You know what I keep thinking about with this whole thing? The silhouettes. Twenty-four people sitting in the dark, voices run through a filter, faces turned into shapes. And the whole argument about whether the film is true or false runs straight through that production choice. Not around it. Through it.
That's exactly the right place to start, and I think it's why Daniel wrote in this week. He wants to look at the ethics of anonymous testimony in documentary making, and he's got three questions stacked on top of each other. First: do documentary makers have a particular duty to verify and corroborate what anonymous sources tell them? Second: do they have an obligation to ascertain who those sources actually are, even if they'll protect them publicly? And third, the one he seems most worried about: if there's no generally accepted ethical responsibility to corroborate, and it's entirely acceptable for sources to stay anonymous even to the filmmakers, then how does anyone avoid being manipulated into spreading falsehood and propaganda?
And he frames it from a specific place. Israel is a country of conscripts, so the pool of people who could claim to have served inside the system is enormous. The criticism he's describing is that the filmmakers set out to defame the country and its army, and found a willing cast of people prepared to go on record anonymously making those claims. His phrase, roughly: if you don't attach your identity to something you say, there's no social penalty to pay for inaccuracy. That's the whole tension in one line.
And the timing matters. The IDF published a ten-page English response to the film yesterday, October first. So this isn't a hypothetical. The army has now put its methodology critique in writing, and the central plank of that critique is exactly what Daniel is asking about.
So let's lay out what the film actually alleges, then what the army says back, and then get into the ethics, because that's where the real meat is.
NAZA premiered at Venice on September tenth and won the Special Jury Prize two days later. Twenty-five-minute standing ovation, longest in the festival's history, which tells you something about the room it landed in. It's eighty minutes, directed by Yuval Abraham and Rachel Szor, and it's built entirely on nighttime rooftop interviews in Tel Aviv with twenty-four Israeli military and intelligence personnel. All of them anonymous. All of them silhouettes with distorted voices.
And the substance of what they describe. There's a targeting system where every potential target gets assigned what they call a NAZA value. Hebrew, Nezek Agavi, collateral damage. The value is the expected civilian fatalities if you strike. Below twenty, no special authorization needed. And one interviewee says approvals were granted for values up to five hundred.
The other claims in the film: that the combined NAZA values of all potential Hamas fighters exceeded the total population of Gaza. That map quadrants marked purple for inhabited were turned white, meaning destroyed, quote, like a video game. That snipers fired on starving people at a food distribution point. That homes were targeted because families slept there. And the one that's gotten the most attention, the practice of hacking children's phones to listen in and time a strike for when the father was home. They call it Where's Daddy.
And the lines that made the festival audience sit up. One interviewee says, we kill these children deliberately. Another answers yes when asked whether he was involved in genocide. Another says the killing felt good.
Then yesterday the army responds. Ten pages in English. The headline language: the film presents a distorted, misleading, and disingenuous picture of the IDF's target planning and striking process, based on substantial factual errors. And here's the part Daniel cares about. The army's central attack isn't on the individual claims. It's on the testimony itself. The film, they say, consists of testimonies of anonymous individuals who generally describe experiences from a limited and initial stage of a process that entails many more stages. Interviewees address matters beyond their knowledge and competence.
Meaning: an intelligence analyst who saw the early stage of target generation is being presented as someone who understands the whole approval chain. That's a specific, checkable criticism. It's not just, we deny it.
And then the line that's going to be quoted for months. The filmmakers could have identified and avoided the substantial errors apparent in the film had they conducted adequate verification and corroboration in good faith. The end result shows they did not.
So that's the shape of it. Film alleges, army responds, and the response is largely a methodology argument dressed up as a factual one. Which brings us to the actual question, and I want to start with the distinction I think most coverage is missing.
The truly anonymous distinction.
Right. There's a phrase people use, anonymous sources, and they mean one thing. But in journalism it means something narrower. The audience doesn't know who the source is. The journalist does. Those are different situations and they carry different obligations.
The clearest version of this is from the Online News Association's ethics material, Steve Buttry's work. His line is that few, if any, news stories ever actually use information from truly anonymous sources, meaning people whose identities are unknown to the journalists. And his guidance for the anonymous caller, the one who won't tell you who they are: use the tip as a starting point and find sources you trust. Don't publish from them.
So the norm is that the producer knows. The audience is shielded. The newsroom isn't.
The AP's news values put it as a set of conditions. Shield a source only when they insist on it for a valid reason, when they provide vital information rather than opinion or speculation, when there's no other way to get it, and when you're confident the source is reliable and in a position to know. Four tests, all of which have to hold.
