
Israel’s May 2015 report, its own account, states the rule the definition applies. 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 for when such persons are hors de combat or entitled to special protection due to their particular function).” The Foreign Ministry’s November 2023 paper states that the IDF “only targets persons who are members of organized armed groups or civilians directly participating in the hostilities.”
On that stated rule, the rank in the Guardian’s allegation, “low-ranking,” does not decide whether a member may be attacked. What the rule leaves to the facts is membership itself: who is a member, and on what evidence.
The 2015 report describes how that is decided for a group but not for a person. Whether an entity counts as an organised armed group rests on “comprehensive, timely intelligence assessments (which are reviewed and updated as necessary) and consultations with military lawyers as well as with the highest military echelons,” under oversight “by Israel’s highest legal echelons, including the Ministry of Justice.”
For an individual, the records say less. The IDF describes Lavender as “not a list of confirmed military operatives eligible to attack,” which says what the database is not rather than how any person’s membership was confirmed. That confirmation is the step the film’s account of large numbers of low-ranking targets turns on, and the step the IDF’s records do not describe.
Does the IDF say each strike is weighed on its own?
The IDF states, on a page first published in 2024, its own account, that its procedures require “an individual assessment of the anticipated military advantage and the incidental damage expected to civilians and civilian objects” for each attack, and “do not permit carrying out strikes on a category basis, without conducting individual assessments per strike.”
The rule answers the film’s allegation in its own terms. A system that produced targets by category, such as every member of a group, would on the IDF’s stated procedure still require a separate weighing for each strike. Israel’s Foreign Ministry stated the same in December 2023: proportionality is “to be applied to each and every attack independently.”
The treaty text sets the duty on people. Article 57(2)(a)(i) of Additional Protocol I requires that “those who plan or decide upon an attack shall do everything feasible to verify that the objectives to be attacked are neither civilians nor civilian objects.” Israel is not a party to Additional Protocols I or II; its May 2015 report states that it “abides by their provisions inasmuch as they reflect rules of customary international law.”
The IDF answered a New York Times investigation in December 2024 in similar terms. The Times of Israel reports that the IDF told the newspaper that “officers always verified the information provided” by AI systems, and denied that AI tools were anything more than a “starting point” of the target selection process. It also reports that the IDF confirmed its rules of engagement changed when the war broke out, while maintaining that its practices stayed consistent with international law.
An individual assessment is a rule about process, and the IDF states it plainly. Whether it held in each strike is the question the film raises and the records examined here do not answer, because none of them records an individual assessment for any strike of the current war.
What do the IDF’s clips show about who stops a strike?
In 6 of the 18 clips of IDF targeting footage provided to the Hanover Institute, the IDF’s own record, the subtitles use the language of a chain of decision: a recommendation, “I recommend,” or an acknowledgment, “Copy that” or “Copy.” No subtitle in any of the 18 refers to a system, a database or a score.
| Wording in the subtitles | Clips | Dates |
|---|---|---|
| “I recommend” | 2 | 22 May 2024, 27 June 2025 |
| “Copy” or “Copy that,” without a recommendation | 4 | 17 May, 24 May, 8 August and 14 August 2024 |
| Neither | 12 | The remaining clips |
- There are people there
- I recommend to abort, there is a group of five people now, even more, possibly children and more. A lot
- A lot, it’s an appropriate time to abort
On 22 May 2024 one voice says “I recommend to abort, there is a group of five people now, even more, possibly children and more.” On 27 June 2025: “There is a woman, I recommend, a child walking nearby,” then “Copy that, divert [ammunition].” The recommendation and the order that follows it come in separate lines, which the subtitles attribute to no one by name.
- There is a woman, I recommend, a child walking nearby
- There are two children
- Copy that, divert [ammunition]
Article 57(2)(b) of Additional Protocol I provides that an attack “shall be cancelled or suspended if it becomes apparent” that it may be expected to cause excessive civilian harm. The clips show that kind of cancellation, subtitled as spoken exchanges, in the last seconds before a strike.
