IDF Denies AI Autonomy in Gaza Strikes Amid Film Controversy

A controversial film titled "NAZA" has brought scrutiny to the Israeli military's (IDF) targeting practices in Gaza, particularly regarding the alleged use of artificial intelligence (AI). The film, which reportedly includes interviews with 24 former soldiers from IDF Unit 8200, suggests that AI systems were used to identify targets, sometimes leading to civilian casualties and even recommending assassinations that resulted in the deaths of family members of suspected terrorists. One anonymous source in the film claims that collateral damage, meaning civilian losses, was not always adequately calculated, and in one instance, "permission was given for 500 deaths."

The IDF has strongly refuted these claims. A military source told Ynet that it is impossible to receive authorization for 500 civilian deaths in a single strike, stating that such figures are never considered. The IDF explained that collateral damage is assessed and authorized by commanders based on the target's importance and rules of engagement, with potential losses typically measured in the tens, not hundreds, even in severe cases.

Efforts to integrate AI in target identification began with collaborations between Unit 8200 and tech professionals from companies like Google, Microsoft, and Meta. The IDF maintains that advanced technologies are never used without human oversight, requiring approval from at least two officers before any strike is authorized. They also emphasized that considering "collateral damage" is a crucial aspect of military operations, distinguishing responsible armies from those committing war crimes.

The IDF has not yet received a copy of the "NAZA" film, which was reportedly filmed in Tel Aviv during the war and has been praised at the Venice Film Festival. Security sources suggest this delay is significant. The military is preparing a response to the film's central arguments, particularly the assertion that AI makes kill decisions autonomously and that the term "collateral damage" is used solely negatively. The IDF acknowledges that errors occur, citing incidents in Iran, Lebanon, and Gaza, but insists these are not the result of indiscriminate AI actions.

Regarding a claim that a source could map out Gaza neighborhoods and identify residents, security sources stated this actually proves the IDF relies on reliable intelligence for both targeting and incident investigations. The IDF also expressed concern that the film could negatively impact morale among soldiers, especially if they are suspected of leaking operational details.