Autonomous weapons systems powered by artificial intelligence are no longer the exclusive domain of science fiction or classified defense laboratories—they are active participants in contemporary armed conflicts, deployed in Ukraine, Libya, Gaza, and beyond. This article provides a comprehensive, research-grounded analysis of what lethal autonomous weapons systems (LAWS) are, how they work, where they are already deployed, what international law says about them, and what the global community is doing—and failing to do—to govern them. It draws on primary sources including the ICRC’s March 2026 position paper, UNGA Resolution A/RES/79/62 (December 2024), the November 2025 UNGA First Committee vote, the CCW Group of Governmental Experts’ rolling text, SIPRI’s report on responsible military AI procurement, and Benjamin Perrin’s analysis published in ASIL Insights (January 2025). The purpose is to equip policymakers, researchers, strategists, and military professionals with a clear and current picture of one of the most consequential governance challenges of the twenty-first century.

Abstract

Lethal autonomous weapons systems (LAWS) have moved from theoretical concern to operational reality, deployed across active conflicts in Ukraine, the Middle East, and Africa. This article examines the technological foundations, legal status, and governance challenges of weapons that, once activated, can select and engage targets without further human intervention. It surveys deployed systems—the Phalanx, HARPY, Lancet-3, and KARGU—and analyzes how the international humanitarian law principles of distinction, proportionality, and precaution apply to autonomous targeting. The analysis tracks accelerating diplomatic momentum, including UN General Assembly Resolution A/RES/79/62 (December 2024, adopted 166–3) and a November 2025 follow-on resolution backed by 156 states, alongside the decade-long Convention on Certain Conventional Weapons (CCW) process. The article argues that existing law applies but leaves dangerous interpretive gaps, that the consensus-bound CCW model has slowed binding action, and that a two-tiered treaty—prohibiting autonomous targeting of people while regulating other systems under meaningful human control—is both necessary and achievable. It concludes that the 2026 CCW Review Conference represents a closing window for codifying human accountability before unilateral state practice sets the de facto rules.

Key Takeaways

• LAWS already exist and are deployed: Systems such as the Phalanx (US), HARPY (Israel), Lancet-3 (Russia), and KARGU (Türkiye) demonstrate that autonomous targeting is not a future threat—it is a present reality.
• Existing international humanitarian law applies, but gaps remain: The IHL principles of distinction, proportionality, and precaution govern LAWS, but the absence of LAWS-specific rules creates dangerous interpretive ambiguity.
• International momentum is accelerating: UNGA Resolution A/RES/79/62 (December 2024) passed with 166 votes in favor. A follow-on resolution in November 2025 secured 156 states in support. Negotiations toward a binding treaty are now widely expected by 2026.
• The ethical stakes are profound: Delegating life-and-death decisions to algorithms raises fundamental questions about human dignity, moral responsibility, and accountability.
• A treaty framework is both necessary and achievable: Civil society, major international bodies, and a growing coalition of states agree on the core elements—prohibition of fully autonomous targeting of people, and regulated use of other LAWS with meaningful human oversight.

1. What Are Lethal Autonomous Weapons Systems?

Lethal autonomous weapons systems are weapons platforms that, once activated, can independently identify, select, track, and engage targets without requiring real-time human authorization. The capability is made possible by converging advances in artificial intelligence, machine learning, computer vision, sensor fusion, and robotics—technologies that have matured rapidly over the past decade. As Professor Benjamin Perrin of the University of British Columbia’s Allard School of Law observed in ASIL Insights (January 2025), LAWS emerge from “advances in artificial intelligence, robotics, sensor technologies, and real-time data processing, enabling systems to independently identify, track, and engage targets.”¹

It is important to note at the outset that there is currently no consensus international legal definition of LAWS.² However, the working definition widely used in diplomatic and academic forums is clear enough in its core elements: a LAWS is a weapons system that, once activated, “selects targets and applies force without human intervention.”³

This definitional framework carries significant practical implications. Not all automated or remotely operated weapons qualify as LAWS. A remotely piloted drone operated by a human controller in real time is not autonomous. A missile that follows a pre-programmed trajectory is not autonomous. What distinguishes LAWS is the capacity for machine-initiated target selection and engagement under conditions that may be dynamic, unpredictable, and legally complex.

How Do the Three Tiers of Autonomy Differ?

The most analytically useful taxonomy organizes weapons systems along a spectrum of human involvement:

Semi-autonomous (human-in-the-loop): The system can identify and track targets but requires explicit human authorization before applying force. Human judgment remains the essential gate between identification and engagement.

Supervised autonomous (human-on-the-loop): The system selects targets and applies force without requiring human authorization but operates under human supervision. The operator retains the ability to override or abort the system, though this may require timely intervention in rapidly evolving scenarios.

