Autonomous Weapons Policy: International Negotiations on Lethal AI Systems

Executive Summary

UN Secretary-General Antonio Guterres set 2026 as the target year for a legally binding treaty prohibiting lethal autonomous weapons systems (LAWS), a deadline that expired on July 6, 2026, at the precise moment he arrived in Geneva for the inaugural UN Global Dialogue on AI Governance. That deadline passed without a treaty, and formal negotiations have never begun, only consultations. The governance gap is not a product of disinterest: supermajority voting blocs at the UN General Assembly have repeatedly endorsed restriction frameworks. The blockage originates from a small coalition of technologically advanced military powers that have used procedural leverage within the Convention on Certain Conventional Weapons (CCW) to prevent formal negotiation from ever opening.

This structural stalemate has direct implications for defense investors, technology companies operating in the autonomous systems market, and governments setting AI policy, because the absence of binding law leaves the field governed by national military doctrine rather than international accountability.

• Defense-sector risk officers: Monitor the CCW Seventh Review Conference (late 2026) and any shift by the United States from its "existing IHL is sufficient" position, as either would materially change liability exposure for autonomous-systems contractors.
• Technology companies: The Anthropic-Pentagon dispute unsealed in July 2026 illustrates that internal corporate guardrails on autonomous lethal use now occupy the governance space that treaty law has failed to fill; companies with dual-use AI capabilities should map their autonomous-weapons exposure ahead of any legal shift.
• Policy advisors: The UNGA track, not the CCW, is now the forum where a binding instrument is most moderate-to-high confidence to originate; resource engagement there accordingly.

The 2026 deadline passed without a treaty, and the realistic near-term path to binding law has narrowed to a General Assembly process that major military powers are still contesting.

The Architecture Of Non-Governance: Ccw Vs. Unga

The international governance framework for LAWS currently rests on three interlocking but inadequate pillars: existing international humanitarian law (IHL), the CCW process, and non-binding voluntary norms. The American Society of International Law notes that existing IHL, including international criminal law and international human rights law, provides foundational principles that technically apply to LAWS, but that gaps in specificity and enforcement highlight the pressing need for a dedicated treaty. No specific protocol governing autonomous weapons exists within the CCW framework. The UN Office for Disarmament Affairs confirms that as of May 2026, the CCW has 128 states parties, but its Protocols addressing existing weapons categories do not cover LAWS.

In 2013, CCW states parties agreed to begin discussing autonomous weapons. From 2014 to 2016, they held informal meetings of experts, and in 2017 established the Group of Governmental Experts on Lethal Autonomous Weapons Systems (GGE on LAWS), which has met regularly since. After nine years of GGE meetings, the group is mandated to produce a set of "elements of an instrument" but has been explicitly prevented from beginning formal negotiations. The Reaching Critical Will monitoring organization reports the CCW GGE held its first 2026 session from March 2-6 in Geneva, and its second session is scheduled for August 31 to September 4, 2026, feeding into the Seventh Review Conference. The UN Office for Disarmament Affairs confirms the GGE's current mandate requires it to submit a report to that Review Conference.

The UNGA track represents a structurally different avenue. Unlike the CCW, which requires consensus, the General Assembly operates by majority vote. On December 2, 2024, 166 countries voted in favor of Resolution 79/62 on lethal autonomous weapons systems, while 3 voted no and 15 abstained. The resolution creates a new forum under UN auspices to discuss challenges and concerns raised by autonomous weapons systems. The 2025 follow-on resolution, Resolution 80/57, passed with a similarly wide margin. However, both resolutions stopped short of mandating formal treaty negotiations, a choice that reflected continued pressure from the United States and other major powers opposed to binding law.

What is not being reported: Public coverage of the LAWS debate focuses almost entirely on the CCW process and its failures. The UNGA First Committee track, where procedural momentum now actually lives, receives significantly less analytical attention. This creates a false picture of a frozen diplomatic situation when in fact the forum of action has shifted.

