The Global AI Governance Structure
This document establishes the overarching governance architecture for artificial intelligence as applied to weapons systems, arms control verification, treaty compliance monitoring, and strategic stability. It defines the institutional framework, decision-making hierarchy, and normative principles that govern the fourteen companion instruments across four tiers — global institutions, anchored by the International Agency for AI Safety and Benefit (IAASB); standards and verification; active promotion of beneficial AI; and adaptive long-horizon governance — together with four non-negotiable pillars that every companion instrument answers to: human dignity, ecological integrity, democratic accountability, and epistemic humility. It articulates the foundational commitment of signatory states to transparency, proportionality, and human oversight in all AI-assisted military and arms control functions, and establishes the legal relationship between this Structure and existing international law, including the UN Charter, the Geneva Conventions, and applicable arms control treaties.
Successor to New START (expired 5 February 2026) — Draft for multilateral negotiation
Successor to New START (expired 5 February 2026) — Draft for multilateral negotiation. This instrument establishes a successor framework to the New Strategic Arms Reduction Treaty, which expired on 5 February 2026 without renewal. It extends and deepens verified limits on strategic nuclear warheads and delivery systems — ICBMs, SLBMs on deployed submarines, and heavy bombers — between the United States and the Russian Federation, while introducing provisions for the inclusion of additional nuclear-armed states as observers and prospective signatories. The treaty incorporates AI-assisted verification protocols governed by the Global AI Governance Structure, mandates continuous data exchange on strategic force postures, establishes a Joint Verification Commission empowered to investigate anomalies flagged by automated monitoring systems, and directs the IAASB to coordinate with the Comprehensive Nuclear-Test-Ban Treaty Organization’s monitoring infrastructure to corroborate flagged anomalies independently. It expressly excludes several categories of new delivery and warfare technology from its central limits — sub-strategic nuclear weapons, offensive cyber capabilities directed at nuclear command-and-control, and conventional precision-strike systems including hypersonic glide vehicles — recording the basis for each exclusion and indicative pathways toward their eventual governance. Seabed-deployed launch systems and autonomous underwater nuclear delivery vehicles, including Russia’s Poseidon/Status-6, fall outside this Treaty’s scope entirely and are instead governed by the companion Sea-Based Launch Treaty.
This instrument re-establishes verified prohibitions on ground-launched ballistic and cruise missiles with ranges between 500 and 5,500 kilometres, rebuilding the arms control framework dismantled by the mutual withdrawal of the United States and the Russian Federation from the 1987 Intermediate-Range Nuclear Forces Treaty. It extends the prohibition to all signatory states, addresses the gap created by the exclusion of China and other missile-capable states from the original INF regime, and incorporates AI-assisted detection and attribution mechanisms to identify prohibited development, testing, or deployment. The treaty establishes a dedicated verification body, defines permissible dual-use research exceptions, and creates a structured dispute resolution process for contested attribution findings.
Drone Governance Treaty, Draft
This instrument establishes a comprehensive international legal framework governing the development, testing, deployment, transfer, and use of unmanned aerial, surface, and underwater systems in contexts with potential military application. It defines categories of regulated systems by autonomy level, payload capacity, and operational range, and establishes binding obligations on signatory states regarding human oversight requirements for lethal engagement decisions. The treaty prohibits fully autonomous lethal systems lacking meaningful human control, mandates registration and identification protocols for dual-use unmanned platforms, and creates an international inspection regime for declared unmanned systems programmes. AI-assisted monitoring provisions are integrated with the Global AI Governance Structure’s verification architecture.
The Algorithmic Warfare Treaty, Draft
This instrument governs the use of artificial intelligence and algorithmic decision-support systems in the conduct of armed conflict, military planning, and targeting processes. It establishes binding standards for human oversight of AI-assisted targeting recommendations, prohibits the delegation of lethal engagement authority to fully autonomous systems, and requires signatory states to maintain auditable logs of AI system inputs, outputs, and human decisions at each stage of the targeting cycle. The treaty mandates pre-deployment ethical review of military AI systems, establishes an international incident reporting mechanism for AI-related military failures or unintended engagements, and creates liability and accountability frameworks for violations attributable to algorithmic system behaviour.
