On April 3, 2024, the Israeli-Palestinian outlet +972 Magazine, in partnership with Local Call, published an investigation by journalist Yuval Abraham headlined “Lavender: The AI machine directing Israel’s bombing spree in Gaza.” Based on testimony from six Israeli intelligence officers who said they had first-hand involvement, the report described an AI system that generated lists of suspected militants to be targeted in airstrikes during the war that began in October 2023.
According to the report, Lavender “clocked as many as 37,000 Palestinians as suspected militants - and their homes - for possible air strikes,” and during the first weeks of the war the army “almost completely relied” on its output. One source said human personnel often served “only as a rubber stamp for the machine’s decisions,” devoting “only about 20 seconds” to each target before authorizing a strike, “just to make sure the Lavender-marked target is male.” This was despite knowing the system made errors in “approximately 10 percent of cases,” sometimes flagging people with only a loose connection to militant groups or none at all. The same investigation described a related system, sometimes called “the Gospel,” used to mark buildings and structures.
The claims rest on anonymous testimony and have not been confirmed in primary first-party military documents; the Israeli military disputed parts of the account, including the characterization of Lavender as a system that produces a list of people to kill. The story is treated here as reported investigative journalism, the primary source being the +972 article itself, not as an established factual record of how the systems operated.
The episode matters because it is one of the most detailed public accounts of AI being used to generate targets in an active war, and it gave the abstract debate over “meaningful human control” a concrete and disturbing illustration: a human review measured in seconds, on top of a system with a known error rate. Whatever the precise details, it sharpened the central worry that automation can hollow out human judgment even when a person nominally remains in the loop.