The Blocking Coalition: Positions, Interests, And Fracture Lines

The states opposing binding restrictions are not a monolithic bloc, and understanding where they diverge is as important as understanding what currently aligns them. According to Automated Decision Research, the monitoring team of Stop Killer Robots, the states opposing a binding instrument include the United States, Russia, Israel, India, Japan, South Korea, Australia, the United Kingdom, Belarus, Estonia, Poland, and North Korea.

The United States has consistently argued, per the Congressional Research Service, that existing IHL is sufficient, that LAWS could provide humanitarian benefit through more precise targeting, and that human operators remain accountable for deployment. A Trump administration national security memo, reported by Gizmodo in July 2026, accused previous administrations of implementing "undue bureaucracy" around autonomous weapons AI development, signaling continued US opposition to restrictions. The US has, however, endorsed non-binding frameworks: the OECD's AI Recommendations and NATO's Principles of Responsible Use of AI in Defense, referenced by Trends Research Group, represent the category of soft-law instruments the US prefers.

Russia's position is more categorical. The Russian delegation has, per the Congressional Research Service, rejected the term "meaningful human control" entirely and opposed any preemptive ban. Civil society observers, noted by the American Society of International Law, have alleged Russian stalling tactics within the CCW process. Russia was one of only three countries voting against the 2024 UNGA Resolution 79/62, alongside Belarus and North Korea, placing it in a significantly more isolated position than the US, which abstained on some votes.

China's posture is the most analytically interesting. Beijing distinguishes between "acceptable" and "unacceptable" LAWS, per the Congressional Research Service, defining unacceptable systems as those that are simultaneously lethal, autonomous, incapable of termination, capable of indiscriminate killing, and capable of autonomous learning. This five-part definition is narrow enough that it would exempt most operational systems either country currently develops. China abstained on the 2024 UNGA vote and on the 2025 vote, placing it in a different category from Russia but still outside the majority coalition pressing for binding restrictions. The May 2025 China-Russia joint statement on global strategic stability, published by China's Ministry of Foreign Affairs, noted the two governments' interest in coordinating on "military application of artificial intelligence technologies" in multilateral venues, which the CCW GGE process provides.

Coalition fracture point: The blocking coalition is not unified by common doctrine. The UK and Australia, both in the opposing group, have participated in joint working papers within the CCW GGE alongside states calling for regulated restrictions, suggesting they occupy a softer obstructionist position than Russia or Israel. The Arms Control Association notes that geopolitics and the pace of domestic AI military investment are the primary drivers shaping each country's negotiating posture, and those interests diverge significantly between, say, Australia and North Korea.

Why Battlefield Deployment Is Outpacing Diplomatic Constraint

The governance challenge is not purely diplomatic; it is structural. Technology deployment in conflict zones is generating precedents faster than negotiating cycles can accommodate. According to Defense One's July 2026 documentary analysis, Ukraine has combined air and ground drones, autonomy, and swarming in the spring and summer of 2026 to recapture territory with reduced human decision-making at the engagement level. The operational effectiveness of these systems creates military-industrial incentives that translate directly into negotiating resistance.

On July 9, 2026, the US Department of Defense awarded Lockheed Martin a Joint Laser Weapon System contract to build a 500 kW containerized laser weapon, with a total program ceiling of $847 million, per The Defence Blog. The Pentagon also consolidated all Unmanned and Autonomous Systems oversight in July 2026, per Reuters investment analysis, signaling accelerated procurement rather than restraint. These procurement decisions, taken together, demonstrate that the US is institutionally accelerating autonomous weapons development while simultaneously blocking the treaty process that would regulate it. This dynamic, accelerating capability alongside blocking governance, is precisely what Human Rights Watch warned would normalize autonomous killing and make future restrictions harder to impose.

