NAZA Fact Check: What the Public Evidence Shows

The dispute over NAZA goes beyond its anonymous witnesses. It centers on a contested 500-civilian allegation, unclear AI statistics, and a central question: when does repeated reporting become independent corroboration?

Analysis | September 22, 2026

An anonymous interviewee in NAZA describes approval for a strike despite an estimate of 500 civilian deaths. The Israeli military explicitly denies that such a strike was planned, approved or carried out. The filmmakers maintain that their sources and accounts were verified.

The publicly reported allegation does not identify the target, date, location or authorization record needed to examine that particular case independently.

Israel News Pulse’s claim-by-claim examination of NAZA argues that several of the documentary’s most consequential allegations remain unproven rather than established facts.

That criticism raises a question worth separating from the political reaction to the film: what does the available evidence establish about each allegation, and where does the public record stop?

NAZA, directed by Yuval Abraham and Rachel Szor, presents testimony from 24 people described as Israeli military and intelligence insiders. Their accounts concern targeting systems, surveillance, civilian casualties and alleged unlawful killings. But the number of participants, the circulation of their accounts and the evidence supporting a particular incident are three different things.

The two different 500-civilian claims

The allegation at the centre of the dispute concerns anticipated casualties, not a documented death toll.

In the excerpt reported by Ynet, an interviewee describes a system for estimating how many people could be harmed by attacking a building. Asked about the highest number he remembers being approved for a single strike, he says there was approval for 500, while acknowledging that he cannot say how many people were killed.

On September 14, Ynet reported the IDF’s specific denial: the military said it had never planned, approved or conducted a Gaza strike in which 500 civilians, or anything close to that number, were expected to die.

That leaves two opposing accounts without a publicly identifiable operation through which to resolve them. The denial does not independently disprove the testimony. The testimony does not independently establish the authorization.

A separate figure requires particular care. The New York Times’ December 2024 investigation into Israel’s bombing rules reported a brief early-war order allowing forces to cumulatively endanger up to 500 civilians per day across strikes.

A cumulative daily limit is not the same proposition as approval for 500 civilian deaths in one attack.

The earlier report may be relevant to questions about casualty thresholds and changes in military policy. It cannot, by itself, corroborate the particular authorization described in NAZA.

Establishing that allegation would require evidence connecting the number to a specific decision: an identifiable operation, an authorization record, sufficiently detailed corroborating testimony or other material that establishes what was approved and under what circumstances.

What the 24 witnesses do and do not establish

The film’s interviewees appear with digitally altered faces and voices. That prevents viewers from identifying them, but it does not demonstrate that they are fictitious.

Abraham has also made a more substantial claim than simply asking audiences to trust him. In a September 14 Channel 13 interview, reported by The Times of Israel, he said the participants’ identities and the cases they described had been verified, and that Guardian editors had access to the raw material.

In an interview with The Hollywood Reporter, he further said that the film’s stories and claims were heard repeatedly from multiple sources.

Those statements belong in any account of the controversy. It would therefore be inaccurate to describe every allegation as necessarily supported by only one witness because only one person is quoted publicly.

However, the filmmakers’ public explanations do not provide a claim-by-claim account of how many witnesses corroborated each incident, what their access was or whether they supplied independent information.

The distinction matters. An intelligence analyst may have firsthand knowledge of a database but not of the final authorization for a particular strike. A participant may know what happened within one unit without being able to establish a military-wide policy.

The IDF’s official response to NAZA, dated September 11, challenges precisely this issue. It argues that some reported claims concern strategy beyond junior personnel’s knowledge or procedures in which the described intelligence personnel would not participate.

These are objections requiring examination, not independently demonstrated findings about each interviewee.

The unresolved question is not simply whether the sources are anonymous. It is how their knowledge, access and independence support each specific allegation.

One allegation can generate many articles

The Guardian is a producer of NAZA, and the film builds on reporting published by The Guardian, +972 Magazine and Local Call.

Abraham’s earlier reporting is part of that record. His April 2024 investigation for +972 and Local Call described Lavender and a tracking tool known as “Where’s Daddy?”

This creates a source-tracing requirement, not automatic evidence of misconduct.