Fully autonomous (human-out-of-the-loop): The system selects and engages targets entirely without human authorization, supervision, or intervention. Once activated, its operation is effectively independent of human decision-making.

Both supervised and fully autonomous systems meet the operational definition of LAWS, since both are capable of applying lethal force without direct human authorization. The policy debate centers primarily on whether fully autonomous systems should be prohibited outright and under what conditions supervised autonomous systems can be lawfully and ethically deployed.

2. The Technology Behind AWS: AI, Robotics, and Sensors

Understanding the governance challenge requires a working grasp of the enabling technology. Four technological domains converge in modern LAWS:

Artificial Intelligence and Machine Learning

AI algorithms—particularly deep learning systems trained on large labeled datasets—enable LAWS to perform tasks once reserved exclusively for human cognition: distinguishing a tank from a civilian vehicle, identifying a combatant from a non-combatant in complex terrain, or determining whether a target poses an imminent threat. Computer vision systems trained on millions of images can perform these discriminations with remarkable speed. The concern is not merely their speed, but their opacity: the decision logic of deep learning models is often difficult to interpret, predict, or audit—what AI researchers call the “black box” problem.

Sensor Fusion

Effective autonomous targeting requires integrating data from multiple sensor streams simultaneously—radar, infrared, electro-optical cameras, sonar, lidar, and communications intercepts. Sensor fusion algorithms combine these streams into a coherent situational picture, enabling the system to track targets across changing conditions. When individual sensors fail or produce ambiguous data, however, fused outputs may still be wrong—with lethal consequences.

Robotics and Mobility Platforms

LAWS are deployed across multiple physical domains: aerial (loitering munitions and UAVs), naval (surface vessels and submarines), ground (armored vehicles and infantry robots), and increasingly in the space and cyber domains. The miniaturization of robotics components has dramatically lowered the cost and increased the accessibility of autonomous platforms, enabling deployment in large swarms.

Communications and Network Architecture

Modern LAWS typically operate within larger command-and-control networks, receiving targeting parameters, environmental data, and mission updates via encrypted communications links. The architecture raises its own vulnerabilities: systems may be subject to electronic jamming, spoofing, or hacking—raising the prospect of LAWS being commandeered or misdirected by adversaries.

The combination of these technologies creates weapons platforms capable of operating at speeds, scales, and levels of complexity that fundamentally challenge the human cognitive capacity required for lawful and ethical targeting decisions.

3. Real-World Examples of Lethal Autonomous Weapons Systems

The discussion of LAWS is sometimes dismissed as speculative. It should not be. Multiple operational systems already meet or approach the definition of lethal autonomous weapons. The following examples, drawn directly from Benjamin Perrin’s 2025 ASIL Insights analysis, illustrate the range of deployed capabilities.⁴

The Phalanx Weapon System (Raytheon, United States)

The Phalanx is a close-in weapons system with both naval and land-based configurations. In its naval role, it “automatically detects, evaluates, tracks, engages and performs kill assessment against anti-ship missiles and high-speed aircraft threats.”⁵ On land, the Counter-Rocket, Artillery, and Mortar (C-RAM) variant intercepts “rockets, artillery and mortar rounds in the air before impact.”⁶ Phalanx operates in a supervised autonomous mode: a human operator activates the system and sets engagement parameters, but the system autonomously executes targeting and firing decisions within those parameters. The speed of incoming threats—fractions of a second to impact—makes meaningful real-time human override effectively impossible.

HARPY (Israel Aerospace Industries, Israel)

The HARPY is a fully autonomous anti-radiation loitering munition. Once launched, it “is equipped to hunt—seek targets in a designated area, locate and identify their frequency, and autonomously pursue a strike from any direction, at shallow or steep dive profiles.”⁷ It is a “fire and forget” autonomous unmanned aerial vehicle (UAV) designed to destroy enemy radar installations. Crucially, the HARPY does not require prior intelligence on the target’s location before launch. It identifies and attacks radar emitters it detects autonomously—a paradigmatic example of a human-out-of-the-loop system operating within predefined parameters.

Lancet-3 (Zala Group / Kalashnikov Group, Russian Federation)

The Lancet-3 is a loitering munition “for reconnaissance, surveillance, and strike missions” described as “a smart multipurpose weapon, capable of autonomously finding and hitting a target.”⁸ It transmits real-time video—allowing for confirmation of target engagement—but its target-selection and attack logic operates autonomously. The Lancet-3 has been used extensively in the conflict in Ukraine, where it has targeted armored vehicles, artillery, and air-defense systems.