Capability without confirmed intent: The US, Russia, and China all possess demonstrated and growing capabilities to deploy LAWS at scale. What remains unverified is whether any of these governments has decided, at the political level, that the absence of a binding treaty is their preferred long-term equilibrium or a negotiating position they would abandon under sufficient diplomatic pressure. Conflating capability with settled intent would cause analysts to over-predict the permanence of the current blockage.

This military-technology pressure translates directly into the diplomatic domain: as more systems approach fully autonomous operation in active conflict, the definitional task confronting treaty negotiators becomes harder, because every new system introduced creates a new category that must be either prohibited, regulated, or grandfathered.

Counterarguments

1. The blocking coalition argument overstates US intent and understates negotiating space: A credible counter-reading holds that the US, UK, and allied positions are not absolute vetoes but rather opening positions in a negotiation that has never formally begun. The US submitted a working paper to the CCW GGE in 2025, CCW/GGE.1/2025/WP.6 on "The Autonomy Aspect of LAWS," and co-signed a joint paper on draft articles on autonomous weapons prohibitions and regulations based on IHL alongside Australia, Canada, Japan, South Korea, and the UK (CCW/GGE.1/2025/WP.7). That joint paper proposed specific prohibitions on systems incapable of IHL compliance, which is substantively closer to a regulatory framework than to a categorical rejection of all international rules. Analysts who treat the US as an unambiguous blocker may be importing the CCW procedural record onto a different negotiating question.

2. The "no treaty exists" framing understates the constraining effect of existing norms: The ASIL notes that existing IHL, including rules on distinction, proportionality, and precaution derived from the Additional Protocols to the Geneva Conventions, legally applies to LAWS right now. The CCW GGE has, per Reaching Critical Will, produced a "rolling text" incorporating these principles in operational terms. The Guiding Principles endorsed by CCW High Contracting Parties affirm that IHL "continues to apply fully to all weapons systems, including the potential development and use of lethal autonomous weapons." A state that deploys a LAWS that kills civilians indiscriminately is already in violation of existing treaty law; the question is whether a dedicated instrument would add enforceable specificity. Overstating the governance vacuum could inadvertently argue for urgency that the existing IHL framework partially addresses.

3. Civil society monitoring introduces selection bias in how state positions are characterized: Organizations including Human Rights Watch, Stop Killer Robots, and Reaching Critical Will generate the most granular public record of CCW and UNGA deliberations. These organizations are advocacy actors, and their framing of states as "blockers" or "supporters" collapses nuanced negotiating positions into binary categories. India, for example, abstained on some votes and submitted papers engaging constructively with regulatory frameworks, while being categorized alongside Russia as a blocker by advocacy groups. This characterization shapes the analytical picture available to outside observers and may obscure paths to partial agreement.

Indicators To Watch

Near-term watch list: (1) CCW Seventh Review Conference, scheduled for late 2026, will determine whether nine years of GGE work produces any actionable mandate or is simply noted and shelved; this is the single most consequential near-term milestone for the CCW track. (2) UNGA First Committee session, November 2026, where Austria and co-sponsors will moderate-to-high confidence introduce a fourth consecutive LAWS resolution; the question is whether this year's text moves from "encouraged exchanges" to mandated formal negotiations. (3) US DoD Directive 3000.09 review, which governs American military requirements for human control in autonomous weapons; any revision weakening the human-control would signal a US doctrinal shift that makes its CCW and UNGA positions harder to negotiate around.

Decision Relevance

Scenario A (~55%): CCW produces limited elements document; UNGA begins formal consultations but not formal treaty negotiations. If you advise on corporate compliance for AI or defense technology companies, this scenario means the regulatory landscape remains patchwork national doctrine for at least two more years; map your products against IHL requirements now, because national liability standards, not international law, will be the operative risk for the foreseeable future. If you are in the autonomous systems supply chain, this scenario does not produce new legal obligations but increases reputational and contractual risk as corporate clients, particularly following the Anthropic-Pentagon dispute unsealed in July 2026, begin imposing their own human-control standards.

