Introduction
The future battlefield is not emerging gradually; it is arriving in the present at extraordinary speed. In only a few years, unmanned systems have evolved from intelligence-support tools into capabilities that shape modern warfare. Alongside tactical reconnaissance systems and loitering munitions, commercial drones that began as photography platforms have become inexpensive, accessible, lethal, and difficult to stop. The war in Ukraine, the fighting in the Middle East, and strategic competition among the major powers all show that this is not merely an incremental technological improvement. It is a profound transformation in operational concepts, force structure, the economics of warfare, and the law of armed conflict.
Four trends are driving this revolution. The first is the rapidly growing threat posed by small, inexpensive, expendable drones. Second is the spread of small unmanned systems into the land and maritime domains. Third is the shift from directly operating individual systems to employing swarms in the air, on land, and at sea, including integrated swarms controlled with the aid of artificial intelligence. Finally, artificial intelligence is being integrated into detection, target identification, target nomination, prioritization, and attack to an extent that increasingly pushes humans out of the decision-making process. This raises a particularly serious dilemma: as systems gain greater autonomy to determine who or what constitutes a legitimate target, and when force may appropriately and lawfully be used, they increasingly place the core principles of international humanitarian law—distinction, proportionality, precaution, and human responsibility—under strain.[1]
This article argues that the principal threat lies not in the autonomous platform itself but in the system it represents, which includes mass production, decentralized communications, learning software, inexpensive sensors, continuous connectivity, and an operational willingness to deploy masses of unmanned systems instead of soldiers.[2] A military swarm is not merely a collection of platforms. It is a distributed, networked, scalable system that is resilient to failure. It can operate with dozens, hundreds, or even thousands of platforms, distribute tasks among them, replace platforms that have been lost, and continue the mission even after part of the swarm has been destroyed. In this environment, the advantage shifts to actors that can shorten the sensor-to-shooter chain, employ coordinated swarms, and adapt rapidly to countermeasures.
The Threat Posed by Unmanned Systems Is Escalating
Drones have become a defining feature of contemporary warfare because they change the battlefield cost-benefit equation. A platform costing only hundreds or thousands of dollars can locate a force, adjust artillery fire, strike an armored vehicle, disrupt logistics, or generate sustained psychological pressure. In Ukraine, FPV drones, loitering munitions, unmanned ground vehicles, and unmanned surface vessels have demonstrated an operational impact far disproportionate to their cost. Ukraine has become a “laboratory of future warfare,” where inexpensive, expendable systems are reshaping combat and forcing large militaries to adapt rapidly.[3]
This escalation is evident across several dimensions. First, drones are now widely accessible. They are employed not only by states, but also by non-state organizations, militias, and small tactical units. Second, they have been integrated into every stage of the strike cycle: detection, identification, target nomination, attack, and battle damage assessment. Third, they operate in an environment saturated with countermeasures, including electronic warfare, interception systems, GPS jamming, small-arms fire, and detection systems. As a result, they are rapidly evolving toward optical navigation, alternative communications, autonomous flight, fiber-optic guidance, and attacks along unpredictable flight paths. Finally, they are reshaping the cognitive battlespace: every movement is exposed to sensors, and the forward arena has become transparent, threatened, and continuously vulnerable to attack.