KARGU (STM, Türkiye)

The KARGU is a rotary-wing attack drone designed for tactical intelligence, surveillance, reconnaissance (ISR), and precision strike for ground troops.⁹ While its manufacturer, Savunma Teknolojileri Mühendislik (STM), contests certain characterizations, a UN Panel of Experts stated in March 2021 that the KARGU-2 was used by forces affiliated with the Libyan Government of National Accord to autonomously attack militias—marking what may be one of the first documented instances of a fully autonomous lethal engagement of human targets.¹⁰

These four systems collectively illustrate several critical points: LAWS span multiple countries and development ecosystems; they operate in both defensive and offensive roles; and the line between supervised and fully autonomous behavior is technically thin and situationally variable.

4. International Humanitarian Law and Autonomous Weapons Systems

Does Existing IHL Apply to LAWS?

The starting point in any legal analysis of LAWS is the unambiguous international consensus that existing international law—including international humanitarian law—fully applies to autonomous weapons systems.¹¹ This includes the UN Charter, IHL (comprising the Geneva Conventions and their Additional Protocols), international criminal law, international human rights law, the law of state responsibility, and treaties governing specific categories of weapons (chemical, biological, nuclear).

The ICRC’s March 2026 position paper, Autonomous Weapon Systems and International Humanitarian Law: Selected Issues, reinforced this position while emphasizing that the application of existing rules to LAWS surfaces concrete legal challenges that demand specific regulatory responses.¹²

Distinction, Proportionality, and Precaution

Three foundational principles of IHL are particularly consequential for LAWS:

Distinction requires parties to armed conflict to distinguish at all times between civilians and combatants, and between civilian objects and military objectives. LAWS must be capable of making this discrimination reliably, across the full range of contexts in which they are deployed. For AI-driven systems trained on data from prior conflicts, the challenge is substantial: civilian behavior is complex, dynamic, and context-dependent in ways that defy simple algorithmic classification.

Proportionality prohibits attacks expected to cause civilian casualties or damage to civilian infrastructure that would be excessive in relation to the concrete and direct military advantage anticipated. This is an inherently contextual judgment—one that requires weighing competing considerations in real time. Whether an AI system can perform this analysis with the necessary legal and moral sophistication is deeply contested.

Precaution requires that parties to a conflict take all feasible precautions to avoid or minimize civilian harm. This includes, among other things, selecting means and methods of warfare that are least likely to cause incidental civilian casualties, and canceling or suspending attacks when civilian harm becomes apparent. A LAWS operating at machine speed may be architecturally incapable of incorporating precautionary pauses.

The Martens Clause and the Principle of Humanity

The Martens Clause, first articulated in the 1899 Hague Convention and incorporated into the 1977 Additional Protocol I, provides that in situations not expressly regulated by treaty, combatants remain “under the protection and the authority of the principles of international law derived from established custom, from the principles of humanity and from the dictates of the public conscience.”¹³

This clause has special salience for LAWS. As Perrin observes, the ICRC has argued that “life-and-death decisions in armed conflict ceded to machines” may cross the line established by the principles of humanity and the dictates of public conscience.¹⁴ The Martens Clause does not, by itself, prohibit LAWS—but it establishes a normative floor against which their use must be evaluated, even in the absence of LAWS-specific treaty law.

Weapons Reviews Under Article 36 of Additional Protocol I

Article 36 of Additional Protocol I to the Geneva Conventions requires States to conduct legal reviews of “new weapon[s], means or method[s] of warfare” to determine whether they would violate international law. This obligation applies directly to LAWS development programs. A credible Article 36 review would need to assess whether a proposed LAWS can reliably perform the distinction, proportionality, and precaution analysis required by IHL—across the full range of scenarios in which it might be deployed, including degraded operating environments, electronic jamming, and unexpected civilian presence.

The practical reality is that such reviews are conducted differently across states—if at all. The absence of shared standards for what constitutes a rigorous Article 36 review creates inconsistency and risk.

5. The International Regulatory Landscape

The CCW Group of Governmental Experts on LAWS

The primary multilateral forum for LAWS-specific negotiations has been the Convention on Certain Conventional Weapons (CCW) Group of Governmental Experts (GGE) on Lethal Autonomous Weapons Systems, which has been meeting since 2014. Over more than a decade, the GGE has produced valuable analytical work, including detailed discussions of IHL applicability, human-machine interaction frameworks, and guiding principles for the development and use of LAWS.