Scenario B (~30%): CCW Seventh Review Conference and UNGA vote together generate a formal negotiating mandate, launching treaty talks outside the CCW consensus requirement. If you hold positions in defense technology companies with significant LAWS exposure, treat this as the scenario requiring hedged action now: begin scenario-planning which product categories would face prohibition versus regulation under the Austria-led two-tier framework, and assess which NATO-aligned customers would need alternative capability. If you advise governments on weapons procurement, this scenario means acquisition contracts for fully autonomous offensive systems signed in 2026-2027 could face treaty compliance challenges within five years.

Scenario C (~15%): No meaningful movement; CCW, UNGA, and bilateral channels all stall; proliferation continues without international law. If you are a policy researcher tracking arms control, this scenario is not merely diplomatic failure; it produces the conditions the Lieber Institute describes as unconstrained proliferation, with China's "intelligentized warfare" doctrine, Russian loitering munitions use in Ukraine, and Kargu-2 autonomous engagement in Libya setting doctrinal precedents that become the de facto international . The Arms Control Association notes this outcome mirrors the pattern before the Ottawa landmines process and before the Chemical Weapons Convention: the worse the humanitarian record, the more political space eventually opens for a treaty.

Expert Integration

Expert Consensus Assessment

Legal scholars, including Benjamin Perrin at the University of British Columbia writing for the American Society of International Law, and policy researchers at the Lieber Institute at West Point converge on two points: no binding LAWS treaty exists, and existing IHL technically applies but lacks enforcement specificity. They diverge on whether that gap is practically significant and whether the CCW or UNGA is the more viable forum.

Expert Disagreement Areas

• Adequacy of existing IHL: The Congressional Research Service and the Lieber Institute assess existing IHL as providing a workable framework requiring national implementation, while Human Rights Watch and ICRC assess it as creating accountability gaps that only dedicated treaty law can close.
• Probability of a binding treaty: The Lieber Institute's March 2026 assessment characterized the probability as "slim to none" given great-power opposition, while the ASIL piece from January 2025 characterized momentum as "building" toward a new treaty. The same evidence base produces opposite forecasts depending on whether analysts weight the majority coalition's preferences or the blocking coalition's procedural leverage.
• China's effective position: Defense analysts drawing on CRS and Lieber Institute sources treat China as a practical blocker given its abstention and narrow "unacceptable LAWS" definition. Civil society monitors, drawing on UNGA voting records, place China in a softer category than Russia and the US.

Systematic-Expert Alignment

Alignment: MIXED

This analysis aligns with the Lieber Institute's near-term pessimism on the CCW track while also crediting the ASIL and HRW assessment that the UNGA track has structural momentum that could produce a formal negotiating mandate. The divergence from pure pessimism rests on the observable fact that each successive UNGA resolution has moved incrementally closer to mandating formal talks, and the 2026 Review Conference creates a decision point that did not previously exist.

• The negotiating positions of Russia, China, and India inside closed CCW sessions are reported primarily through civil society monitoring organizations with advocacy mandates; their characterizations of "stalling tactics" or "blocking behavior" reflect a particular interpretive frame that cannot be independently verified against diplomatic records.
• No classified intelligence is available on whether the US, Russia, or China has reached internal government decisions about the long-term acceptable governance outcome for LAWS; the stated positions reported through CCW submissions and UNGA votes may not reflect ultimate red lines.
• The CCW GGE's second 2026 session (August 31 to September 4) and the Seventh Review Conference will produce outcomes that may materially change this assessment; this analysis cannot account for those proceedings.
• The operational definition of "meaningful human control" remains unresolved in any international forum, which means that the question of what counts as prohibited autonomy is itself contested; assessment of what a treaty would actually ban depends on definitional outcomes that have not been reached.
• Civil society organizations including Stop Killer Robots and Human Rights Watch generate the most detailed open-source record of CCW and UNGA deliberations; this creates a selection-bias risk where the analytical framing of state positions reflects advocacy priorities rather than neutral diplomatic history.