The Ukraine conflict illustrates the new “economy of attrition.” Instead of relying on a small number of expensive platforms, both sides employ large numbers of expendable systems. Ukraine is developing a concept of AI-enabled autonomous warfare, although a gap still exists between technological ambition and broad, mature operational capability. The lesson is important: this revolution does not occur overnight. It advances through thousands of small operational experiments that generate learning faster than any traditional procurement process.[4]
Israel, too, has recognized the growing threat. The October 7 attack demonstrated how small, improvised drones can disrupt defensive systems, damage surveillance assets, and facilitate a ground invasion. Since then, Israel has accelerated development of tools for detection, electronic disruption, and interception, alongside the integration of artificial intelligence into small arms and counter-drone defense systems. Israel is focusing on technologies intended to transform personal weapons and tactical systems into effective means of engaging small aerial targets by integrating sensors, software, and improved firing and targeting capabilities.[5]
Drone defense cannot rely solely on an expensive interception layer. When the attacker employs large numbers of small, inexpensive platforms, the defender cannot always respond with costly missiles. What is needed is a multilayered architecture combining early detection, automatic classification, electronic warfare, jamming, takeover capabilities, inexpensive kinetic fire, directed energy, physical protection, deception, and dispersion and hardening of infrastructure. The challenge is not merely to intercept the individual platform, but to break the attacker’s cost-benefit equation.[6]
Unmanned ground systems are rapidly moving beyond limited support roles to become an organic component of combat teams. Ground robots were initially employed primarily for sabotage missions, explosive ordnance disposal, reconnaissance in hazardous areas, and carrying equipment. Today they are increasingly used for forward logistics, casualty evacuation, opening routes to allow forces to advance, armed reconnaissance, and attack missions. Integrating sensors, communications, autonomous navigation, and artificial intelligence allow these systems to operate ahead of manned forces, reduce risk to soldiers, detect enemy forces, hold terrain, and rapidly close the fire cycle. The next stage will be the deployment of synchronized groups of ground systems, namely, robotic swarms. In a synchronized swarm, one platform conducts reconnaissance, other carries munitions or payloads, a third serves as a communications relay, and a fourth supports evacuation or logistics. Another model is an entirely distributed swarm that attacks a military force like a pack of wolves. Ground robotics is thus evolving from a point solution into a capability that could transform the nature of ground maneuver.
Unmanned surface and underwater systems are likewise moving from isolated applications toward a systemic concept encompassing reconnaissance, threat, and attack. As demonstrated in Ukraine and Iran, small, inexpensive unmanned surface vessels can threaten warships, ports, and coastal infrastructure without exposing crews to risk. At the same time, unmanned underwater systems are extending the threat beneath the sea by laying mines, gathering intelligence, and conducting attacks against cables, pipelines, and critical infrastructure. Their future integration into maritime and underwater swarms will make it more difficult for an adversary to identify the source of the threat, defend extensive maritime areas, and preserve freedom of action at sea.
From Aerial Swarms to Multidomain Swarms
The shift from managing individual unmanned systems to employing swarms is a turning point. A swarm is not simply “a large number of drones” or “many unmanned systems.” Rather, it is a system in which numerous platforms share information, divide tasks, exchange roles, and adapt to changing circumstances. It may also operate in a distributed manner, with all platforms pursuing a common objective: to locate and destroy. A swarm is defined by decentralized control without a single leader, but with continuous communication among platforms, scalability across large numbers of systems, and relative autonomy based on algorithms and artificial intelligence. An effective swarm can confuse defensive systems, saturate radar coverage, conduct deception and attack simultaneously, detect targets across a wide area, and continue functioning even after losing part of its force. Its operational significance lies not merely in the number of platforms but in their coordination and collective behavior.
Four principal types of swarms are emerging: aerial swarms of UAVs and drones; ground swarms of robotic and unmanned ground vehicles; maritime and underwater swarms of unmanned surface and underwater vessels; and integrated swarms—or a “swarm of swarms”—in which air, land, and maritime platforms operate under a single networked command structure. The last poses the most serious challenge because it blurs the boundaries between operational domains and requires the defender to confront simultaneous threats arriving from different dimensions, at different tempos, and with different signatures.
Aerial swarms can saturate air defense systems, detect targets, and employ “hunters” and “spotters” simultaneously. On land, unmanned systems can conduct reconnaissance, forward logistics, casualty evacuation, laying of explosive devices, or opening of routes under fire. At sea, unmanned surface vessels have already demonstrated their ability to disrupt conventional fleets, strike infrastructure, and pose a persistent threat to ports, bridges, offshore platforms, and shipping. Ukraine’s battlefield is shaping the future of autonomous warfare across the air, maritime, and land domains.[7]
The maritime domain is especially significant because it disrupts traditional assumptions about naval superiority. A state or organization without a conventional navy can employ small, inexpensive, suicide unmanned surface vessels to threaten major warships, ports, and energy infrastructure. In the Russia–Ukraine war, unmanned surface vessels became sufficiently effective to force the Russian fleet out of certain areas of the Black Sea.[8] This provides a glimpse of a future in which maritime and underwater swarms operate alongside aerial and ground swarms as components of a single integrated system.
The central challenge in operating swarms is command and control. As the number of platforms increases, direct human control over every individual platform becomes impossible. Swarms therefore require at least partial autonomy in functions such as route selection, maintaining formation, target allocation, obstacle avoidance, and adaptation to platform losses. Some Ukrainian programs already enable drones and ground systems to participate in coordinated operations. These programs reduce the number of required operators while transferring part of the planning, task allocation, and execution process to algorithms.[9] Even when a human remains “in the loop” as a supervisor, that person becomes increasingly removed from the actual decision-making process.