What they cannot show is where the target came from. Every clip begins at the end of the process, at the ninth stage, execution, and none shows how the target was proposed or approved. The 18 are a selection made before they reached the Institute, published here exactly as received, and the absence of a system in their subtitles is evidence about the moment of stopping, not about selection.
What can the IDF’s records not show about target selection?
The IDF’s records examined here publish no count of targets suggested by a system, approved as targets, approved for strike, suspended or aborted, for any period of the war. Each record describes a rule or an example; none gives the number of times the rule was applied.
The gap falls where the dispute is. The film’s insiders, as The Guardian reports them, speak of the first stages, where software proposed buildings and people. The IDF’s presentation names those stages in a word or two, its data page describes the two systems in general terms, and its 2015 report describes the collection of intelligence in a few sentences.
The testimony on the other side cannot be checked either. The IDF states that the film’s interviewees cannot be verified as to whether they served, in what capacity, or whether they were involved in what they describe, and the Institute has not viewed the film. Neither account of the first stages can be set against a published record of them.
What the records do settle is the stated structure. On the IDF’s own account, a system suggests, an analyst researches, an intelligence officer approves a target, a legal adviser reviews a pre-planned attack, and a commander approves, suspends or aborts; the footage shows strikes stopped in their last seconds.
The allegation, as The Guardian reports it, concerns how many targets the systems produced and what followed, and on that the records hold no figures. A report of how many system suggestions reached each stage, and how many stopped at each, would test the film’s account and the IDF’s at once.
Methodology and limitations
This report sets the published account of the documentary NAZA against the IDF’s own records: an IDF targeting presentation provided to the Institute, 18 clips of IDF footage, the IDF’s statements on the film, the IDF’s page on its data systems, the Military Advocate General’s Corps page on alleged misconduct, and the Israeli government’s published statements of its targeting law, principally the Foreign Ministry’s May 2015 report on the 2014 war and its papers of November and December 2023. The film’s claims come from The Guardian, which produced it; the film was not viewed.
The presentation is undated, and its worked legal review is a 2014 case that Israel published in 2015. The clips appear exactly as received, with their checksums, and the Institute has not verified when or where they were recorded. A clip is counted for recommendation or acknowledgment where its subtitles contain “I recommend,” “Copy” or “Copy that”; clips containing both are counted once, under recommendation.
Every record here is a party’s own account of its own procedures, evidence of the stated standard and not of compliance with it. The Times of Israel’s account of the IDF’s reply to the New York Times is that outlet’s report of a response.
Conclusion
The IDF’s own records answer half of the title’s question and leave the other half open. Does AI decide who the IDF strikes? On the IDF’s account, no: a system suggests, and four roles of people stand between the suggestion and the strike, from the analyst to the commander who may abort. The presentation names no software at any of its nine stages, the IDF describes its database as not a list of targets, and in the footage every stop is subtitled as something said, none as a system’s output.
The film’s claim, as The Guardian reports it, is about the step before them: that software identified large numbers of targets, who were then bombed. On that, the IDF’s records hold rules and no counts. They say each strike is assessed individually and that suggestions are verified; they do not say how many suggestions there were, or how many were stopped.
The IDF’s answer and the film’s both rest on the first stages of a process whose later stages alone are documented. If the IDF published how many of its systems’ suggestions reached a commander, and how many a commander stopped, which of the two accounts would the figures leave standing?
Frequently Asked Questions
What is Habsora?
The IDF describes Habsora as a system first used operationally in 2021 that suggests to an intelligence analyst which physical objects to research for a possible link to an armed group. The IDF states that its suggestion cannot on its own identify a target without an analyst’s research and an authorized intelligence officer’s approval.
What is Lavender?
The IDF describes Lavender as “a general-purpose database” and states that it is “not a list of confirmed military operatives eligible to attack.” The IDF’s published page gives no further account of what the database holds.
Does the IDF use AI to select targets in Gaza?
The IDF states that it does not, and that decisions on the selection and approval of strikes are made by people only. The Guardian reports that NAZA’s interviewees describe AI-based systems that identified large numbers of targets.
