Minhaz Samad Chowdhury

International Law & Human Rights Analyst
Editor, HR Defender

11 Sept 2026
Venice 83 | Documentary

The Architecture of Deniability: NAZA and the Bureaucratization of Mass Killing

At Venice 83, a film does not ask you to watch Gaza burn. It asks you to watch how the burning was approved, coded, filed, and made to look inevitable.

There is a moment in NAZA—screened out of competition at Venice but already the most discussed documentary of the Lido—where a former intelligence officer describes printing a target list. He does not describe the targets. He describes the printer. How it jammed. How someone had to shake the tray. The banality is the point. The film, directed with forensic restraint, argues that mass killing in Gaza has not been a breakdown of systems, but a triumph of them.

NAZA (an Arabic term the filmmakers intentionally leave untranslated, evoking both displacement and stripping) assembles 24 insiders from across the military-intelligence-political chain. No archival shock montage. No narrator. Just testimony, documents, and the infrastructure that connects them. Its thesis is stark: we are witnessing not a war without rules, but a war where the rules have been engineered to permit everything.

I. The Film and Its Method: 24 Insiders Break the Seal

The power of NAZA lies in its sourcing. These are not dissident academics or humanitarian workers—though their analysis has long warned of this trajectory. The film’s subjects are former officers in target generation, legal advisors who wrote proportionality memos, operations planners who managed “strike banks,” and tech contractors who maintained the AI-assisted recommendation systems. Their faces are often shadowed, their voices altered, but their documents are not.

The method is deliberately bureaucratic. One insider walks through a PowerPoint slide deck used for morning target approval: green, yellow, red. Another shows an email chain where a village is reclassified from “civilian infrastructure” to “Hamas-affiliated support infrastructure” with a single checkbox, instantly changing the permissible civilian harm threshold. The film never says “genocide” in voiceover. It shows you the administrative instrument that makes the question unavoidable.

“We stopped debating whether we could strike. The system told us we already had. The lawyer’s job was to write why that was okay afterward.”

— Former Military Legal Advisor, NAZA

Venice audiences have called it chilling because it refuses catharsis. There is no climactic revelation of a secret order. The revelation is that there was no need for a secret order. The order was the workflow itself.

II. The Machine: AI Targeting and the Production of Casualties

The second act of NAZA is its most technical and most devastating. It deconstructs the AI targeting ecosystem—systems reported in investigative journalism as Lavender, Gospel, and Where’s Daddy—not as science fiction, but as labor-saving devices for mass target creation.

One engineer explains it plainly: traditional targeting required 20-30 hours of analyst work per target to establish positive identification and pattern of life. The new system produces 100-200 targets per day by lowering confidence thresholds to as low as 10-15%, and by treating behavioral proxies—changing SIM cards, gathering in certain buildings, living in certain neighborhoods—as affiliation.

The Core Mechanism

When you automate target generation but keep human approval as a 20-second rubber stamp, you have not kept a human in the loop. You have kept a human as the liability shield. The machine produces scale; the human produces deniability.

Crucially, NAZA documents how “Where’s Daddy”-type logic inverted International Humanitarian Law. Instead of tracking fighters to strike them away from families, the system flags when targets return home—precisely because the family home is where collateral data is most predictable, and where strike authorization is fastest.

The algorithm doesn’t know what a child is. It knows what a heat signature cluster is. And someone decided a cluster under 8 people is acceptable if the target is above a certain rank. That decision was typed in a spreadsheet cell.

Under IHL, precaution and distinction are not suggestions. They are obligations to take constant care. NAZA shows how automation has been used to systematically lower the care, then document the lowered care as compliance.

III. Executive Design: How War Becomes Workflow

The film’s title—Architecture—becomes literal here. A former operations center designer sketches the room for the Venice audience: three screens per station, a queue, a timer, a “throughput” KPI. Strikes per shift. The language is not military. It is logistical.