From an Israeli perspective, maintaining a qualitative advantage requires superiority in sensors, artificial intelligence, secure communications, counter-drone technologies, and offensive swarm capabilities. At the same time, Israel faces adversaries that can rapidly adopt commercial technologies, learn from Ukraine, Iran, and Hezbollah, and deploy simple systems in large numbers. Israel must consequently regard artificial intelligence and autonomous systems—including drones, unmanned ground vehicles, and drone swarms—as essential components of future force development, while simultaneously preserving the human role in the decision-making process.[10]
IDF Chief of Staff Eyal Zamir’s remarks regarding the next multi-year force development plan are particularly relevant in this context. He stated that the plan will strengthen robotic capabilities and integrate them into combat teams, enabling remote operation, improving terrain control and enemy detection, and accelerating the sensor-to-shooter cycle. This reflects an understanding that ground robotics is no longer merely an auxiliary or experimental capability, but an organic component of combat teams—one that integrates terrain control, target detection, maneuver, lethality, and shortening of the sensor-to-shooter cycle. [11]
Multidomain swarms transform the concept of defense. It is no longer sufficient to defend a nation’s airspace or borders. Bases, ports, power stations, water facilities, communications networks, transportation routes, and maneuvering forces must all be protected against threats that may arrive simultaneously from air, sea, and ground. Defense must therefore become networked, localized, mobile, and adaptive. It must integrate intelligence, cyber capabilities, electronic warfare, firepower, physical protection, and deception, while operating at the pace at which the offensive threat evolves.
Case studies show that no single swarm model prevails. China is developing an intelligent, multidimensional swarm concept;[12] Russia relies on mass, attrition, and adaptation;[13] Ukraine demonstrates rapid innovation;[14] and the United States is pursuing a concept of “intelligent mass” that combines quantity, software, autonomy, and human control.[15] Across these models, the common lesson is that swarms shift the center of gravity of warfare away from the individual platform and toward systemic behavior. As the land, air, and maritime domains become increasingly integrated, defenders must do more than intercept individual platforms. They must understand patterns of operation, disrupt networks, strike production and launch infrastructure, and be able to make decisions faster than the enemy.
The Ethical Dimensions of Artificial Intelligence, Autonomous Targeting, and Removing Humans from the Decision-Making Process
The most dangerous development is not the mass use of unmanned systems, but the combination of swarms with AI-enabled decision-making. Artificial intelligence can analyze enormous quantities of information, identify patterns, detect anomalies, nominate and prioritize targets, and even operate unmanned platforms without specific human instruction. In doing so, it shortens military decision cycles while transferring part of human judgment to machines.
The legal debate focuses on Lethal Autonomous Weapons Systems (LAWS),[16] which can select and attack targets without human intervention. According to the United Nations Office for Disarmament Affairs (UNODA), there is still no internationally agreed definition of such systems. Yet states are increasingly developing and fielding systems with autonomous functions, including loitering munitions, defensive systems, unmanned ground vehicles, and unmanned maritime platforms.[17] The lack of an agreed definition does not diminish the problem. On the contrary, it allows states to operate in a gray area in which technology advances faster than the law.
International humanitarian law assesses new technologies against the principles of distinction, proportionality, precaution, and responsibility. Autonomous systems, however, struggle to make context-dependent assessments, particularly in environments where combatants and civilians are intermingled, information is incomplete and changes rapidly, and the legal significance of an action depends on subtle contextual factors. In 2025, the International Committee of the Red Cross called for preserving meaningful human control over decisions involving the use of force, arguing that the window for effective regulation is rapidly closing.[18]
The United Nations has described the danger in particularly strong terms. It has called for a prohibition on systems capable of causing human death without human control or oversight, describing them as politically unacceptable and ethically problematic.[19] This appeal is operational as well as moral. A machine that cannot understand human context, intent, surrender, injury, civilian status, or changing circumstances in real time risks turning the use of lethal force into an automated process while undermining command responsibility.
