10 Questions NAZA Raises About Targeting
NAZA is a starting point for specific questions. Instead of treating the documentary as a single proposition to accept or reject, these ten questions separate intelligence, automation, command approval, law and civilian-harm estimation.
1. What does an intelligence “target” mean at each stage?
A person flagged by an analytical system, a developed target file and an operationally approved strike are not necessarily the same thing.
2. What role did Lavender actually play?
Was it a database, a ranking tool, a recommendation system or something more? The film and IDF use different descriptions that should be compared precisely.
3. Could AI ever approve a strike?
The IDF says no. The question then becomes how much weight human reviewers placed on AI-assisted outputs.
4. What was the known error rate?
Any system used at scale needs a way to measure false positives and uncertainty.
5. What happened after intelligence nomination?
Who checked the target, what other intelligence was required, and what operational or legal review followed?
6. How was expected civilian harm calculated?
What data fed the estimate, and how close were estimates to actual results?
7. Were there fixed thresholds?
The film includes allegations about numerical tolerance. The IDF denies binding automatic thresholds of the kind described.
8. Were rules different for senior and junior targets?
Higher military value can affect proportionality analysis, but that is not the same as a universal casualty quota.
9. How were warnings and evacuation orders treated?
An evacuation warning does not make everyone who remains a combatant. The IDF says its rules did not treat warnings as permission to ignore civilians who stayed.
10. What evidence would resolve the dispute?
Operational directives, strike files, contemporaneous legal reviews, source identities and complete after-action data would be far more decisive than rhetoric from either side.