Autonomous Weapons Policy: International Negotiations on Lethal AI Systems

Executive Summary

No binding international treaty governs lethal autonomous weapons systems today. Although states broadly agree that international humanitarian law applies to such systems, specific internationally binding standards remain largely absent, and a key structural divide has hardened, with Russia and the United States opposing new legally binding instruments while a growing majority argues new rules are necessary. The CCW's Seventh Review Conference, scheduled for 16 to 20 November 2026 at the Palais des Nations in Geneva, is the most consequential diplomatic deadline in a decade of talks. Whether that conference produces a negotiating mandate or another deferred outcome will determine whether international law catches up to weapons already deployed in active conflicts. The underlying tension is between states that view autonomous weapons as strategically essential assets requiring maximum operational flexibility, and a supermajority consensus that existing international humanitarian law is insufficiently specific to govern these systems and that new binding rules are urgently necessary before proliferation becomes irreversible.

The Architecture Of Non-Governance: Why A Decade Produced No Treaty

International law on conventional weapons rests on a framework adopted in 1980. The CCW, adopted in 1980, has long sought to limit weapons with indiscriminate or excessively injurious effects, such as landmines and blinding lasers. LAWS fit within that humanitarian mandate on paper, but the forum was never designed for a technology moving at the pace of artificial intelligence.

Discussions on LAWS formally began in 2013, when the Meeting of State Parties to the CCW agreed on a mandate to discuss questions related to emerging technologies in the area of lethal autonomous weapons systems in the context of the objectives and purposes of the Convention. Three rounds of expert meetings followed in 2014, 2015, and 2016, producing no binding output. In 2018 and 2019, the GGE identified and adopted eleven principles to guide future work, reaffirming the relevance of international humanitarian law to such technologies, a reaffirmation of existing law rather than the creation of new rules.

Despite years of debate on killer robots that began in 2013, when Christof Heyns, a UN special rapporteur, recommended establishing national moratoriums on lethal autonomous robotics, there is still no legal global framework to regulate or prohibit the development and deployment of LAWS.

The structural explanation is straightforward. The CCW operates by consensus, meaning any one state can block progress. Key challenges include definitions, what exactly constitutes an "autonomous weapon"?, and scope: should the outcome be an outright ban on certain systems or a framework of restrictions and regulations? Both questions are politically charged because they determine which weapons currently in development or deployment would be affected.

The interplay between arms race dynamics and regulatory paralysis is mutually reinforcing. The United States leads in sophisticated, high-intelligence autonomous systems. For fiscal year 2026, the Pentagon requested a record 4.2 billion for AI and autonomous research, and the Replicator program received billion in 2025 to fast-track the deployment of thousands of expendable autonomous drones and surface vessels. These capital commitments translate directly into institutional resistance to binding prohibitions, since any treaty would constrain programs already funded and in production. Both economic and strategic dimensions of this dynamic make near-term agreement structurally difficult.

The Blocking Coalition: Strategic Logics Behind Opposition

The five states that voted against the November 2025 UNGA resolution, and the broader informal coalition resisting binding rules in Geneva, are not monolithic in their reasoning.

The United States frames its resistance around preserving technological superiority and operational flexibility. The U.S. focus is on superior software and algorithms that allow weapons to operate in environments where GPS and communications are completely severed, a capability that depends on high-end semiconductors that the U.S. and its allies closely control. The U.S. Mission to the UN in Geneva noted concerns that the resolution endorses informal consultations that were not helpful, and proposes to take up the resolution again next year when the CCW process will be ongoing and the outcome of the CCW 7th Review Conference will be unknown, a framing that prioritizes process integrity over outcome urgency.

Russia combines operational deployment with procedural obstruction. Russia's strategy is driven by the immediate need for survivable technology on the front lines. Their primary autonomous platform is the S-70 Okhotnik-B ("Hunter"), a heavy stealth drone designed for autonomous deep-strike missions. The ASIL notes that civil society observers have questioned the ability of the CCW process to culminate in consensus on new international rules for LAWS, citing alleged stalling tactics by the Russian Federation. Stop Killer Robots documents that a 2023 CCW meeting saw more than a whole day lost because Russia blocked confirmation of the Rules of Procedure, forcing the meeting to operate informally and resulting in observers, international organisations and civil society being excluded.