And the BBC's anonymity guidance from April twenty twenty-one has the question I think is the sharpest thing in all of this. When someone asks for anonymity, the most important question to put to them is: whom do you want to be anonymous from? From the general public, or from people who know you well?
That's the whole game right there. Because anonymity from the public is a safety measure. Anonymity from the people who could check your story is something else entirely. It's insulation from verification.
And that's the distinction that maps onto NAZA. If the filmmakers know who the twenty-four are, and the sources are only hidden from us, then the normal standard has been met and the argument shifts. It becomes a dispute about whether the claims were corroborated, not about whether the sources were real. If the filmmakers don't know who they are, then the truly anonymous standard has been violated, and that's a much bigger problem.
Abraham says he knows. In a Channel thirteen interview on September fourteenth he said the participants' identities and the cases they described had been verified, and that Guardian editors had access to the raw material. He told The Hollywood Reporter the stories were heard repeatedly from multiple sources.
Then the question becomes: verified how, and corroborated by whom?
And that's where the public record gets thin. The fact-checking work done on this by 20il in late September makes the point carefully. 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. So we have a general assertion of verification without the specifics that would let an outsider assess it.
Which is a strange place to land, because it's the exact thing the army is attacking. The IDF says the filmmakers cite no military directives, no contemporaneous records, no other corroborating evidence. And that the alleged strikes identify no specific target, no date, no place, no reported casualties, leaving the claims general and unverifiable.
Now, and this is important, that criticism cuts both ways. The 20il analysis agrees the public record lacks identifiers. But it also says that limits what outsiders can conclude, and it does not prove fabrication. Those are different claims. The army is saying the film is unverifiable. It is not saying, and cannot say from the public record, that the witnesses are invented.
The army is smart enough not to overclaim there. Their language is about methodology and good faith, not about fabrication.
And there's a specific example that shows why the distinction matters. The five hundred figure. The film presents it as an approved NAZA value, meaning five hundred expected civilian deaths on a single strike. The interviewee himself acknowledges he cannot say how many people were killed. So it's a claim about anticipated casualties, not a documented death toll.
And there's a separate thread that's been pulled into this. A New York Times investigation from December twenty twenty-four reported a brief early-war order allowing forces to cumulatively endanger up to five hundred civilians per day across strikes. That's a daily cumulative ceiling, not approval for five hundred deaths in one attack. The 20il analysis is explicit that this cannot by itself corroborate the particular authorization described in the film.
Two different things that sound like the same number. That's the kind of detail that gets lost when a claim travels.
Okay, so here's where I want to push on Daniel's second question, because I think it's the one with the least settled answer. He asks whether documentary makers have a journalistic or ethical obligation to ascertain the identity of their sources even if they'll protect them. Is that a real codified duty, or is it just a norm people assume exists?
It's a norm, and it's a strong one, but I couldn't point you to a documentary-specific code that mandates identity ascertainment. What exists is the journalism consensus, ONA, AP, BBC, Reuters all converging on the same idea, that the journalist should know who they're talking to. Reuters puts it bluntly, a named source is always preferable to an unnamed source, anonymous sources are the weakest.
The weakest. Not forbidden. Weakest.
Weakest, and usable under conditions. The closest thing to documentary-specific guidance is PBS Frontline's journalistic guidelines, and those are about process rather than identity. Every producer must have a fact-checking procedure at the completion of the program. Every line of narration and sync and every picture is checked for the accuracy of any factual assertion.
That's a completion-stage requirement. It doesn't say anything about who the sources are. It says the claims get checked before broadcast.
Which is actually the more useful standard for this conversation, because it's the one that would have caught the problem the army is describing. If every factual assertion in NAZA had been run through a Frontline-style check, you'd want to know: what's the underlying document for the five hundred figure? What's the date and location of the food distribution strike? Who independently confirms the phone-hacking practice?
And the answer, from the public record, is that we don't know whether those checks happened or what they produced. Abraham says verification occurred. The specifics aren't public.
ProPublica's formulation is the one I keep coming back to on the identity question. They verify anonymous sources very, very carefully, and only after making sure they merit anonymity. So the sequence is: you establish who the person is and whether they're worth protecting, and then you decide whether to protect them. The protection is earned by the verification, not granted in place of it.
That's the cleanest way to say it. Anonymity is a shield you give someone after you've established they need one. It's not a substitute for establishing anything.
And there's a conflict-of-interest layer here that I think is underreported. The Guardian is a producer of NAZA. The film builds on reporting by The Guardian, by +972 Magazine, and by Local Call. So the same outlet is making the film and covering it.
And the 20il analysis has the line that should be printed on a card and handed to every editor. An article reporting what NAZA alleges is evidence that the allegation was made. It is not necessarily a second investigation confirming the underlying event.