In November 2024, the GGE issued a revised rolling text—a draft framework document—outlining potential regulatory measures for LAWS. Key elements include:

• Requiring LAWS to be predictable, reliable, traceable, and explainable to maintain lawful use under IHL.
• Maintaining “context-appropriate human oversight, particularly in morally and legally significant decisions, such as during the identification and/or engagement of targets.”
• Limiting the operation of LAWS by restricting target types, duration, geographical scope, and scale of operations.
• Enabling human operators to deactivate LAWS after activation.
• Incorporating self-destruct, self-deactivation, or self-neutralization mechanisms.
• Avoiding deployment in areas densely populated with civilians or civilian objects.
• Mitigating biases in AI and automation through regular evaluations and corrective measures.¹⁵

The CCW process operates by consensus, meaning any one member state can effectively block progress. Civil society observers—including Stop Killer Robots—have documented what they characterize as stalling tactics by the Russian Federation, contributing to a perception that the CCW process alone cannot deliver the binding framework the situation demands.¹⁶

UN General Assembly Resolution A/RES/79/62 (December 2024)

On December 2, 2024, the UN General Assembly adopted Resolution A/RES/79/62 on Lethal Autonomous Weapons Systems with 166 votes in favor, 3 opposed (Belarus, the Democratic People’s Republic of Korea, and the Russian Federation), and 15 abstentions.¹⁷ The resolution represented the most significant UN General Assembly action on LAWS to date. It acknowledged the potential for a two-tiered treaty approach—prohibiting certain categories of LAWS outright while regulating others—and mandated informal consultations among member states in 2025, drawing on the work of the CCW GGE.

The 156-State UNGA Resolution (November 2025)

On November 6, 2025, the First Committee of the UN General Assembly adopted Resolution L.41 on autonomous weapons systems for the third consecutive year—this time with 156 states in favor, 5 against, and 8 abstentions.¹⁸ Tabled by Austria and 30 co-sponsoring states, the resolution expressed concern about the “consequences and impact of autonomous weapon systems on international peace and security, including the risk of an emerging arms race, of exacerbating conflicts and humanitarian crises, miscalculations, lowering the threshold for and escalation of conflicts and proliferation, including to unauthorized recipients and non-State actors.”¹⁹

The resolution stressed “the importance of the role of humans in the use of force to ensure responsibility and accountability” and called upon CCW High Contracting Parties to “work towards completing the set of elements for an instrument being developed within the mandate of the Group of Governmental Experts, with a view to future negotiation.”²⁰

Civil society critics, including Stop Killer Robots, expressed disappointment that the resolution did not mandate formal negotiations—reflecting the persistent gap between broad political consensus on the urgency of the problem and concrete commitments to binding solutions.

The 2026 GGE on LAWS Sessions

The 2026 GGE sessions represent the final phase of the current GGE mandate, which concludes ahead of the CCW Seventh Review Conference. Working Paper 2 (CCW/GGE.1/2026/WP.2) and associated documents reflect continued progress on the rolling text—particularly regarding the elements of meaningful human control, operational restrictions, and accountability mechanisms. All eyes within the LAWS diplomatic community are on the Review Conference as the moment at which states must decide whether to launch formal treaty negotiations.

The ICRC Position Paper (March 2026)

The ICRC’s March 2026 position paper, Autonomous Weapon Systems and International Humanitarian Law: Selected Issues, represents the most comprehensive and authoritative non-state articulation of the IHL challenges posed by LAWS to date.²¹ The ICRC argues that while existing IHL applies in full, the specificity and predictability of a LAWS-specific treaty instrument is essential to close interpretive gaps. The paper addresses, among other issues, the challenge of verifying IHL compliance in autonomous systems; the conditions under which human oversight can be considered “meaningful”; and the legal implications of AI bias in targeting algorithms.

6. The Case for a New International Treaty

Calls by the UN Secretary-General and the ICRC President

The case for a new treaty on LAWS has been articulated with unusual clarity and directness by two of the international community’s most credible voices. In October 2023, UN Secretary-General António Guterres and ICRC President Mirjana Spoljari issued a joint call for states to “establish new prohibitions and restrictions on autonomous weapon systems.”²² They specified a two-tiered framework: outright prohibition of autonomous targeting of people without human involvement, combined with regulation of other LAWS—”limiting where, when and for how long they are used, the types of targets they strike and the scale of force used, as well as ensuring the ability for effective human supervision, and timely intervention and deactivation.”²³ Both leaders have called for the conclusion of negotiations on a new international treaty on LAWS by the end of 2026.

The logic of their position is straightforward. Existing general rules and principles of international law apply to LAWS, but “without specific rules,” Guterres and Spoljari warned, “too much will be left to varied interpretations by States”—a situation that is already materializing as major military powers develop LAWS programs with divergent and often opaque internal policies.

CCW Rolling Text Proposals

The GGE’s revised rolling text from November 2024—and its further development in 2025 and 2026—provides a concrete starting point for treaty negotiations. The rolling text’s emphasis on predictability, reliability, traceability, explainability, and context-appropriate human oversight offers a regulatory architecture that could accommodate both defensive systems like the Phalanx and more aggressive offensive LAWS, while establishing clear outer limits.

