Israel’s AI advantage is strategic, but humans in the IDF remain responsible - opinion

Recently, a controversial film was released overseas and even won an award at the Venice Film Festival. The film, “The Causality of War,” portrays IDF soldiers and the State of Israel in an extremely negative light, without providing evidence that can be independently examined. The movie is relying, among other things, on anonymous interviews allegedly conducted with 24 low-ranking intelligence soldiers.

The film accuses Israel of genocide and of disregarding human life in the Gaza Strip. It focuses on intelligence personnel who provide targets to the air, ground, and naval forces that use this precise intelligence to eliminate immediate terrorist threats to Israel to save Israeli lives.

I have not watched the film, I do not intend to watch it, and I would not recommend that any Israeli who respects himself or herself watch it.

According to reports circulating online and in several irresponsible Israeli media outlets that have echoed the film favorably, it directs its accusations not only at the IDF, but also at artificial intelligence (AI), which the film allegedly portrays as being used to direct fire and indiscriminately harm innocent people.

It is therefore important to explain the actual role of artificial intelligence in military systems, particularly in the intelligence systems criticized by the film, and to clarify why the claims presented in such a biased manner do not reflect the way these systems are actually operated.

The relevant authorities, headed by the prime minister, IDF chief of staff, the military advocate-general, the Culture and Sports minister, and other officials, were right to respond firmly to the publication. Some have also announced that they will consider the appropriate steps available under Israeli law in response to what they regard as efforts to damage Israel’s reputation around the world.

AI in the IDF and the Defense Establishment

Artificial intelligence is an extremely important tool, subject to the necessary safeguards and limitations, in a wide range of military applications, some classified and therefore beyond the scope of this article, and others unclassified.

Intelligence

AI tools can be highly effective in processing natural language, including reading, writing, and analysis, when dealing with rich and dynamic context.

A significant portion of intelligence output consists of text embedded in precisely this kind of rich, dynamic context. AI can therefore provide advantages over traditional algorithms, which are generally better suited to stable, repetitive tasks involving relatively little contextual complexity.

AI can also review existing intelligence and identify gaps between available information and the intelligence assessment. This can reduce the risk of serious analytical errors, such as those that may have contributed to the failure to anticipate the events of October 7.

AI can also accelerate the process of identifying an enemy and developing potential targets. Here, however, it is particularly important to understand the role of the human being in the system; the human remains firmly in the loop. The only entity that makes a decision concerning life and death is a human being.

AI can also accelerate information-processing tasks that require a high degree of contextual understanding or the integration of different types of information. Large language models (LLMs), in particular, are increasingly capable of processing and fusing multiple modalities, linking text, audio, images, and structured information in ways that can improve the integration and understanding of complex data.

AI tools are also useful in detecting particularly weak signals and connecting them to one another. This has traditionally been an especially difficult task for human intelligence, particularly in the field of early warning.

The reason is straightforward: AI can review extraordinarily large quantities of data and identify indications that may be difficult for a human analyst to notice.

AI-assisted software development

AI has additional important applications in the IDF and the defense establishment, all of which are monitored and operated with appropriate caution.

AI tools can help write code for military systems quickly and efficiently. Programming languages are highly structured and supported by large amounts of data, with relatively low variability compared with natural human language.

This can reduce software development costs and accelerate development, allowing military systems to adapt more quickly to the dynamic operational requirements characteristic of modern warfare.

Because the military deals with missions that can directly affect human life, it is essential to ensure, through rigorous professional and human oversight, that software generated with AI and ultimately integrated into operational systems performs at the required level, can withstand heavy loads and edge cases, and remains stable and robust under extreme conditions.

Physical AI

Many military missions involve operating in and influencing the physical environment, often under dangerous conditions and at considerable distances.

Physical AI can support capabilities such as navigation, the operation of autonomous and robotic systems, and rapid responses to threats, including detection and mitigation.