This is the bureaucratization Hannah Arendt warned of, updated for Jira boards. Responsibility is distributed so thinly that no single person feels they killed anyone. The analyst who tags a building, the lawyer who sets the collateral threshold, the commander who clicks approve in under a minute, the pilot who never sees the building—each can plausibly claim they were only doing a small, lawful part.

One of the most damning documents shown is a standard operating procedure that defines “post-strike assessment” as complete upon “munitions impact confirmation,” not civilian impact assessment. Death is not counted because it was never assigned to be counted by that desk.

We built a system where saying no took more paperwork than saying yes. In any bureaucracy, that means yes becomes the default.— Former Targeting Officer

IV. The Language of Evasion: Euphemism as Doctrine

NAZA devotes an entire chapter to lexicon. The filmmakers overlay official statements—“precision strikes,” “knocking on the roof,” “evacuation corridors,” “pressure cooker procedures”—against the insider definitions.

“Roof knocking,” we learn, was logged in the system not as a warning, but as a strike itself, allowing the next, lethal strike to be classified as a “secondary strike on already warned structure.” “Human shields” is described by a legal advisor as a “status-laundering term”—once applied, it reclassifies every person in a building as voluntary shield, reducing their legal protection to zero.

The most insidious euphemism is “collateral.” In the film, a linguist consultant notes that collateral in military English does not mean accidental. In the system’s coding, it means “pre-calculated and budgeted.” Families are not collateral damage. They are collateral allowance.

Euphemism Log — From Film Evidence

Power TargetCivilian infrastructure whose destruction pressures population

Pressure PotWhole neighborhood designated for systematic destruction

Legitimization CellUnit whose task is to produce legal justification post-facto

V. What Must Follow: Policy Imperatives Beyond Outrage

NAZA does not end with a call for a ceasefire—though it documents why one is legally required. It ends with a blueprint for accountability that directly targets the architecture itself.

For legal accountability, the film argues:

1. The ICC must investigate the system, not just incidents. Under Article 28 of the Rome Statute (command responsibility) and Article 25 (co-perpetration through organizational control), the designers of target-generation thresholds are not advisors; they are co-authors of the attack pattern.
2. AI complicity needs a legal standard. States supplying cloud infrastructure, model training, or data fusion for systems where civilian harm thresholds are pre-set to 15-20+ per low-ranking target cannot claim dual-use neutrality. NAZA cites the standard from the ICJ Bosnia case: if you know and you assist, you bear risk.
3. Euphemism must be pierced in court. Terms like “power targets” and “pressure” have no IHL basis. Prosecutors should treat them as evidence of indiscriminate attack policy under Article 8(2)(b)(i) and (iv).

For states and platforms:

• Mandate human-level substantive review time minima (minimum 45 minutes per lethal target) and ban batch approval. Speed is not efficiency when lives are at stake; it is recklessness.
• Require independent logging of civilian harm that cannot be marked complete by the attacking force alone.
• Impose a moratorium on AI target recommendation systems operating below 95% confidence in densely populated areas, as a precautionary measure under IHL Article 57.

The Venice Question

“If a bureaucracy can make mass killing feel like inbox management, what does that make us who receive its press releases and call it defence?”

The audience in the Sala Grande reportedly sat in silence for nearly a minute after credits. Not from shock—this war has exhausted shock—but from recognition. NAZA does not show us something alien. It shows us the endpoint of a logic we have all been asked to accept: that technology makes killing cleaner, that procedure makes it lawful, and that complexity makes it no one’s fault. The architecture of deniability is built precisely so that everyone can deny they built it.

SOURCE NOTE: This review is based on the Venice 83 press screening of NAZA (2026, 112 min). Testimonies cited are from the film. Legal analysis reflects the author’s interpretation of IHL and Rome Statute provisions. Film documents referenced include redacted SOPs and target approval screenshots presented on screen.

First published on hr-defender.blogspot.com — Human Rights, International Law & Accountability.