Forty-two states delivered a joint statement at the September 2025 GGE session calling for the group to begin negotiations of an instrument on the basis of the rolling text—signaling that a negotiating coalition is forming even as broader consensus remains elusive.²⁴

Challenges: The Consensus Model and Geopolitical Stalling

The CCW’s consensus-based decision-making model has been both its strength and its most significant structural limitation. Universal buy-in produces durable and broadly legitimate treaties; the requirement for unanimity, however, gives any single state an effective veto. In the LAWS context, this has allowed a small number of powerful states with advanced autonomous weapons programs to slow the process.

Russia’s repeated obstruction has been documented by Stop Killer Robots and acknowledged in diplomatic reporting. The US, while publicly committed to responsible AI in defense, has resisted binding prohibitions, preferring a framework of voluntary political commitments supplemented by existing IHL. China has expressed support for some prohibitions on fully autonomous lethal systems, while developing an extensive LAWS industrial base—a tension that will need to be resolved in any serious negotiation.

The consensus model may ultimately require a workaround. One precedent is the Anti-Personnel Mine Ban Treaty (Ottawa Treaty, 1997) and the Convention on Cluster Munitions (2008), both of which were negotiated outside the CCW framework through a coalition of willing states—demonstrating that determined multilateral coalitions can achieve binding agreements even when consensus forums fail.

7. Ethical and Moral Dimensions

Delegating Life-and-Death Decisions to Machines

The most fundamental ethical objection to LAWS is also the most straightforward: the decision to take a human life is a decision of such moral gravity that it should require human moral agency. Delegating targeting decisions to an algorithm—however sophisticated—constitutes a categorical shift in the nature of organized violence. It removes from the chain of lethal force the element of human moral responsibility that underpins both ethical warfare and legal accountability.

This argument does not depend on the assumption that AI targeting will necessarily be less accurate than human targeting. It rests instead on the premise that accuracy and legality are necessary but not sufficient conditions for legitimate lethal force—that human dignity requires the act of killing in war to remain a human act, accountable to human moral standards.

Human Dignity and the “Meaningful Human Control” Standard

The concept of “meaningful human control” has emerged as the central normative standard in LAWS governance debates. The term appears prominently in the CCW rolling text, the ICRC’s position papers, and diplomatic statements from dozens of states. Its core requirement is that a human being must make a genuine, informed, and contextually appropriate decision about the use of lethal force—not merely activate a system and step back.

What constitutes “meaningful” human control is itself contested. A human operator who sets targeting parameters, monitors a feed, and retains the technical ability to abort does not necessarily exercise meaningful control if engagement decisions occur faster than human reaction time allows. The ICRC has argued that meaningful control requires not only technical override capability but also the cognitive capacity to understand, evaluate, and if necessary countermand the system’s targeting logic before force is applied.²⁵

AI Bias and Accountability Gaps

AI targeting systems trained on historical data inherit the biases embedded in that data. If training datasets overrepresent certain demographic profiles as threats, or were collected in contexts of racial or ethnic conflict, the resulting systems may systematically misidentify civilians from particular communities as combatants. The CCW rolling text explicitly acknowledges the need to “mitigate biases in AI and automation by implementing measures to reduce potentially harmful bias of AI-driven decisions” and “conducting regular evaluations to detect and address harmful biases.”²⁶

The accountability question is equally pressing. If a LAWS kills a protected person in violation of IHL, who bears legal responsibility? The developer? The manufacturer? The commanding officer who activated the system? The state that deployed it? Existing command responsibility doctrine provides partial answers: commanders who knew or should have known that a weapon system was likely to be used unlawfully bear responsibility for failing to prevent its use. But the opacity of AI targeting logic creates genuine evidentiary and attributional challenges that existing frameworks are not fully equipped to handle.

8. Geopolitical Dimensions and the Arms Race

Major Power Competition: United States, China, and Russia

The LAWS arms race is, at its core, a great-power competition. All three of the world’s leading military powers are investing heavily in autonomous weapons capabilities, and all three bring different postures to international governance discussions.

The United States has articulated a policy of responsible AI in defense through directives including DoD Directive 3000.09 on Autonomous Weapon Systems, which requires “appropriate levels of human judgment over the use of force.” The US has invested in autonomous systems across air, sea, ground, and undersea domains, and has been among the more resistant major powers to binding international prohibitions—arguing that existing IHL provides an adequate framework.

China has publicly supported the prohibition of fully autonomous lethal systems at the conceptual level, while simultaneously pursuing one of the world’s largest and most sophisticated military AI programs. China’s dual posture—diplomatic constraint-and-regulate positioning alongside robust LAWS development—reflects a calculated approach to shaping the normative environment while preserving operational flexibility.