An article reporting what NAZA alleges is evidence that the allegation was made. It is not necessarily a second investigation confirming the underlying event. Similarly, a publication’s involvement in producing the film needs to be distinguished from independent reporting about it.

The distinction works in both directions. Journalists within the same organization can gather additional evidence. A different publication can merely repeat an existing account. The masthead alone does not establish independence.

For the 500-civilian allegation, for example, multiple reports describe the interviewee’s statement and the IDF’s denial. Their repetition does not resolve the dispute.

Nor does this source trail establish that all the film’s interviewees are the same people who spoke to Abraham previously. The extent of that overlap has not been publicly disclosed.

The useful question is what additional evidence each publication contributes, not how many publications carry the story.

The AI numbers need definitions, not just repetition

The numerical claims surrounding NAZA illustrate another problem: figures that appear precise can remain difficult to interpret.

Public reviews have described a 15% figure in different ways, including an acceptable error threshold and a measured rate of incorrect results. Those are not interchangeable descriptions.

A permitted error rate describes a tolerance. An observed error rate describes a measurement. Neither can be evaluated properly without knowing what counts as an error, which cases were examined and when the assessment occurred.

Abraham’s April 2024 +972 investigation reported approximately 90% accuracy, based on sources describing a manual check of several hundred randomly selected cases. The underlying audit was not published, but the reporting did describe a sampling procedure.

The difference between that account and a later 15% figure does not, by itself, prove a contradiction. The figures could concern different periods, samples, systems or definitions. Establishing a contradiction requires showing that they measure the same thing under comparable conditions.

The target-count discussion needs the same discipline.

Some commentary attributes up to 70,000 human targets to the film. The earlier +972 investigation reported approximately 37,000 people flagged as suspected militants during an early stage of the war.

Those figures cannot be meaningfully compared without their dates and counting criteria. A count of flagged individuals, a database of devices, a list of approved targets and a casualty total describe different populations.

The available descriptions do not justify treating those categories as interchangeable. Equally, the appearance of similar numbers in different contexts does not prove that one was mistakenly copied from the other.

The incidents that remain difficult to identify

Several reported accounts concern individual victims or particular shootings. Their evidentiary problem is different from the numerical disputes: the published descriptions lack enough identifying detail to connect them to independently documented events.

Reported allegation	What the published accounts do not identify	What remains unresolved
A 12-year-old girl’s phone was used to locate her father before a bombing	The family, date, location and associated strike record	Whether a particular documented attack matches the account
A teenage boy was killed after entering a restricted area	The boy, location, date and relevant military unit	The circumstances of the alleged killing
Soldiers shot unarmed people approaching an aid site	A specific site, date, unit and victims tied to the testimony	Which incident the account describes and how it can be corroborated

The Hollywood Reporter describes an interviewee recounting the use of the girl’s phone and a subsequent bombing. That establishes what the interviewee is reported to have said; it does not supply the identifiers needed to reconstruct the incident independently.

The same distinction applies to the accounts involving restricted areas and aid distribution.

Evidence of other shootings can establish relevant context or a broader pattern. It does not automatically identify the incident described by a particular witness.

Conversely, missing public identifiers do not prove that the event never happened. They limit the conclusions an outside observer can draw from the published account.

The term “annihilation zone” also needs attribution. Public coverage uses different descriptions, including “no-go zones.” Whether a phrase was an official designation, an informal expression or a journalist’s translation cannot be settled by repeating it as military terminology.

A surveillance alert is not the entire targeting process

“Where’s Daddy?” is not a claim that appeared for the first time with NAZA. The tool’s reported function was described in the April 2024 +972 investigation and discussed subsequently by Human Rights Watch.

Human Rights Watch described a tool intended to indicate when a targeted person was at a particular location, often a presumed family home. Its account also expressly acknowledged that it could not document when and where the military used the digital tools under discussion.

That supports examining the reported surveillance function. It does not independently reconstruct every subsequent operational decision.

Several questions remain separate: whether a device entered a location, whether the intended person carried it, whether that person was a lawful target, what civilian presence was anticipated and who authorized an attack.