Israel occupies a distinct strategic position among resistors. In late 2025, Israel accelerated deployment of the "Iron Beam," a laser system that uses autonomous targeting to neutralize incoming threats at a speed no human operator could match. Israel abstains from a total ban because its national survival is increasingly tied to these defensive, human-out-of-the-loop systems. To Israel, a restrictive treaty is not just a legal hurdle; it is a potential threat to its ability to defend its citizens from high-speed modern attacks.

India supports the GGE process but resists binding commitments. Carnegie Endowment research examining India's position notes concerns among countries that the emergence of multiple parallel processes could fragment the normative and regulatory debates on LAWS, hindering consensus on common approaches to regulate them. India's preference is for the CCW track, which it can navigate under consensus rules, over UNGA processes where it carries less structural weight.

The broader security and technology implications of this coalition extend beyond arms control. The interplay between military AI investment and diplomatic resistance spills into export control regimes, cybersecurity vulnerability, and the proliferation of these systems to non-state actors. Without robust international export controls and tracking, there is a high risk that nonstate actors in Africa could buy and misuse autonomous weapons, further complicating peace efforts.

The Pushing Coalition: What Treaty Supporters Actually Want

More than 120 countries support calls to negotiate a treaty that prohibits and regulates autonomous weapons systems. Their positions are not uniform, but the core architecture is broadly shared.

The ICRC, civil society groups, and many governments say the treaty should prohibit autonomous weapons systems that operate without meaningful human control or that target people. The UN Secretary-General has pressed this position repeatedly. He has stated that human control over the use of force is essential and that we cannot delegate life-or-death decisions to machines. In his New Agenda for Peace, the Secretary-General laid out recommendations on the governance of AI and autonomous weapons systems, including calling for a legally binding treaty to prohibit LAWS that function without human control or oversight, to be concluded by 2026.

The regional diversity of the treaty-seeking coalition matters analytically. The September 2025 joint statement in favor of negotiations was signed by states from across regions, including Austria, France, Germany, Mexico, New Zealand, Pakistan, South Africa, Spain, Sweden, Switzerland, and Uruguay, among others. This is not a Global South bloc or a European bloc, it spans both.

The accountability question is among the treaty supporters' strongest arguments and proves difficult for resistors to counter directly. Austria, Costa Rica, El Salvador, Guatemala, and other countries raised concerns about the accountability gap created by autonomous weapons systems, noting that it is unclear who could be held legally responsible if such systems violate international humanitarian law. There are obstacles to holding individual operators criminally liable for the unpredictable actions of a machine they cannot understand, and significant legal challenges to finding programmers and developers responsible under civil law.

The broader systemic implications also extend to financial stability and cybersecurity. As Amnesty International's Matt Mahmoudi notes, "AI-powered technologies are often highly invasive, biased and discriminatory and lack the ability to parse contexts," risking mass violations of international law at scale. Taken together, the humanitarian case and the systemic security case are mutually reinforcing arguments for binding governance.

The November 2026 Chokepoint And What Comes After

The Seventh Review Conference of the CCW will take place from 16 to 20 November 2026 at the Palais des Nations in Geneva. The GGE is scheduled for two final sessions before that conference: from 2-6 March 2026 and from 31 August to 4 September 2026.

From 2 to 6 March 2026, the group held its penultimate session under the group's three-year mandate. Delegations discussed the chair's rolling draft text, circulated in December 2025, on elements of a possible instrument or other measures concerning lethal autonomous weapon systems. The Lieber Institute at West Point and SIPRI have both engaged the March 2026 session, with SIPRI presenting findings on responsible procurement of military AI, noting that deliberations within the CCW GGE on LAWS have centered on the conditions under which autonomous weapons can be developed and used consistently with international humanitarian law.

The RSIS assessment of the structural pathways is sobering. If the international community wants a legally binding instrument, the GGE would first need to reach consensus on the elements. After that, the GGE would need to consider how to incorporate these elements into its report for the 2026 Seventh Review Conference. The GGE could incorporate the elements in a few ways, depending on whether it reaches consensus. Since the CCW forum, including its review conference and the GGE, works on a consensus basis, minority-opposed elements would not be adopted over objection.