Russia has been the most openly obstructive actor in the CCW GGE process, blocking consensus positions and rejecting draft framework language. Russia’s LAWS programs—including the Lancet-3 family and the Uran-9 autonomous ground combat vehicle—are among the most combat-tested in the world, developed and refined through deployment in Ukraine.

Proliferation Risks and Non-State Actors

The proliferation of LAWS technology beyond major-state militaries represents one of the most dangerous medium-term risks in the arms control landscape. Loitering munitions—the weapon category most immediately relevant to LAWS—have proliferated to dozens of states and, via state intermediaries, to non-state armed groups. The November 2025 UNGA resolution explicitly cited the risk of “proliferation, including to unauthorized recipients and non-State actors” as a core concern motivating the call for new rules.²⁷

The entry costs for LAWS development are falling rapidly. Commercial drone technology, open-source machine learning frameworks, and widely available sensor components create pathways for non-state actors—including terrorist organizations—to develop crude but potentially deadly autonomous targeting systems. A governance framework focused exclusively on state military programs will prove inadequate to this challenge.

9. The Path Forward: Governance, Accountability, and Human Control

Elements of a Viable Treaty Framework

The case for a dedicated international treaty on LAWS has been articulated by Guterres, Spoljaric, Perrin, and a growing coalition of states and civil society organizations. The broad outlines of a viable framework are now well-established in the literature and in the GGE rolling text:

Prohibition of LAWS that autonomously target people without human involvement—a categorical prohibition on the full automation of lethal force against human beings. This is the “bright line” that the ICRC, the UN Secretary-General, and many delegations have identified as non-negotiable.

Regulation of other LAWS through requirements for:

• Predictability and reliability (the system must perform as intended across the full range of anticipated operational environments).
• Traceability and explainability (targeting decisions must be auditable and explicable to human operators and legal reviewers).
• Context-appropriate human control (the degree of human involvement must be calibrated to the legal and ethical complexity of the targeting decision).
• Operational restrictions on target types, duration, geography, and scale.
• Mandatory deactivation and self-destruct mechanisms.
• Regular bias assessment and mitigation.

Accountability mechanisms, including clear allocation of legal responsibility along the chain of command and development, supported by mandatory incident reporting and investigation procedures.

Universality and verification provisions, modeled on existing arms control treaty regimes, to ensure that the framework binds all relevant actors and provides mechanisms for compliance monitoring.

Responsible Military AI Procurement: Insights from SIPRI

The Stockholm International Peace Research Institute (SIPRI) has produced important analytical work on the intersection of military procurement and responsible AI.²⁸ SIPRI’s research establishes that the procurement process is a critical—and underutilized—governance lever. Military procurement decisions determine which AI systems enter service, under what operational constraints, and with what oversight mechanisms. Building accountability requirements and bias-testing standards into procurement specifications can operationalize governance commitments that treaty frameworks articulate but may not be able to directly enforce.

SIPRI’s work also highlights the challenge of AI literacy within military institutions: procurement decision-makers often lack the technical expertise to evaluate AI system capabilities and limitations rigorously. Developing AI-literate acquisition workforces—across allied and partner militaries—is a practical governance priority that sits alongside treaty negotiations rather than depending on them.²⁹

The Role of Civil Society: Stop Killer Robots

The Campaign to Stop Killer Robots, a global coalition of NGOs established in 2012, has been the primary civil society driver of international LAWS governance discussions. Its member organizations have contributed to the CCW process, the UNGA First Committee debates, and public awareness campaigns across more than 60 countries.

Stop Killer Robots has consistently called for a legally binding treaty that establishes meaningful human control over the use of force and prohibits fully autonomous weapons that select and engage human targets without human authorization. Following the November 2025 UNGA vote, the organization expressed disappointment that the resolution did not mandate formal negotiations, while acknowledging the significance of 156 states expressing concern about the humanitarian, legal, security, technological, and ethical dimensions of LAWS.³⁰

The campaign’s advocacy has contributed to a normative shift that would have been difficult to predict a decade ago: fully autonomous lethal weapons are now broadly acknowledged—across state delegations, international organizations, academic communities, and the defense industry itself—to require new and specific international regulation. The question is no longer whether rules are needed, but what rules, made by whom, and on what timeline.

What Governance Must Deliver Before 2027

The international community faces a narrow and closing window. The CCW GGE mandate concludes in 2026, ahead of the Seventh Review Conference. The Secretary-General and ICRC President have called for completed treaty negotiations by the end of 2026. Military AI programs in multiple major states are advancing faster than governance frameworks can track.

