On the morning of Saturday, 7 October Hamas launched a co-ordinated and devastating terrorist operation against Israel that caught it completely by surprise. The country was celebrating Shabbat, and its defence forces were totally unaware and unprepared for the swift assault. How did the much-celebrated Israeli intelligence miss the coming of the attack that needed huge resources, co-ordination and planning?

This question is puzzling people beyond the expert community, and while certainties about it may only be established afterwards, intelligence theory and history may provide some first, plausible, explanations. One of the factors that led to this catastrophic intelligence failure was probably an overreliance on AI by the Israeli intelligence services.

The historical parallel with the last big trauma of intelligence failure – the 1973 unforeseen attack starting the Yom Kippur War – sounds even more compelling, as Saturday’s Hamas attack on Israel happened exactly on the 50 anniversary of those events.

But while some factors are “classics” of intelligence failure, which can be traced to the Yom Kippur precedent and to other such historical instances, new times introduced a novelty. A new scope of intelligence vulnerabilities which is the massive employment of AI.

The causes behind the tactical surprise achieved by Hamas are substantially three.

Hubris strikes when a state and its security apparatus are extremely successful. When everyone has wonderful things to say about you, you start to believe them. Due to Israel seeing off the Arab states in 1848, 1956 and 1973, they believe their own press. The problem is that Hamas and Gaza are a very different issue than the Arab states. Israel thought that walling Gaza off and getting rocket/missile defence was sufficient to contain Hamas. Of course Hamas saw this as a challenge and worked with their IRGC/Quds Force handlers in Iran to solve their challenge. Every few years Hamas builds up enough munitions and trains enough replacements to be able to mount an operation. The Israelis smack them down and Hamas goes back to nurturing its grievances and rebuilding an offensive capability. The three differences this time were: 1) Hamas got even more rockets/missiles than before and 2) the international situation is more unstable, and 3) the Israeli population and government are riven by internal factions.

Internal factionalisation is inherent in heterogeneous democracies like the US and Israel; however, they are currently more of an issue. The political process is that of prioritising approaches/policies and apportioning resources; heterogenous states are more difficult because each “in group” seeks to prioritise its policies and resources as compared to the other groups. Russia/China/North Korea/Iran are pressing on the democracies that seek to contain them, including the US and Israel. The current situation combines both external efforts to generate internal frictions (led by Russian information confrontation operations) together with hard-core right-wing and left-wings politicians seeking to promote their own agenda. One example is the current effort by the Netanyahu government to disempower the Israeli Supreme Court. This effort has driven a major wedge within Israeli society. This could be perceived by adversaries such as Hamas and Hezbollah, as well as their sponsor Iran, as an opportunity.

Over-reliance on Artificial Intelligence.

The massive employment of AI, Big Data and Machine Learning that characterises Israel's defence system may have constituted a blinding factor. First, because the enormous amount of data that is collected necessarily requires an AI-operated analysis to process it. This results in a process that can only partially be supervised by human brains, that would be in turn overwhelmed by this “data dumping”. Secondly, as in the case of “Iron Dome”, the air defence system operating since 2011 in Israel – that has in the last attack failed to intercept a number of initial strikes against the country – Israeli defence systems have been carefully studied by Hamas since the 11-day war of 2021, the last instance of big-scale confrontation.

In the case of the 2021, 11-day war – dubbed as the first AI-enabled conflict of history – it followed a 5-year Israeli effort to overhaul and upgrade its defence capabilities, combining changes in the field of organisation, doctrine and tools. Israel assumed that Hamas, or any attacker, would not know how to hit its defence capabilities.

But after the much publicised successes of the “AI war” of 2021, Hamas – likely properly helped by the IRGC – has done its homework, understanding how to deceive the artificial intelligence collection system, and consequently manipulating it into flawed analysis; not “going dark”, as the system would be alerted by such an abrupt change, but feeding the “panopticon” with tonnes of misleading business-as-usual information, that would not alert about any intentions of attack.

After the traumatic Yom Kippur War intelligence failure, Israel got rid of its previous approach deriving from a “postulate” – the “concept” that the enemy would never start a conflict engendering a stark destructive response from Israel – that made its analyst overlook the several warnings collected about an imminent attack.

Here, the question is upside down. Massive collection and analysis may have not been handled properly for the very reason they were just too much.

Therefore overconfidence, overreliance and underestimation of the enemy’s capabilities are the old new mistakes in the much-reputed, new-generation defence system of Israel.

Adding to these is the probable focus on threats from the West Bank, rather than Gaza; the conflictual internal situation in the country; and not least the moment of religious celebration – that’s how it has been possible for one of the much celebrated intelligence services of the world to be caught by surprise.

 

Dr Claudia Palazzo is a PhD candidate in intelligence studies. Dr Alexander Crowther is an intelligence expert and a non-resident Senior Fellow with the Transatlantic Defence and Security Programme at the Centre for European Policy Analysis.