In practice, most armed forces are not adopting a model in which humans are entirely “out of the loop.” Instead, they progress through intermediate stages: the human in the loop, the human on the loop, and finally the human out of the loop. In the first stage, every strike requires human approval. In the second, humans supervise and can intervene, but the system performs a substantial portion of the process. In the third, the system selects targets and employs force on its own. The danger is that the transition from the second stage to the third may occur not through an explicit policy decision, but through the cumulative effects of software improvements, operational pressure, and the demand for shorter response times.
In Israel, a debate has emerged over the use of AI systems to support target nomination and manage fire cycles. A distinction must be drawn between employing AI as an analytical support tool and granting autonomous authority to employ lethal force. However, even a support tool can create “automation bias,” in which human operators tend to adopt the system’s recommendation because it appears more accurate, faster, and more objective. Integrating autonomous systems of this nature may create an accountability gap and accelerate the kill chain.[20]
The legal and ethical problem does not begin with the use of force. It begins when the algorithm identifies a pattern, designates a person or platform as a target, assigns tasks among platforms, and proposes priorities for attack. The less time, context, and capability the human operator must question the machine’s recommendation, the more human control becomes merely formal. The central question, therefore, is not simply whether there is “a human in the system,” but whether that human can understand the system’s decision, stop it, and accept responsibility for it.
The greatest danger is “normalization of the exception.” Initially, AI is introduced to help human operators cope with information overload. Subsequently, review times are shortened. Later, the system permits force against predefined categories of targets. Finally, under the pressure of swarms, saturation attacks, and response times measured in seconds, commanders may come to prefer systems that make decisions faster than human beings. When that point is reached, international law will confront a fait accompli: machines will determine who is harmed, while humans will be left to explain afterward why they failed to intervene.
An exclusive focus on “killer robots” misses the central challenge. Autonomy is not a binary condition, but a continuum of capabilities that can be incorporated into weapons systems at varying levels and in partial applications. Consequently, the principal danger is not confined to the moment when a weapon is fired. It lies in the gradual accumulation of capabilities for searching, identifying, classifying, and prioritizing targets, until the human operator remains formally present but no longer serves as the relevant decision-maker.[21]
Recommendations: Preparing for the Future Battlefield
Preparing for the future battlefield requires a conceptual shift, not merely the acquisition of new capabilities. A multilayered defense must integrate detection, classification, electronic warfare, inexpensive interception, directed energy weapons, physical protection, and deception. The objective is not simply to intercept every drone at any cost, but to identify swarm behavior, disrupt synchronization, disrupt the network, and defeat the enemy’s attack.
Israel should establish an independent swarm capability. A military that cannot operate swarms will struggle to understand how to defend against them. This capability should span air, land, and maritime domains and rest on domestic production, continuous operational experimentation, and the integration of frontline units into the development process. The lesson from Ukraine is that the pace of learning is almost as important as the quality of the platform itself.
Consistent with the Chief of Staff Zamir’s vision, the renewed emphasis on ground forces should be translated into a practical model for an integrated robotic-human combat team. Every battalion- and brigade-level combat team should therefore organically employ drones, ground robots, sensors, communications systems, and fire-support assets, ensuring that robotics does not remain confined to specialized units. These robotic-human teams will improve terrain control, accelerate enemy detection, enable remote operations in hazardous areas, and shorten kill chains, while preserving human oversight over sensitive targeting decisions.
Israel’s threat environment differs from that of larger states. It has limited strategic depth, multiple nearby fronts, and a maritime domain that is critical to energy security and trade. At the same time, it faces both state and non-state adversaries capable of generating serious multidomain threats. Israel cannot wait for expensive, fully mature generations of autonomous systems to emerge. Instead, it must develop swarms through rapid operational learning based on experimentation, operational deployment, lessons learned, continuous refinement, iterative production, and rapid integration into frontline units.
Offensively, the IDF should develop a multidomain swarm capability extending well beyond drones. In the air, it should develop swarms for reconnaissance, deception, electronic disruption, and attacks that can operate against air defense networks, locating launchers, and pursuing enemy cells even in GPS-denied environments. On land, it should develop robotic maneuver teams capable of operating ahead of manned forces, clearing routes, detecting explosive devices, searching buildings and tunnels, transporting equipment, and enabling precise attacks while reducing risk to soldiers. At sea, it should develop swarms of unmanned surface and underwater vessels for reconnaissance, protection of offshore platforms and ports, threat detection, deception, blocking maritime routes, and attacks against enemy vessels or infrastructure used to launch unmanned systems.