Here too, however, decisions involving human life are made only with human involvement. Qualified personnel must assess the circumstances and make informed decisions while considering the specific environmental conditions and applicable ethical constraints.

This area is particularly relevant to military organizations, and the IDF is among the most advanced in this field.

AI and Cybersecurity

Because AI tools are increasingly capable of writing code, they can also identify vulnerabilities and weaknesses in software and potentially exploit them, alongside other development capabilities required in cybersecurity.

AI tools that can identify and exploit vulnerabilities can also help defenders detect attackers.

For many years, this capability was largely associated with militaries and intelligence organizations. More recently, leading AI companies have developed tools that can discover vulnerabilities and direct software developers’ attention to closing security gaps. This development may make the work of intelligence organizations around the world, and the IDF, more difficult.

Everything described above demonstrates that, contrary to the picture presented by the film, the use of AI in IDF systems is accompanied by clear and highly cautious operational methodologies.

Military orders and procedures explicitly prohibit relying exclusively on AI applications in situations involving human life without human involvement.

AI-based algorithms do not make decisions to launch an attack and are unlikely to do so in the coming years. Authorized human personnel make such decisions under close legal oversight, much as human professionals remain responsible for AI applications involving human life in medicine.

Decision-makers can and should use AI to maximize the information available to them and to provide recommendations, including relevant precedents and historical examples, so that they can make the most informed decision possible.
But although it may sound attractive, AI systems are not intended to replace the professionals responsible for making these decisions.

Beyond the scientific research itself, the use of AI has led to the development of extensive evaluation methodologies designed to ensure that an AI product meets predefined performance thresholds before it can be used operationally.

These processes include testing the stability of algorithms over time and under different conditions, creating randomized test samples, and establishing multiple regulatory and approval mechanisms before an AI system is authorized to provide recommendations for human use.

The system is providing recommendations, not making the decision.

In situations where intelligence may lead to an attack, the identification of a target must in any case be supported by at least two separate indicators from independent sources, and the final decision must be made by a human authority.

AI as a strategic capability, the need for skilled personnel

As AI models become more powerful and acquire increasingly far-reaching capabilities, the technology, both hardware and software, is beginning to shift from being a global commodity whose distribution is driven primarily by supply, demand, and economic efficiency, toward becoming a strategic national asset.

Export controls are increasingly being imposed on such technologies, while governments are promoting domestic production of critical components in order to support national needs first.

AI models themselves are now becoming strategic national assets. Under these conditions, an important question arises: how much independence does Israel need in the field of AI models, both civilian models and models used for military purposes?

This issue was examined extensively by the Nagel Committee, which addressed AI in Israel’s government and academic systems.

Elite IDF programs such as "Alonim" and "Bareket" are specifically designed to train the personnel the IDF needs for data-science projects at all levels. Israel’s leading technological units across the military branches highly value graduates of the programs.

AI in civilian systems

As in military systems, civilian systems can also include applications in which AI may ultimately influence life-and-death decisions, usually indirectly.

Civilian systems are subject to extensive monitoring and regulation. This is probably one of the reasons why autonomous civilian systems, such as autonomous vehicles, are entering widespread use much more slowly than earlier forecasts suggested.

At the same time, regulatory and supervisory capabilities are more difficult to enforce in some civilian environments. We may therefore see, in the future, civilian systems that rely on AI in which operational errors could, under certain circumstances, result in loss of human life.

What is Israel’s position in the global AI race?

Unfortunately, Israel is not yet the world leader in AI, as it is in cybersecurity, particularly in civilian AI, although not exclusively.

The establishment of the AI Directorate in the Prime Minister’s Office is intended, among other things, to help position Israel among the world’s leading countries in this strategically important field.

Brig. Gen. (Res.) Prof. Jacob Nagel served as National Security Adviser to Prime Minister Benjamin Netanyahu and as Acting head of Israel’s National Security Council. He is a Visiting Professor at the Technion, and a Senior Fellow at the Foundation for Defense of Democracies (FDD) in Washington, D.C.