Building on years of investigative reporting, Yuval Abraham and Rachel Szor’s documentary NAZA offers a rare glimpse into the socio-technical systems behind Israel’s targeting campaign in Gaza. Drawing on reporting previously published by +972 Magazine, Local Call, and The Guardian, the film uses 24 anonymous interviews with former soldiers and intelligence personnel to explore how AI-assisted systems accelerated target production, normalized mass killing, and reshaped understandings of responsibility for violence.
The film takes its name from the term used by the Israeli military for collateral damage. Former personnel describe acceptable collateral damage thresholds of 20 civilians per target, while some targets were reportedly approved despite expectations that 200 or even 500 civilians would be killed. The term “NAZA” sanitizes a cruel and unusual targeting practice: people potentially associated with Hamas (either in the present or in the past) are killed in their homes, along with their entire families. Applying the term “collateral damage” to the resulting deaths stretches the term to its breaking point – their deaths can hardly still be understood as unintended or accidental.
Much of the film is devoted to the software that makes this practice possible and, for those who operate it, palatable. Central to this process was Lavender, an AI system that identified 70.000 potential Hamas “members” based on phone-derived data. According to the IDF itself, interviewees describe, an error rate of 15% was deemed acceptable. Although it is tempting to criticize this staggering willingness to accept erroneous killing, the system would be deeply problematic even if it were a 100% accurate, as it marks people for killing on loose inferences from phone data, without establishing in any way that the people to be killed are involved in military activities.
Lavender is accompanied by the system Where’s Daddy? which allows military personnel to surveil the phones of the immediate family of the “suspect” for indications that the “suspect” is home. Once a daughter texts her mom that “Daddy is home”, an alert goes out and the house is struck. The result is one of the documentary’s central paradoxes: unprecedented intimacy existing alongside profound detachment. Operators often knew more about the lives of those they were charged with killing than ever before, yet this knowledge tends to facilitate these killings rather than prevent them: listening in on the phone call of a grieving sister serves to confirm the target was successfully killed.
Perhaps the film’s strongest contribution is its exploration of responsibility. ‘If I hadn’t been there, someone else would have done it,’ says one of Lavender’s developers. Responsibility, he argues, lies with the person who drops the bomb. Similar reasoning appears throughout the documentary. Developers build the tools. Analysts review the targets. Commanders approve the strikes. Pilots release the weapons. Responsibility becomes fragmented across a vast socio-technical apparatus: the “targets” are killed by everyone and by no one in particular.
This diffusion of responsibility is also reflected in the soldiers’ own accounts. Several describe themselves as progressive, open-minded, and working in a unit free from “Arab-haters”. Yet when confronted with the consequences of their actions, many struggle to explain how individuals who consider themselves moral became participants in a campaign that some of them explicitly describe as genocidal. One interviewee even reflects on his visit to Poland: “I still see those Nazi’s as evil. But what does that mean I am?”. The film repeatedly exposes a dissonance between self-perception and action, between acknowledging immense suffering and denying personal responsibility for it.
Importantly, NAZA demonstrates that the technologies themselves are insufficient to explain the violence they enabled. Following October 7, interviewees describe a political and social atmosphere in which the idea that “no one in Gaza was innocent” became increasingly accepted. Although AI enabled the mass violence, the policy of destruction existed prior to it. In some cases, algorithms helped indicate which buildings had been left by 75% or more of their residents. These buildings were then bombed with the remaining inhabitants in them, to facilitate the displacement of the civilian population to the south. This policy however was not reliant on the algorithm: other units moved through the north of Gaza going from house to house, burning them down as fast as possible. Ostensibly, the algorithmic mediation of this mass destruction provides some distance, but the burning of houses is described with the same (if not more) detachment than the mass bombing of homes is.
The documentary is also visually effective in communicating this argument. Interviews are conducted on rooftops overlooking the illuminated skyline of Tel Aviv. City lights, restaurants, offices, and apartment buildings provide the backdrop as participants discuss surveillance, targeting, and destruction in Gaza. The contrast becomes even more striking as the film juxtaposes these images with scenes of devastation. The message is clear: the people operating these systems are not distant monsters but ordinary individuals embedded within ordinary lives.
Ultimately, NAZA is less a film about artificial intelligence than about the social and political conditions that allow AI to facilitate mass violence. Its most troubling insight is that genocide is not automated. It remains a profoundly human project. As Charli Carpenter noted in a 2024 piece: the real ‘killer robots’ are already here, and they’re us. What technologies such as Lavender and Where’s Daddy? offer is not autonomy, but scale, distance and deniability. The effects of the distancing that acronyms, policies and algorithms achieve is felt in the documentary itself. Only in the last scene, that shows the devastating, embodied reality of a single strike the viewer is truly able to grasp the scale of the destruction and suffering these systems facilitate. For us as RAW researchers, the film is also a reminder not to let acronyms and the technical detail of algorithms make us lose sight of what they produce: the radical undoing of life.
Martine Jaarsma is a PhD researcher at the University of Antwerp. She holds a master degree in Conflict Studies and Human Rights as well as a master in Public International Law, both from Utrecht University. In her interdisciplinary research she examines how Palantir’s Maven for NATO (re)shapes the application of the principle of distinction in aerial targeting.
Milutin Đurašković is a PhD Candidate at the University of Antwerp and Utrecht University, and Project Officer for the Realities of Algorithmic Warfare program. He holds a masters degree in Conflict and Development Studies from Ghent University. His current research interrogates how algorithmic technologies produce new forms of violence and civilian harm in Gaza and Ukraine.