Defensively, Israel should move from point interception toward defense against swarm systems. The objective should be not merely to detect individual platforms, but to identify patterns of operation, including launch locations, communications relays, avenues of approach, the distribution of functions among platforms, and attempts to saturate existing defensive systems. To support this approach, Israel should develop localized defensive packages for maneuvering forces, military bases, ports, offshore platforms, power stations, and other critical facilities. These packages should integrate short-range detection, automatic classification, jamming and electronic disruption, accurate light weapons, inexpensive interceptors, deception measures, physical protection, and operating procedures designed for saturation attacks. Success should be measured not only by the number of platforms intercepted, but by the ability to disrupt the swarm. This capability is, of course, intended to complement—not replaced defenses against established threats such as missiles and rockets.
Israel should also develop the ability to strike every element in the swarm threat’s value chain, including production centers, storage facilities, launch sites, communications relays, command-and-control systems, ground stations, motherships, and navigation infrastructure. This would complement localized defensive measures by shifting the focus from individual platforms to broader systems that enable swarm operations.
The swarm era will also require organizational change. Alongside existing development directorates, the IDF should establish an operational body responsible for integrating ground, air, and naval forces with the C4I Directorate and the Directorate of Defense Research and Development (Mafat-DDR&D). This body should define a common architecture for operations, communications, friend-or-foe identification, safety, cybersecurity, documentation of human decisions and control. Individual military corps should not be permitted to develop independent swarm capabilities in isolation from one another. Israel’s comparative advantage will emerge only if swarm capabilities become components of a single integrated command-and-control system.
Israel should also establish clear legal and operational boundaries for employing artificial intelligence and swarms. A distinction should be maintained between decision-support systems and decision-making systems, with explicit rules governing which actions a swarm may perform autonomously, which actions require human authorization, and under what circumstances a mission must be terminated automatically. Any system that influences target nomination or strikes should undergo legal review, reliability testing, comprehensive documentation of data sources, explainability assessments, incorporation of human override mechanisms, and assignment of clearly defined command responsibility. No unmanned system should be permitted to employ lethal force against human beings without effective human control.
Israel should develop a “responsible human in the loop” doctrine. It is not enough merely to state that a human is part of the system. Doctrine should define what information a human in the loop possesses, when he should provide authorization, how he can terminate an attack, and what responsibility he bears in the event of failure. Effective human control is not just the formal act of pressing a button, but the genuine ability to understand, evaluate, and intervene in decisions. At the same time, democratic states should develop common standards for evaluating autonomous systems, sharing operational lessons, and restricting systems that endanger civilians or undermine human responsibility.
Finally, Israel should strengthen international cooperation in force buildup. In an earlier article addressing Israel’s need for greater armaments independence, we proposed establishing an international consortium dedicated to the development of critical military capabilities.[22] The development of autonomous systems across air, land, maritime, and underwater domains appears to be an appropriate candidate for such cooperation.
Conclusion
The future battlefield is already here. It is defined by inexpensive systems deployed at scale, multidomain swarms, artificial intelligence that analyzes information and makes decisions at machine speed, and a widening gap between the possibilities offered by technology and the capacity of law and ethics to regulate it. Swarms constitute a systemic transformation. They combine mass, tempo, redundancy, autonomy, and networking, allowing them to generate effects that depend not on any individual platform, but on the collective behavior of the system. This revolution does not eliminate the human role, but unless that role is redefined it will make humans increasingly redundant
Drones have transformed the economics of warfare. Swarms are transforming its tempo. Artificial intelligence may ultimately transform the very nature of responsibility in war. Future advantage will not necessarily belong to the actor with the most expensive platform, but to the one able to build a learning, networked, inexpensive, redundant, and rapidly adaptable system—while denying the adversary the ability to operate collectively.
The challenge is therefore twofold: to achieve technological superiority while preserving the moral foundations of military decision-making. Precisely because machines can see, calculate, and strike faster than humans, stronger human control is required—not as a legal ornament, but as an operational and moral safeguard. The future is arriving at extraordinary speed. The question is whether we will be prepared for it, and whether we will succeed in keeping responsibility in human hands.
[1] Distinction requires parties to distinguish between combatants and military objectives, on the one hand and civilians and civilian objects on the other. Weapons or systems incapable of making this distinction are prohibited.
Proportionality requires that the expected incidental harm to civilians and civilian objects not be excessive in relation to the concrete and direct military advantage anticipated.
