A look at Israel's AI-generated 'mass assassination factory' in Gaza

The two-month Israeli military operation has caused more damage than the years-long carpet bombing of German cities during World War II.

Militaries worldwide will be watching closely and learning lessons from Habsora's application. Only time will tell what those lessons will be.
Axel Rangel Garcia
Militaries worldwide will be watching closely and learning lessons from Habsora's application. Only time will tell what those lessons will be.

A look at Israel's AI-generated 'mass assassination factory' in Gaza

Since Hamas attacked southern Israel on 7 October, the subsequent Israeli military campaign on Gaza has killed more than 18,000 Palestinians and maimed more than 50,000.

In terms of scale, scope and destructive effect, the two-month Israeli military operation has caused as much — if not more — damage than the years-long carpet bombing of German cities during World War II.

Satellite analysis of Gaza indicates that as much as 70% of all buildings in northern Gaza and 60% of residential buildings across all of the Gaza Strip have been damaged or destroyed.

The Israeli military has made no secret of its intent in Gaza, with Air Force Chief of Staff Omer Tishler asserting that “Operation Iron Swords” involved “relentless, around-the-clock air strikes” conducted “on a large scale, and not in a surgical manner.”

In the first five days of Israeli bombing, at least 6,000 bombs were dropped on Gaza, with a combined explosive weight of 4,000 tons. After 35 days, 15,000 separate targets had been struck – more than the 12,600 hit in all of Israel’s 92 days of four wars in Gaza in 2008, 2012, 2014 and 2021.

The enormity of the Israeli military campaign has given rise to claims that the Israeli army has been acting indiscriminately, particularly after the Israeli army spokesman Daniel Hagari said in the early days of the campaign that the "emphasis was on damage and not on accuracy."

In reality, the scale of destruction and death is due to a dramatic increase in Israeli-identified targets – for which artificial intelligence (AI) has been responsible.

In terms of scale, scope and destructive effect, the two-month Israeli military operation has caused as much — if not more — damage than the years-long carpet bombing of German cities during World War II.

AFP
This picture, taken on October 11, 2023, shows an aerial view of buildings destroyed by Israeli air strikes in the Jabalia camp for Palestinian refugees in Gaza City.

Secretive AI unit

Since 2019, Israel has steadily expanded a secretive artificial intelligence unit (the Targets Administrative Division), which relies upon a sophisticated AI platform called Habsora, or The Gospel.

Utilising human, geospatial, signals, and open-source intelligence, as well as aerial surveillance, Habsora has reportedly developed a database of more than 40,000 targets across Gaza it considers to be associated with Hamas or other militant groups.

The AI system was first tested in the May 2021 war, when Habsora was first used — sometimes in conjunction with AI-driven drone swarms collecting real-time intelligence to identify new targets.

According to the then Israeli Minister of Defence Aviv Kochavi, the AI system identified an average of 100 targets per day in the 2021 war, compared to the army's previous non-AI capacity to generate 50 militant targets across Gaza per year.

The effect of AI was apparent then, and since 7 October 2023, it has facilitated a dramatic expansion in the Israeli army's perceived ability to attack Hamas. It alone is likely why Israel's leadership felt confident enough to project the "elimination" or "eradication" of Hamas as the objective of Operation Iron Swords.

According to reporting within Israel and, in particular, a groundbreaking investigation by 972 Magazine, Habsora's database contains four categories of targets: "tactical" (militant cells, weapons caches, rocket launchers, local militant command posts etc.), "underground" (tunnels), "power targets" (high rise apartment blocks, government offices, banks and public buildings), and "homes" (buildings containing the residences of Hamas fighters of any rank).

In the first five days of Israel's assault, 6,000 bombs dropped on Gaza had hit AI-generated "power targets" in an attempt to "create a shock" among Palestinian civilians in Gaza. 

These included the University of Gaza, the Palestinian Bar Association, several United Nations buildings, the Ministries of Culture, Economy, and Telecommunications, and dozens of residential tower blocks.

AFP
Palestinians drive a motorcycle past the destroyed building of the Islamic University in Gaza City on November 26, 2023.

Beyond its reliance on typical intelligence sources, the AI system's sophisticated machine learning also perpetually monitors the movement of suspect individuals and groups and analyses their behaviour and patterns.

As a militant group commanding at least 30,000 fighters, the vast majority of Habsora data is therefore likely associated not with militant leaders but with the more visible foot soldiers and their movement to and from their homes – thereby explaining the very high level of targeting of residential buildings.

After two months, the Israeli army claims to have killed as many as 5,000 fighters, but it is believed that only approximately a dozen mid-level or senior Hamas commanders have actually fallen.

In the first five days of Israel's assault, 6,000 bombs dropped on Gaza had hit AI-generated "power targets", including the University of Gaza, several UN buildings, and dozens of residential tower blocks, among other targets.

Effectiveness remains to be seen

Given the fact that the 7 October attack exacted one of the highest death tolls in Israel's history, its response was always guaranteed to be harsh, but it is clear that AI has been instrumental in facilitating what has been an extraordinarily ruthless military campaign.

How effective that will prove to have been – both in the immediate and long-term – remains to be seen, but the mass civilian death toll, humanitarian catastrophe, and infrastructural devastation is clear.

It is hard to envision those effects alone proving to support any semblance of stability for Israel, which raises the central question: might AI's otherwise extraordinary capabilities create a false sense of security and confidence?

Five months before 7 October, the Director General of the Israeli army, Eyal Zamir, claimed the Israeli security establishment was becoming an AI "superpower."

However, despite its world-leading integration of AI, facial recognition and surveillance capabilities vis-à-vis Palestinian territories, Hamas still invaded Israel with little more than commercially available drones, a few paragliders, and fighters equipped with explosives.

AI neither predicted, detected or prevented the attack itself.

Israel's response to 7 October was always guaranteed to be harsh, but it is clear that AI has been instrumental in facilitating what has been an extraordinarily ruthless military campaign.

'Mass assasination factory'

The ease with which AI can identify purportedly viable targets also raises questions. According to a former Israeli intelligence officer, Habsora has merely helped establish a "mass assassination factory" where every chosen target is assigned a "score card" colour-coded with red, yellow or green lights and in which there is a natural "emphasis on quantity, not on quality."

Does AI take into consideration the long-term effects of recommending that every apartment or house linked to Hamas foot soldiers be bombed or the long-term security implications of rendering Gaza a strip of rubble?

All new technologies take time to find their place within broader structures, mechanisms and best practices, but AI appears to have been the primary driver behind a military campaign in Gaza that will unquestionably have profound implications for years to come – for Palestinians, the region at large, and also particularly for Israel.

In media commentary and analysis centred around AI's role in the ongoing war in Gaza, it has often been suggested that militaries worldwide will be watching closely and learning lessons from Habsora's application.

Only time will tell what those lessons will be.

font change

Related Articles