The argument for urgency is not theoretical. LAWS are already deployed in active conflicts. The Lancet-3 is destroying Ukrainian armored vehicles. The KARGU may have autonomously attacked human targets in Libya. Loitering munitions with varying degrees of autonomy are proliferating to state and non-state actors across the Middle East, Africa, and Asia. Each week without a binding framework is a week in which the normative vacuum is filled by unilateral state practice—and each instance of autonomous lethal engagement that goes uninvestigated and unaccounted for erodes the international legal framework that 75 years of arms control diplomacy has built.

The path forward requires three parallel tracks. First, states must commit—at the CCW Review Conference—to launching formal negotiations on a legally binding instrument, on the basis of the existing rolling text and the emerging consensus around prohibition and regulation. Second, states must strengthen national implementation of existing IHL obligations as applied to LAWS, including through rigorous Article 36 reviews and AI-literate procurement processes. Third, the broader community of states, civil society organizations, technology companies, and research institutions must sustain the political and normative pressure that has brought 156 countries to express support for international action.

The stakes could not be clearer. The delegation of lethal force to autonomous machines, without adequate legal frameworks and meaningful human accountability, represents not merely a military or technological challenge—it is a challenge to the foundations of the international humanitarian law regime, to the principle of human dignity in armed conflict, and to the aspiration, however imperfect, that war can be conducted within legal and moral limits. Meeting that challenge requires the same combination of political courage, technical expertise, and moral clarity that has produced every landmark achievement in international arms control—from the Chemical Weapons Convention to the Treaty on the Prohibition of Nuclear Weapons.

The machines are already in the field. The diplomats need to catch up.

---

Frequently Asked Questions

What is the difference between a drone and a lethal autonomous weapon system?

A drone operated by a human pilot via remote control is not a LAWS—a human being is making the targeting and engagement decisions in real time. A LAWS is distinguished by its capacity to select and engage targets independently, without real-time human authorization. Some drone platforms can operate in both modes, depending on mission parameters.

Is there currently a ban on lethal autonomous weapons systems?

No binding international ban exists. However, UNGA Resolution A/RES/79/62 (December 2024) passed with 166 states in favor and mandated informal consultations on a possible treaty. A follow-on resolution in November 2025 received 156 votes in support. The CCW GGE is developing a rolling text that could form the basis of a binding instrument. Negotiations toward a binding treaty are widely expected, with a key deadline at the CCW Seventh Review Conference in 2026.

Which countries are most advanced in developing autonomous weapons?

The United States, China, Russia, Israel, and Türkiye are among the states with the most advanced and operationally deployed LAWS programs. South Korea, the United Kingdom, France, and India are also significant developers. Commercial technology diffusion means that autonomous weapons capabilities are spreading rapidly across additional state and non-state actors.

What does “meaningful human control” mean in practice?

Meaningful human control requires that a human being exercises genuine, informed judgment over the decision to apply lethal force—not merely activates a system. At minimum, it requires that the operator understands the system’s targeting logic, has sufficient situational awareness to evaluate the legality and proportionality of a proposed engagement, and has a realistic opportunity to abort or override before force is applied. The ICRC and many states argue that engagement timescales that preclude genuine human review are inconsistent with meaningful control, regardless of technical override capability.

Can existing international humanitarian law govern autonomous weapons?

Existing IHL applies fully to LAWS—this is the unanimous position of the international community. The principles of distinction, proportionality, and precaution govern their use. However, existing IHL does not contain LAWS-specific rules, creating interpretive gaps and enforcement challenges. The ICRC, UN Secretary-General, and a growing coalition of states argue that a dedicated treaty is necessary to close these gaps and provide the clarity and specificity that effective governance requires.

What role does the ICRC play in autonomous weapons governance?

The International Committee of the Red Cross participates in the CCW GGE process as an expert observer and has been one of the most influential voices calling for new treaty law. The ICRC’s March 2026 position paper provides detailed legal analysis of the IHL challenges posed by LAWS. ICRC President Mirjana Spoljaric has jointly called with UN Secretary-General Guterres for the conclusion of treaty negotiations by the end of 2026.

What are the risks if no treaty is negotiated?

Without a treaty, LAWS development will continue on the basis of divergent national policies and varied interpretations of existing IHL. This creates risk of escalation, accidents, and civilian harm from systems whose targeting logic cannot be adequately supervised or audited. Proliferation to non-state actors will accelerate. Accountability for violations will remain legally ambiguous. The normative environment for future armed conflict will be shaped by the unilateral practice of the most militarily advanced states, rather than by multilateral legal standards.

