American AI helped choose who died in Gaza

American generative AI is named in a new film as part of how Palestinian civilians were killed in Gaza.

American generative AI is named in a new film as part of how Palestinian civilians were killed in Gaza. That is the hard fact at the center of NAZA, the latest documentary from the No Other Land team, Israeli directors Yuval Abraham and Rachel Szor. It lands like a stone in your lap. You look down. You do not move for a second.

The film runs about 80 minutes and it is built from interviews with 24 Israeli intelligence officers and soldiers. They speak at night on rooftops in Tel Aviv. Their faces stay in shadow. Their voices do not. They describe a system that turned neighborhoods into data, then into targets, then into rubble. They use a Hebrew acronym, nezek agavi, shortened to NAZA. It means collateral damage. In their telling, it also meant a number you could plan around before a strike.

Here is the detour I keep coming back to, because it makes the rest feel colder. The title is not a slogan. It is a ledger line. A value attached to a building before the bomb falls. The film says that value could be 20 civilians for a low-ranking operative, or far more for someone higher up the chain. It says an AI system called Lavender helped rank suspects, and another tool called Gospel generated target lists. It says human officers made the final call, but the call often followed the machine’s suggestion.

Then the story comes back to the point I cannot shake. The directors say American technology enabled parts of this. Not just any tech. Generative AI. Cloud services. The film points to American companies as part of the stack that made mass targeting possible at speed. That claim sits beside testimony that entire apartment blocks were hit to kill one person, that strikes waited for nightfall when families were home, that children saying “daddy’s home” on a tapped phone became part of the timing.

You can argue about the numbers. You should. The Israel Defense Forces rejected the film’s core claims in a detailed statement. They said no strike was planned with an expected death toll near 500 civilians. They said AI never selected targets on its own. They said human officers always make the final decision and that binding rules require distinguishing terrorists from uninvolved civilians. They called the film distorted and misleading. They also noted that major news outlets have reported the IDF used AI-driven systems in Gaza, and that safeguards were loosened at times.

So what is documented, and what is still a claim? The documentary documents 24 insiders describing their roles in targeting, surveillance, and strike planning. It documents their use of the term NAZA for expected civilian casualties. It documents their account of AI tools like Lavender and Gospel feeding into the process. It documents their claim that American cloud and AI services were part of the infrastructure. What remains unproven in the film itself is independent verification of each specific strike, each ratio, and the exact role of each American system in each case. The IDF denies the worst of it. The filmmakers say the testimony is the evidence.

This is where the room gets quiet. Not because the film is loud. It is not. The voices are calm. The details are plain. That is what makes it land. You hear a person say they used AI to help decide who to kill, and you realize the future of war is not a robot with a gun. It is a spreadsheet with a confidence score, fed by phone data and surveillance, reviewed by a tired officer at 3 a.m., approved by a commander who trusts the system more than the fog outside.

The wider picture is already in the news. Rights groups have accused Google, Amazon, and Microsoft of enabling Israel’s AI-driven targeting in Gaza. Microsoft has said it provides cloud and AI services to Israel’s Defense Ministry but denies its tools were used to harm civilians. The Center for Constitutional Rights has warned that Google’s services could expose the company to legal liability. The FTC has opened an investigation into OpenAI, Anthropic, and other AI firms over safety risks, though that probe is not specific to Gaza. None of this proves the film’s every claim. It does show that the question is not fringe anymore.

I do not know what to do with the feeling that a tool built to write poems and draft emails also helped choose buildings to level. I only know the feeling does not leave when the credits roll. It follows you out into the street. It sits with you while you check your phone. It makes you wonder how many other systems, sold as efficiency and safety, are quietly rewriting the rules of who counts as a person when the math gets done.

The film does not offer a clean lesson. It offers testimony. It offers a word, NAZA, that turns a life into a line item. It offers a claim that American AI is part of the chain that led to Palestinian civilians dying in Gaza. It offers a denial from the IDF that the worst of it happened as described. It offers a future where war is faster, wider, and easier to outsource to a model that does not sleep.

You can argue the details. You should. But you cannot argue the shape of the thing. Machines that learn from us are now learning how we kill. And somewhere in that training set is a number for how many of us we are willing to lose.