Precaution require parties to take all feasible precautions to avoid or minimize harm to civilians, including through the choice of means and methods, warnings, and the cancellation of attacks.
Responsibility/accountability, is not one of the classic principles governing the conduct of hostilities in the same sense as the three above principles, but it remains a fundamental and binding principle. States and individuals, including commanders and operators, remain responsible for compliance with the law.
[2] Uzi Rubin, “Will Unmanned Aircraft Decide the War in Ukraine?,” Jerusalem Institute for Strategy and Security (JISS), May 27, 2026.
[3] Samuel Bendett and David Kirichenko, “Battlefield Drones and the Accelerating Autonomous Arms Race in Ukraine,” Modern War Institute, January 10, 2025.
[4] Kateryna Bondar, “Ukraine’s Future Vision and Current Capabilities for Waging AI-Enabled Autonomous Warfare,” Center for Strategic and International Studies, March 6, 2025.
[5] Pesach Benson and Omer Novoselsky, “Israeli AI Turns Standard Rifles into Drone Killers,” TPS-IL, February 9, 2026.
[6] Sven Clement, “Mastering the Future of Uncrewed Warfare,” NATO Parliamentary Assembly, October 12, 2025.
[7] Peter Dickinson, “Ukraine Is Shaping the Future of Drone Warfare at Sea as Well as on Land,” Atlantic Council, June 12, 2025.
[8] Tsiporah Fried, “The Impact of Drones on the Battlefield: Lessons of the Russia-Ukraine War from a French Perspective,” Hudson Institute, November 13, 2025.
[9] Patrick Tucker, “This Ukrainian Startup Has Re-Invented Drone Swarming,” Defense One, September 15, 2025.
[10] Anna Ahronheim, “Israel’s Strategic Edge in the Age of AI & Autonomous Warfare,” The Jerusalem Post, July 20, 2025.
[11] Yoni Kempinski, “Chief of Staff: We Will Build a Decisive Ground Force,” Arutz Sheva, July 23, 2026.
[12] Timothy Ditter, “PRC Concepts for UAV Swarms in Future Warfare,” CNA, November 7, 2025. https://www.cna.org/analyses/2025/10/prc-concepts-for-uav-swarms-in-future-warfare
[13] Missile Defense Advocacy Alliance, “MDAA Alert: Russia’s Advancing Geran Family of Attack Drone Capabilities,” June 29, 2026. https://www.missiledefenseadvocacy.org/alerts/mdaa-alert-russias-advancing-geran-family-of-attack-drone-capabilities
[14] Vlad Cherevko, “Ukrainian Forces Using Drone Swarms That Autonomously Locate and Strike Targets—WSJ,” Ukrainska Pravda, September 3, 2025. https://www.pravda.com.ua/eng/news/2025/09/03/7529154
[15] Thomas Novelly, “Anduril, General Atomics Get Air Force Contracts to Build First Drone Wingmen,” Defense One, June 17, 2026. https://www.defenseone.com/defense-systems/2026/06/anduril-general-atomics-get-air-force-contracts-build-first-drone-wingmen/414266
[16] The term LAWS stands for lethal autonomous weapons systems. These are AI-enabled military systems capable of identifying, selecting, and engaging targets without human intervention in the operational decision-making loop.
[17] United Nations Office for Disarmament Affairs, “Lethal Autonomous Weapon Systems,” accessed July 26, 2026.
[18] Mirjana Spoljaric, “Preserving Human Control over the Use of Force,” International Committee of the Red Cross, May 12, 2025.
[19] UN News, “‘Politically Unacceptable, Morally Repugnant’: UN Chief Calls for Global Ban on ‘Killer Robots,’” May 14, 2025.
[20] Ramón Reichert, “Autonomous Occupation: Israel’s AI-Driven Drone Warfare and the Digital Architecture of Authoritarian Power,” Dialogues on Digital Society 1, no. 3 (2025).
[21] Gabi Siboni and Yoni Eshpar, “Dilemmas in the Use of Autonomous Weapons,” Strategic Assessment 16, no. 4 (January 2014).
[22] Gabi Siboni and Erez Winner, “Israel’s Path to Armaments Independence,” Jerusalem Institute for Strategy and Security (JISS), July 24, 2026.
JISS Policy Papers are published through the generosity of the Greg Rosshandler Family.