Reuters’ reporting on NAZA describes officers saying that words equivalent to “daddy’s home,” intercepted through a child’s phone, could be sufficient to trigger an airstrike. The publicly reported formulation does not identify a particular intercepted message and corresponding strike record.

The IDF says humans select targets and approve strikes, rather than artificial intelligence systems.

That response does not, on its own, answer whether human review was meaningful or whether an attack was lawful. But the existence of surveillance technology does not, on its own, establish a fully automated chain from an intercepted phrase to a bombing.

Both propositions require evidence about how the systems were actually used.

Civilian-harm estimates do not settle questions of legality

The documentary’s published synopsis uses the phrase “calculated mass killing.” Evaluating that framing requires separating calculation, knowledge, intent and legality.

Estimating anticipated civilian harm does not by itself establish that civilians were the intended target. It also does not establish that an attack complied with the law.

The ICRC’s explanation of proportionality concerns whether expected incidental civilian harm would be excessive in relation to the concrete and direct military advantage anticipated. Feasible precautions and the identification of a lawful target are separate requirements.

An attack can therefore raise serious legal questions even without proof that civilians were deliberately selected as its objective. Human authorization, a casualty estimate or an asserted military target does not automatically resolve those questions.

The reverse inference is also insufficient. Showing that a commander anticipated civilian deaths does not alone establish an institutional purpose of killing civilians.

Such a conclusion requires evidence connecting particular decisions, practices and knowledge to the broader claim. Written orders can be important, but the absence of a published written order cannot make testimony or evidence of conduct irrelevant.

This is where the distinction between an allegation and a demonstrated conclusion matters most. Accounts of individual attacks, descriptions of operational practices and claims about an institution’s purpose need to be assessed at their respective levels, rather than treated as substitutes for one another.

Israel’s security concerns are not the same as proven damage

The controversy has produced documented security and legal responses.

On September 14, IDF Chief of Staff Eyal Zamir directed an examination of information-security issues and possible legal action relating to the film.

However, The Times of Israel’s September 16 report on the security and legal review described significant limits. Citing Haaretz, it reported that the military’s information-security department said no criminal investigation had been opened at that point and that it lacked the film’s contents and participants’ identities.

The same report said the military advocate general had advised that criminal charges would require security violations or leaks.

A request to investigate is not a finding that classified material was disclosed. It is also not a finding that an operational capability was compromised. Those announcements do not establish either conclusion.

There is a documented international use of the film. Ynet reported on September 15 that Hamas urged the International Criminal Court, United Nations bodies and human-rights organizations to use its testimony in accountability proceedings.

That establishes Hamas’s public attempt to use NAZA in its advocacy against Israel. It does not establish that Hamas supplied the material, that a court accepted the testimony or that the allegations were verified.

The consequences for Jewish communities abroad require a similarly specific evidentiary connection. Neither Hamas’s statement nor the announcement of an Israeli security review demonstrates that the documentary caused threats or attacks against Jews outside Israel. Claims of such harm require identifiable incidents and evidence linking them to the film.

Threats directed at the filmmakers are a separate, documented consequence. On September 18, the Committee to Protect Journalists and ten other press-freedom organizations called for their protection, citing threats and intimidation involving the journalists, their families and associated media outlets.

Those threats do not establish the accuracy of NAZA. They are also not a substitute for answering its allegations.

What remains established, disputed and unresolved

NAZA has brought forward testimony that its filmmakers say was authenticated and corroborated. Some of the systems and practices discussed in its coverage have a reporting history extending well before the documentary.

Other claims remain unresolved in the public record: the particular 500-civilian authorization, the meaning and basis of certain AI figures, and the identities of incidents described without dates or locations.

The IDF’s denials are part of that record. They are not independent verification. The filmmakers’ assurances are also part of it. They do not give the public access to every underlying source or document.

None of these gaps, by itself, establishes fabricated witnesses or invented testimony. Nor does the repetition of an allegation turn it into an independently established event.

The distinction is between what a source alleges, what additional evidence corroborates and what remains disputed. That distinction applies to NAZA’s accusations, the military’s responses and claims about the film’s consequences alike.