Sea-Based Launch Treaty, draft
This instrument establishes verified limits and transparency measures for sea-based ballistic missile launch capabilities, addressing the least-monitored leg of nuclear triads and the growing proliferation of submarine-launched ballistic and cruise missiles among non-traditional naval powers. It requires declaration of submarine classes capable of strategic strike, mandates port inspection rights and at-sea monitoring protocols, and establishes AI-assisted acoustic and signals intelligence sharing arrangements among signatories to support mutual verification. The treaty also establishes the first binding governance regime for autonomous underwater nuclear delivery vehicles — a category that includes Russia’s Poseidon/Status-6, currently carried by the operational Belgorod submarine and the Khabarovsk submarine now completing sea trials — requiring positive human control at the point of terminal delivery, prohibiting automatic-detonation “dead man’s switch” designs, and mandating declaration of control architecture to the IAASB. The treaty prohibits the transfer of sea-based strategic launch technology to non-signatory states and creates a standing naval arms control commission to adjudicate compliance disputes and manage the inspection regime.
This instrument supersedes and substantially expands the Joint Comprehensive Plan of Action, establishing a durable multilateral framework for verified limits on the Islamic Republic of Iran’s nuclear programme and a structured pathway for the normalisation of Iran’s civil nuclear activities under enhanced international safeguards. It incorporates lessons from the JCPOA’s collapse, including an automatic snapback mechanism triggered by IAEA findings without requiring a new Security Council vote or facing a veto, a Joint Commission dispute resolution mechanism for compliance disputes, and a four-schedule graduated sanctions relief structure. The treaty integrates AI-assisted monitoring of declared and undeclared nuclear facilities, mandates continuous environmental sampling and remote sensor data sharing, and is structured around regional non-proliferation governance in the Middle East, with regional parties including Saudi Arabia and Israel invited into its framework.
The CFE Successor Treaty, draft
This instrument re-establishes verified limits on conventional armed forces in Europe, rebuilding the framework that collapsed following the Russian Federation’s suspension of participation in the Treaty on Conventional Armed Forces in Europe in 2007 and its formal withdrawal in November 2023. It updates equipment categories to reflect three decades of technological change, incorporates AI-assisted monitoring and data analysis for declared force postures, and extends the treaty’s geographic scope to address force deployments relevant to current European security conditions. The instrument reconstitutes the Joint Consultative Group under Article XI with enhanced powers to investigate anomalies. Confidence-building measures addressing prior notification for large-scale exercises remain under active development, flagged in the draft’s own text as forthcoming alongside Russia’s re-accession pathway and Ukraine’s post-conflict force ceiling.
The NPT Reinforcement Protocol, draft
This instrument strengthens the non-proliferation and disarmament obligations of the Treaty on the Non-Proliferation of Nuclear Weapons through a binding supplementary protocol addressing AI’s implications across all three NPT pillars, responding to structural gaps the IAEA’s safeguards architecture was not designed for: the 1990s discovery of clandestine programmes within NPT membership, and the absence of any mechanism addressing states that withdraw. It establishes binding AI declaration and audit requirements for civilian nuclear facilities, prohibits AI systems that optimise dual-use research toward weapons-relevant trajectories, and creates dedicated IAEA Safeguards AI Governance and Proliferation Intelligence AI Governance regimes. It requires nuclear-weapon states to develop AI-assisted verification tools capable of confirming warhead reductions with greater confidence than existing methods — coordinated with, rather than duplicating, the Comprehensive Nuclear-Test-Ban Treaty Organization’s monitoring infrastructure — directly addressing the verification concerns cited as the primary political barrier to Article VI disarmament commitments. Each Review Conference is mandated to receive an IAASB technical report on AI in the nuclear domain, supplemented by comparative disarmament tracking drawn from the annual Hiroshima Report. Withdrawal from the NPT triggers an emergency IAASB review of any AI-governed nuclear fuel cycle systems to assess resulting safeguards gaps.
BIOLOGICAL/CHEMICAL WEAPONS AI GOVERNANCE TREATY, draft
This instrument addresses the intersection of artificial intelligence with biological and chemical weapons development, production, and deployment, establishing binding obligations on signatory states to prohibit AI-assisted design, optimisation, or targeting of biological or chemical weapons agents and delivery systems. It supplements the Biological Weapons Convention and the Chemical Weapons Convention by creating AI-specific prohibitions and verification measures, mandating declaration of AI systems used in relevant life sciences and chemical research, and establishing the International CBRN AI Verification Authority (ICAVA) to administer a weapons convergence registry and tiered audit tracks. The treaty creates an international incident reporting mechanism and imposes export control obligations preventing the transfer of CBRN-relevant AI systems to non-state actors. State responsibility for AI-assisted biological or chemical weapons development actually undertaken by non-state actors is explicitly identified in the treaty’s own gap analysis as unresolved, with a future Instrument 15 recommended to address it.