---

Endnotes and References

1. Benjamin Perrin, “Lethal Autonomous Weapons Systems & International Law: Growing Momentum Towards a New International Treaty,” (2025) 29:1 American Society of International Law – Insights 1. Published January 24, 2025. Available at: https://asil.org/insights/volume-29-issue-1/
2. United Nations Office for Disarmament Affairs, Lethal Autonomous Weapon Systems (LAWS) (2023), https://disarmament.unoda.org/the-convention-on-certain-conventional-weapons/background-on-laws-in-the-ccw/
3. United Nations Secretary-General, Note to Correspondents: Joint Call by the United Nations Secretary-General and the President of the International Committee of the Red Cross for States to Establish New Prohibitions and Restrictions on Autonomous Weapon Systems (October 5, 2023), https://www.un.org/sg/en/content/sg/note-correspondents/2023-10-05/
4. Perrin, supra note 1.
5. Raytheon, Phalanx Weapon System, https://www.rtx.com/raytheon/what-we-do/sea/phalanx-close-in-weapon-system
6. Id.
7. Israel Aerospace Industries, HARPY Autonomous Weapon System, referenced in Perrin, supra note 1.
8. Global Defense News Army Recognition Group, Lancet-3 (July 25, 2024), https://www.armyrecognition.com/military-products/army/unmanned-systems/unmanned-aerial-vehicles/lancet-3
9. STM, KARGU Combat Proven Rotary Wing Loitering Munition System, https://www.stm.com.tr/en/kargu-autonomous-tactical-multi-rotor-attack-uav
10. Letter from the Panel of Experts on Libya Established Pursuant to Security Council Resolution 1973 (2011) to the President of the Security Council, at 63, U.N. Doc. S/2021/229 (March 8, 2021), https://documents.un.org/doc/undoc/gen/n21/037/72/pdf/n2103772.pdf
11. Perrin, supra note 1, at 4; Group of Governmental Experts on Emerging Technologies in the Area of Lethal Autonomous Weapon Systems, Existing International Humanitarian Law Applicable to Lethal Autonomous Weapon Systems, CCW/GGE.1/2024/CRP.3 (May 27, 2024).
12. International Committee of the Red Cross, Autonomous Weapon Systems and International Humanitarian Law: Selected Issues (March 3, 2026), https://www.icrc.org/en/article/autonomous-weapon-systems-and-international-humanitarian-law-selected-issues; PDF available at: https://www.icrc.org/sites/default/files/2026-03/4896_002_Autonomous_Weapons_Systems_-_IHL-ICRC.pdf
13. Additional Protocol I to the Geneva Conventions (1977), Preamble and Art. 1(2); Hague Convention (II) (1899), Preamble.
14. Perrin, supra note 1, at 5, citing ICRC, A Legal Perspective: Autonomous Weapon Systems Under International Humanitarian Law (December 3, 2024).
15. United Nations Office for Disarmament Affairs, GGE on LAWS – Revised Rolling Text as of 8 November 2024, https://docs-library.unoda.org/Convention_on_Certain_Conventional_Weapons_-Group_of_Governmental_Experts_on_Lethal_Autonomous_Weapons_Systems_(2024)/Revised_rolling_text_as_of_8_November_2024_final.pdf
16. Stop Killer Robots, First CCW Meeting After UNGA Resolution Maintains Status Quo (November 15, 2024), https://www.stopkillerrobots.org/news/first-ccw-meeting-after-unga-resolution-maintains-status-quo/
17. G.A. Res. 79/62, Lethal Autonomous Weapons Systems, U.N. Doc. A/RES/79/62 (December 2, 2024), https://documents.un.org/doc/undoc/ltd/n24/305/45/pdf/n2430545.pdf; Perrin, supra note 1, at 1.
18. Stop Killer Robots, 156 States Support UNGA Resolution on Autonomous Weapons (November 6, 2025), https://www.stopkillerrobots.org/news/156-states-support-unga-resolution/
19. Id.
20. Id.
21. ICRC, supra note 12.
22. United Nations, UN Secretary-General, President of International Committee of Red Cross Jointly Call for States to Establish New Prohibitions, Restrictions on Autonomous Weapon Systems (October 5, 2023), https://press.un.org/en/2023/sg2264.doc.htm
23. Id.
24. Stop Killer Robots, September 2025 GGE Joint Statement, https://stopkillerrobots.org/news/september-2025-gge-joint-statement/
25. ICRC, supra note 12.
26. UNODA Rolling Text, supra note 15.
27. Stop Killer Robots, supra note 18.
28. Stockholm International Peace Research Institute, Responsible Procurement of Military Artificial Intelligence (2026), https://www.sipri.org/publications/2026/other-publications/responsible-procurement-military-artificial-intelligence
29. SIPRI, Responsible Behaviour in Military AI Starts with Procurement (2025), https://www.sipri.org/commentary/essay/2025/military-ai-responsible-procurement
30. Stop Killer Robots, supra note 18.

---

Meta Data

Meta title
Autonomous Weapons & AI: Law, Ethics & Governance

Meta description
A comprehensive guide to lethal autonomous weapons systems: how they work, what international law says, and what a binding treaty framework needs to include.