This structural constraint creates a parallel track dynamic. The UN General Assembly provides a more inclusive forum that allows more thorough consideration of the human rights, ethical, security, and other concerns raised by autonomous weapons systems compared to the CCW, which has focused narrowly on military, legal, and technological aspects under international humanitarian law. The CISS Pakistan analysis notes that 2026 represents the final year for negotiation before the Seventh CCW Review Conference in the November 2026 sessions.

Securitization Theory Analysis

Securitizing Actor: Multiple actors have moved simultaneously to securitize LAWS. The UN Secretary-General, the ICRC, and a coalition of states led by Austria have been the primary norm entrepreneurs framing autonomous weapons as an existential governance crisis. Civil society organizations including Stop Killer Robots serve as amplification actors, not primary securitizers.

Referent Object: The referent object varies by actor. Treaty-seeking states frame the threat around civilian populations and the principle of human dignity. The ICRC frames it around compliance with international humanitarian law and the laws of armed conflict. Military powers opposing restrictions frame their own referent object as national security and the capacity for self-defense.

Existential Threat Construction: The Secretary-General's repeated framing, describing LAWS as "politically unacceptable, morally repugnant and should be banned", represents the most explicit securitizing speech act from an institutional actor. He has stated that "time is running out to take preventive action" and called for urgent action to prevent the normalization of autonomous weapons deployment. The Usanas Foundation frames the current moment as "the pre-proliferation window, the final moment in history before these weapons become as common and unmanageable as small arms," a framing that constructs urgency by analogy to a proven governance failure.

Target Audience: The primary audience for the securitizing moves is the group of states sitting in abstention, particularly China, which voted against earlier resolutions and abstained on the November 2025 vote. Persuading this middle group is the functional objective of the diplomatic campaign.

Extraordinary Measures: The shift from the CCW to UNGA track represents the operational extraordinary measure: bypassing the consensus veto by moving to a two-thirds vote forum. This is structurally analogous to how the Ottawa Process bypassed the CCW to produce the landmine ban treaty in 1997.

The existential framing has achieved sufficient acceptance among a supermajority of states to justify emergency diplomatic processes. The November 2025 UNGA vote (156-5) demonstrates that the securitization has been accepted by audiences capable of acting. The blocking coalition's resistance does not desecuritize the issue; it positions those states as the source of risk rather than responders to it.

Constructivism Lens Analysis

Actor Identities: Treaty-seeking states project the identity of norm guardians, states that see compliance with international humanitarian law as constitutive of sovereign legitimacy. Major military powers project the identity of security providers whose military capability is inseparable from their identity as great powers. The identity gap is not merely strategic; it is constitutive, meaning the two groups are arguing from different premises about what kind of actor a responsible state is.

Operative Norms: The norm of "meaningful human control" over lethal force has emerged as the central contested element. Discussions at the GGE have demonstrated that states share common ground on the need for a two-tier approach of prohibitions and regulations, on the need for meaningful human control, and on many basic elements of such control, suggesting the norm has achieved conceptual acceptance even where its operationalization is contested. The existing norm of IHL compliance provides the enabling framework within which LAWS governance arguments are made.

Intersubjective Meaning: The sharpest contested meaning is whether autonomous systems represent a qualitatively new legal category requiring new rules, or merely existing weapons subject to existing law. The American Society of International Law notes that while existing international humanitarian law, international criminal law, and international human rights law provide foundational rules and principles, gaps in specificity and enforcement highlight the need for a dedicated international treaty, a framing that accepts the existing-law argument while acknowledging its insufficiency.

Norm Lifecycle Stage: The norm of meaningful human control over lethal force decisions sits at the cascade stage, it has passed its tipping point in terms of state adoption of the concept, even though binding codification has not followed. The critical need for new rules on autonomous weapons is now firmly on the international agenda, with growing global awareness of the issue. The norm of a specific treaty prohibition on fully autonomous weapons remains at the emerging stage, blocked from cascade by the structural opposition of major military powers.

(for the broader norm of human control) / EMERGING (for specific prohibitory treaty law)

Process Tracing Analysis

Cause and Outcome: The cause under examination is the CCW's consensus requirement. The outcome is the absence of a binding treaty on LAWS after over a decade of formal negotiations.