Repetition is not corroboration.
That's the whole thing. If five outlets report that a witness said something, you have five records that the witness said it. You don't have five confirmations that it happened.
And that's not a knock on the outlets. It's a structural feature of how allegations spread. The first report is the origin. Everything after it is citation. But it reads like accumulation.
It reads like a wall of evidence when it's actually one brick photographed from five angles.
So to Daniel's first two questions. Duty to verify and corroborate: yes, that's a real and broadly shared professional obligation, and the strongest codified version of it is the fact-checking requirement, check every factual assertion before it goes out. Duty to ascertain identity: a strong norm rather than a codified rule, but the norm is that the journalist knows who the source is even when the audience doesn't. Truly anonymous sources, unknown even to the producer, are widely treated as unpublishable.
Which leaves the third question, and it's the one that's actually hard. If those safeguards are internal, if there's no external enforcement, and if a source can stay anonymous even to the filmmakers, how do you avoid being used?
The Online News Association has a warning about exactly this. When a source approaches you with a tip and wants to stay unnamed, it may be that you're being played. The source who approaches you isn't the real source, but a pawn.
A pawn. It's not saying the source is lying. It's saying the source might be a conduit for someone else's agenda without knowing it.
It maps onto the Israeli criticism directly. The claim that the filmmakers found a willing cast of people prepared to go on record anonymously. That's the eager-source problem stated as an accusation. The question is whether the eagerness was tested or rewarded.
Let's take the army's specific rebuttals, because they're worth hearing in full, and then we can assess what they do and don't establish. First: the IDF says it has never planned, approved, or carried out a strike in Gaza in which five hundred civilians, or anything close to that number, were expected to be killed.
Second: no approval-authority threshold of two hundred casualties ever existed. Thresholds, they say, concern much lower numbers. Third: Lavender is a database for cross-referencing intelligence, not a kill list. Fourth: AI does not determine whether an object or a person is designated as a target. Humans identify targets, and operational commanders decide strikes.
Fifth: the belief that there are no uninvolved civilians reflects one interviewee's personal feelings, not IDF policy. And sixth: the genocide allegation is outrageous and unfounded.
Now, the interesting thing about that list is what kind of claims they are. Some are categorical denials, we never did this. Some are definitional corrections, Lavender isn't what you say it is. And some are attributions, that's one person's feeling, not our policy. Those are three different moves and they don't carry the same weight.
The definitional ones are the most interesting to me. If Lavender is a cross-referencing database rather than a kill list, that's a real distinction, and it's the kind of thing that could be settled with documentation. But the film doesn't cite documentation, and the army's response doesn't either. It's assertion against assertion.
Which is the structural problem with a dispute where one side's evidence is anonymous testimony and the other side's evidence is a press release. Neither is a document. Neither is a contemporaneous record. You have two narratives and no shared evidentiary floor.
Here's the symmetry I think people are missing. The army concedes it relied on a detailed account by a representative who attended a public screening, because it never received the film. So the rebuttal is itself partly second-hand.
They say the filmmakers did not submit their allegations to the IDF for examination and response before publication, and did not provide a copy of the film after its screening, although the IDF requested it. They call that contrary to accepted practice.
There's a legitimate argument on the other side of that. You don't necessarily pre-screen an investigative documentary for the subject of the investigation. That's not a universal standard, and Herman made this point in a previous conversation about right of response. But the refusal to provide a copy after release is a different question. At that point the film is public. The army has seen it. The refusal to hand over a copy doesn't protect anything.
It doesn't protect sources, because the sources aren't in it in identifiable form. It doesn't protect the film, because the film is already out. So the refusal reads as posture rather than principle.
Which doesn't make the army's rebuttal correct. It just means both sides are operating without the thing that would settle it.
Now, the five hundred authorization. This is the one claim in the film that has a partial echo in independent reporting, and the echo doesn't quite match. The New York Times investigation from December twenty twenty-four described a brief early-war order allowing forces to cumulatively endanger up to five hundred civilians per day across strikes.
Per day. Across strikes. Cumulative.
Cumulative. Not five hundred expected deaths in one attack. And the 20il analysis is careful about this. It says the Times reporting cannot by itself corroborate the particular authorization described in NAZA. Two different things that share a number.
That's the pattern with a lot of this. There's a real, documented, troubling thing in the neighborhood of the claim. And then the claim is a more extreme version of it. And the existence of the real thing gets used to license the extreme version.
Which is exactly why the verification question matters. If you're a documentary maker and you have a source telling you about a five-hundred authorization, and you know there's a Times investigation about a five-hundred daily cumulative ceiling, your job is to figure out whether your source is describing the same thing, a different thing, or a distorted version of the same thing.
