The IDF has been using a new AI-based system that assisted in identifying about 37,000 potential targets based on their connection to Hamas, the UK Guardian reported on Thursday. According to report, the IDF authorized the killing of a significant number of Palestinian civilians, especially in the early weeks and months of the war.
The sources provided a rare glimps into firsthand experiences using the system and said it has raised a series of legal and moral questions and changed the relationship between soldiers and technology.
"It's unprecedented," said an intelligence officer who used the AI based Lavender technology which he said was more reliable than soldier on the ground. "We all lost people on October 7. The machine operates objectively, making it easier on us." he said.
The six officers interviewed for the investigative report said that the technology developed by the military Intelligence Directorate, played a central role in the conflict. The system processes vast amounts of data to quickly identify potential low-ranking Hamas operatives for targeting.
Sources said that before the ground incursion, the IDF "allocated approved numbers" for the estimated number of civilians who might be killed. Two sources claimed that during the early weeks of the war, they were "allowed" to kill 15 or 20 civilians in airstrikes for every five "targets." These attacks were typically carried out using unguided munitions, called "dumb bombs," which on several occasions destroyed homes and killed their occupants.
"Contrary to Hamas, the IDF is committed to international law and acts accordingly," the Guardian said. "As such, it directs its strikes only towards military targets and military operatives and carries out strikes in accordance with the rules of proportionality and precautions in attacks. Exceptional incidents undergo thorough examinations and investigations." Since October 7, the model for approving attacks on human targets has been dramatically upgraded due to the escalation of IDF bombings in Gaza.
"They were always pressuring us, to provide more targets because we must dismantle Hamas, no matter the cost. Whatever is possible, bomb it," said an intelligence officer. To meet this demand, the IDF heavily relied on the "Lavender" system to create a database of tens of thousands of individuals marked as Islamic Jihad or Hamas operatives.
The sources said that after random sampling and confirmation of its predictions, the intelligence unit concluded that "Lavender" achieved a 90% accuracy rate, leading the IDF to approve its extensive use. The system, as described by The Guardian.
Last week, three anonymous Israeli intelligence officials confirmed to The New York Times that the security apparatus operates facial recognition technology in Gaza to locate wanted individuals. The extensive article featured the story of Musab Abu Toha (31), a Palestinian poet residing in the Gaza Strip, who claimed that about a month after the start of the war, he was interrogated by IDF forces who blindfolded him and led him to interrogation. "I had no idea how they could know my name," he said. Later, Abu Toha was identified as a suspect through artificial intelligence facial recognition.
4 View gallery
IDF identifies targets on the grounds
( Photo: IDF Spokesperson's Unit, REUTERS/Dawoud Abu Alkas)
The American publication noted that the facial recognition software, operated by Israeli intelligence, relies on technology from the Israeli company Corsight, which assisted hospitals in the wake of the Hamas massacre. According to Israeli intelligence officials, the software also utilizes Google Images, identifying faces from large crowds and unmanned aerial vehicle photos.