Electromagnetic and Cyber Governance Protocol, draft
This instrument establishes international norms, transparency measures, and confidence-building obligations governing the development and use of electromagnetic pulse weapons, directed energy systems, and offensive cyber capabilities with potential to cause widespread civilian harm or to disable critical infrastructure, including maritime navigation systems. It does not prohibit cyber or electromagnetic operations categorically, but establishes the Azerbaijan Airlines Standard (Article VII) requiring pre-deployment procedural safeguards for GPS/GNSS spoofing operations conducted near civilian aircraft or vessels, named for the December 2024 crash in which such an operation left a civilian aircraft disoriented in an active air-defence engagement zone where it was struck by a missile. The protocol creates mandatory incident notification procedures for significant cyber operations affecting signatory state infrastructure and establishes an attribution support mechanism drawing on AI-assisted forensic analysis governed by the Global AI Governance Structure.
This instrument establishes international standards and accountability mechanisms governing the use of legal processes, judicial systems, international institutions, and administrative procedures as instruments of strategic coercion — a practice commonly termed lawfare. It defines four forms of prohibited lawfare conduct by state actors — weaponised litigation, manufactured ambiguity, legal delegitimisation narratives, and AI-enabled legal manipulation — and establishes a review mechanism for complaints by signatory states. The protocol creates transparency obligations for state legal strategies deployed in international forums and, through Article X and its companion Annex III in the Algorithmic Warfare Treaty, establishes technical standards for detecting AI-generated legal filings and procedural floods.
North Korea Governance Framework, draft
This instrument establishes a crisis-management and risk-reduction framework for the Korean Peninsula that treats the DPRK’s nuclear status as a present reality rather than a problem to be resolved through a negotiated denuclearisation timeline. It establishes crisis communication architecture and miscalculation risk-reduction measures, addresses the accelerated decision-making pressures AI-assisted early-warning systems create in the Korean Peninsula’s compressed strategic geography, and governs AI in DPRK troop deployments in light of documented experience from DPRK forces in the Ukraine war. It defines governance obligations for China and Russia and establishes trilateral US-ROK-Japan coordination. A final section sets out the conditions under which a fuller engagement architecture — including eventual treaty-based denuclearisation commitments — could become viable, treating this as a conditional future possibility rather than the framework’s primary near-term objective.
Open Skies Successor Protocol, draft
This instrument re-establishes a multilateral open skies regime providing for scheduled and short-notice observation flights over signatory state territory, rebuilding the confidence-building architecture dismantled by the withdrawal of the United States in November 2020 and the Russian Federation in December 2021 from the Treaty on Open Skies — a collapse the protocol’s own analysis attributes to states choosing withdrawal over engaging the original treaty’s dispute resolution provisions. It modernises the original treaty’s sensor and aircraft provisions to reflect current technology and integrates AI-assisted image analysis governed by the Global AI Governance Structure. The protocol establishes an annual quota of sixty enhanced security observation requests per state party, with no carryover of unused quota, an exception for observation of imminent security threats, and unlimited environmental observation data collection outside the quota system.
CIVILIAN AI AND DIGITAL INFRASTRUCTURE GOVERNANCE TREATY
This instrument establishes the first binding governance regime for civilian AI infrastructure, converting the parent framework’s voluntary Ecological and Gender Impact Assessment schemas into enforceable treaty obligations. It requires certified Ecological Impact Declarations before any hyperscale AI facility receives operational certification, sets binding water and energy consumption standards, and establishes hardware supply chain due diligence obligations. It creates the first treaty-level data provenance and consent regime for civilian AI training, giving individuals and communities a redress mechanism against unauthorised use of their work — including by government agencies — and requires free, prior, and informed consent for infrastructure sited on indigenous or local community land. It establishes the International Civilian AI Infrastructure Authority (ICAIA) as a sister body to IUSGA and ICAVA, operating a continuous environmental monitoring network modelled on the Comprehensive Nuclear-Test-Ban Treaty Organization’s International Monitoring System.