Who approves an IDF airstrike?
The IDF presentation describes an approving commander qualified for the role and holding the authority to authorize strikes. Israel’s May 2015 report states that a senior commander may approve an attack, with or without conditions, suspend it, or abort it.
Are IDF legal advisers independent of the commanders they advise?
Israel’s May 2015 report states that they are not subordinate to those commanders and that the Military Advocate General is appointed by the Minister of Defence. The Foreign Ministry stated in November 2023 that legal advisers report directly to the Military Advocate General on professional matters.
Can an IDF strike go ahead without a legal review?
Israel’s May 2015 report states that in time-sensitive situations fact-specific legal advice “will not be available in real time,” and commanders rely on training and directives. It gives no share of strikes carried out that way.
Why can’t the film’s claims about AI be checked against the IDF’s records?
The IDF’s published records describe the rules for each stage but give no count of targets a system suggested or of how many were approved or stopped. The IDF states that the film’s interviewees cannot be verified.
Sources
- Israel Defense Forces, 2024. The IDF’s Use of Data Technologies in Intelligence Processing. 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 record, a party’s own account of its procedures.
- Israel Defense Forces, 2026. The IDF Spokesperson’s Response Following the Publication of the Movie “NAZA.” 11 September 2026. https://www.idf.il/en/mini-sites/israel-hamas-war-gaza/articles-israel-hamas-war-gaza/distributions-swords-of-iron-war-gaza/the-idf-spokesperson-s-response-following-the-publication-of-the-movie-naza/. Official statement, a party’s own.
- Israel Defense Forces, 2024 to 2026. Targeting footage of strikes diverted or called off, 18 clips. Provided to the Hanover Institute, 17 September 2026, and published as received. Official record, a party’s own footage.
- Israel Defense Forces, undated. Addressing Challenges Through LOAC | IHL, targeting process presentation, 18 slides. Provided to the Hanover Institute. /hanover/media/naza/idf-targeting-process-presentation.pdf. Official record, a party’s own presentation.
- 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. Dated May 2015, released 14 June 2015. https://www.gov.il/en/pages/idf-conduct-of-operations-during-the-2014-gaza-conflict. Official record, a party’s own account.
- 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 record, a party’s own position.
- 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 record, a party’s own position.
- Military Advocate General’s Corps, Israel Defense Forces, 2024. Addressing Alleged Misconduct in the Context of the War in Gaza. 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 record, a party’s own account.
- Protocol Additional to the Geneva Conventions of 12 August 1949, and relating to the Protection of Victims of International Armed Conflicts (Protocol I), 8 June 1977. Article 57. https://ihl-databases.icrc.org/en/ihl-treaties/api-1977/article-57. Treaty text.
- The Guardian, 2026. Secret systems used by Israel in mass killings of Gaza civilians revealed in new film. 10 September 2026. https://www.theguardian.com/world/2026/sep/10/secret-systems-used-by-israel-in-mass-killings-of-gaza-civilians-revealed-in-new-film. International press, the film’s producer.
- The Guardian, 2026. ‘They know the house they are bombing is full of children’: NAZA’s directors on their devastating film about Israeli intelligence. 11 September 2026. https://www.theguardian.com/film/2026/sep/11/naza-directors-film-israeli-intelligence-rachel-szor-yuval-abraham-. International press, the film’s producer.
- The Times of Israel, 2024. IDF loosened rules of engagement after Oct. 7, allowing more civilian deaths, NY Times. 26 December 2024. https://www.timesofisrael.com/idf-loosened-rules-of-engagement-after-oct-7-allowing-more-civilian-deaths-ny-times/. Israeli press, carrying the IDF’s response.
- The Times of Israel, 2026. IDF denies claim in new ‘NAZA’ documentary that it planned Gaza strike with 500 civilian deaths. 14 September 2026. https://www.timesofisrael.com/liveblog_entry/idf-denies-claim-in-new-naza-documentary-that-it-planned-gaza-strike-with-500-civilian-deaths/. Israeli press, carrying the IDF’s statement.