Causal Mechanism Chain:

• Step 1: The CCW's consensus rule is embedded in its procedural framework, requiring all states to agree before any outcome is adopted.
• Step 2: Major military powers, the U.S., Russia, Israel, India, have strategic and material interests in preserving autonomous weapons programs.
• Step 3: These states exercise the consensus veto selectively, accepting conceptual discussions while blocking progress toward negotiating mandates.
• Step 4: Blocking tactics operate at the procedural level, blocking Rules of Procedure, insisting on informal modes, diluting text in final sessions.
• Step 5: The GGE produces non-binding outputs (principles, rolling text) that satisfy the form of deliberation without its substance.
• Step 6: The absence of binding rules allows weapons development to accelerate, increasing the future cost of compliance and further entrenching resistance.

Evidence Assessment:

• Hoop test (passed): Russia's blocking of the 2023 CCW Rules of Procedure must be present for procedural obstruction to explain the stall, it is documented.

• Smoking gun (present): These negotiations have been consistently blocked by a handful of heavily militarised states, including Russia, Israel, India, Australia, the Republic of Korea, and the United States. These states have adopted several actions that have been detrimental to the majority's wish of negotiating an international treaty.

• Straw-in-the-wind (consistent): The vote margin (156-5) against binding restrictions is consistent with the mechanism but could also reflect genuine disagreement on definitions.

Counterarguments

1. The Ottawa Process analogy is weaker than treaty supporters claim. The 1997 landmine treaty succeeded because a coalition of medium powers, led by Canada, organized a conference outside the CCW entirely and produced a treaty that major powers later joined after normative pressure mounted. The LAWS case differs: landmines were politically embarrassing to defend; autonomous weapons are actively celebrated as strategic assets by their developers. The public opinion dynamics that created room for the Ottawa Process are largely absent in the LAWS debate. Treaty momentum in UNGA may not translate into the kind of reputational pressure that brought reluctant states into the landmine regime.

Indicators To Watch

Decision Relevance

Scenario A (~55%): CCW Review Conference produces a negotiating mandate, but only for a weak regulatory framework. The November 2026 conference agrees to begin formal negotiations, but the blocking coalition's participation comes at the price of narrow scope, covering only fully autonomous weapons targeting human beings, exempting defensive systems like missile defense, and leaving major definitional questions to future negotiation. Recommended for risk managers: engage with the rule-setting process now, because any resulting treaty's definitions and exemptions will shape national procurement requirements and export control regimes for a decade. Organizations supplying defense technology to NATO or allied partners should map their product lines against the rolling text's working characterization of LAWS before the August 2026 GGE session.

Scenario B (~30%): The CCW process deadlocks at the Review Conference and the UNGA track activates as the primary venue. The Seventh Review Conference fails to produce a negotiating mandate due to US-Russia opposition. A coalition led by Austria, Brazil, and key European states convenes a standalone conference in 2027 modeled on the Ottawa Process. Recommended: organizations operating in jurisdictions that join a standalone treaty face first-mover compliance obligations. Defense procurement officers and legal counsels in signatory states should begin internal reviews of autonomous systems in development pipelines.

Scenario C (~15%): The diplomatic process collapses entirely with no agreed outcome and no alternative venue. The Review Conference fails, no parallel process gains traction, and LAWS governance reverts to ad hoc IHL application, leaving interpretation entirely to national governments and militaries. Recommended: investors in defense AI should price in maximum regulatory uncertainty as the default. Organizations conducting due diligence on defense AI companies should treat the absence of binding international rules as a long-tail liability, not as a green light.

• The positions of Russia and China inside the GGE are assessed primarily through open-source reporting on public statements and voting patterns. Internal deliberations and any private diplomatic signaling between major powers are not available to this analysis, meaning the true flexibility (or inflexibility) of the blocking coalition may differ from publicly stated positions.
• The rolling text itself, as circulated in December 2025, is the primary document for assessing what a potential treaty might cover. If the August 2026 session produces substantive revisions, particularly on the definition of "meaningful human control", the analysis of scope and compliance obligations would require revision.
• Active conflict deployment data on autonomous systems in Ukraine and Gaza remains partial. Confirmed use of fully autonomous (human-out-of-the-loop) target selection has not been independently verified, which means the humanitarian argument for urgent treaty-making rests on probable but unconfirmed operational scenarios.
• The analysis does not capture bilateral agreements or technical standards emerging from the US-led Political Declaration on Responsible Military Use of Artificial Intelligence, which represents a parallel non-CCW governance track. If that declaration gains traction as an alternative, it could redirect diplomatic energy in ways this assessment does not model.