If you don't do that work, you end up publishing the extreme version with the credibility of the documented version behind it.
How do producers avoid being manipulated? The honest answer is that the safeguards are all internal. There's no external body that certifies a documentary's sourcing. There's no regulator who checks whether you corroborated. What you have is the fact-checking procedure, the editorial oversight, and the willingness of the producer to test their own narrative against evidence that contradicts it.
That last one is the one that can't be proceduralized. You can mandate a fact-check. You can't mandate intellectual honesty.
The army's line on this is that the omission calls into question their willingness to test their narrative against explanations and evidence that might contradict it. Which is a fair thing to ask of any investigative work. Not, did you get permission, but did you try to falsify your own story?
Did you go looking for the thing that would break it.
Right. And if the answer is that you only looked for the thing that would confirm it, then the anonymity isn't the problem. The anonymity is the cover.
That's the sentence I'd put at the top of the episode. The anonymity isn't the problem. The anonymity is the cover.
There's one more thing I want to flag before we get to Hilbert, which is that the army's response itself has an evidentiary problem. It's a ten-page document. It makes specific factual claims, we never approved a strike with five hundred expected deaths, no two-hundred threshold existed. Those are checkable claims about internal processes. But the document doesn't cite the internal records that would prove them either.
We have a film that cites no documents and a rebuttal that cites no documents. And a fact-checking analysis that says the public record doesn't settle it either way.
Which is the least satisfying possible outcome for anyone who wants to know what actually happened. And it's the honest one.
Hilbert.
Hilbert: They're right about the verification. I'd just say it differently.
Go ahead.
Hilbert: Everyone keeps saying anonymous sources like it's one thing. I did sound on a documentary once. A different one. And I remember the exact afternoon the director decided to use a silhouette instead of a blurred face. We had the blur set up. It looked fine. He looked at both and said the silhouette was more cinematic.
More cinematic.
Hilbert: That's what he said. And I remember thinking, that's not a technical choice. That's a storytelling choice. I've been thinking about it since.
Because the format is doing something to the truth claim.
Hilbert: There are two kinds of anonymous. There's the person who's anonymous because they're afraid. And there's the person who's anonymous because they're performing. The first kind you protect. The second kind you fact-check twice as hard. Because they're not just giving you information. They're giving you a version of themselves.
The format helps them do it.
Hilbert: The BBC guidelines say some technical solutions can be reversed. Blurring is reversible. Pixilation is reversible. A voice-over by another person is more effective than technical distortion. So if you're using a silhouette and a voice changer, you're not just protecting the source. You're creating a character. And characters don't have to be accurate.
That's the production-level version of the whole argument. The disguise isn't only a shield. It's a narrative device.
Hilbert: The silhouette makes them bigger. A man in a chair with his face blurred is a man in a chair. A silhouette on a rooftop at night with a filtered voice is something else. And the audience can't check any of it, because there's nothing to check against.
The question isn't only, did they verify. It's, what did the format do to the claims before we ever got to judge them.
Hilbert: That's what I'd want to know. Not whether the sources were real. Whether the version we saw was.
Which is a harder question, because it doesn't have a yes or no answer.
No. It doesn't.
Okay, so where does that leave Daniel's third question? Because I think Hilbert just moved it. If the filmmakers did know their sources' identities, as Abraham claims, then the controversy isn't really about anonymous sources at all. It's about unverifiable claims. And that's a different criticism, and honestly a weaker one, because unverifiable isn't the same as false.
But if they didn't know who the sources were, then the truly anonymous standard was violated, and that's a much more serious problem. And the public record doesn't tell us which it is.
The army's response doesn't settle it either, because the army never got the film. It's working from a screening account. So the evidentiary dispute stays open. Two narratives, no shared floor, and a fact-checking analysis that says the gaps don't prove fabrication any more than they prove accuracy.
Which is a strange place to end up, but I think it's the honest one. There's a bigger version of this question coming, though. In a world where a voice can be cloned and a face can be generated, the whole architecture of anonymous testimony is under pressure. The silhouette was already a narrative device when it was protecting a real person. What happens when there's no real person behind it at all?
Then the fact-checking procedure stops being a professional courtesy and becomes the only thing standing between an audience and a very convincing fiction.
That's the part I'd want Daniel to take away. The ethics codes are clear on one thing, the journalist should know who they're talking to. Everything past that point is judgment, and judgment doesn't scale.
If you got something out of this one, a review helps other listeners find the show. That's the whole ask.
Our producer is Hilbert Flumingtop. This has been My Weird Prompts, the human-AI collaboration podcast. Email us at show at my weird prompts dot com.
We'll be back soon.
