FORMAL CONCEPT NOTE
DEFINITION
Definition of Human Agency
“The practical ability to understand a system, judge its capabilities and limitations, determine the
conditions for its use, question or reject its outputs and intervene when necessary.”
1.
Foundational Capability & Capacity
Directly caters to the initial emphasis on capability and capacity. Ensures countries at diverse stages
of technological, compute and institutional development can participate equitably and govern systems
with genuine sovereign authority.
2.
Comprehensive System Assurance
Human agency encompasses rigorous testing, evaluation, validation and continuous monitoring (TE&M)
to verify that AI models are safe, reliable, fair, transparent and strictly compliant with applicable
international law including IHL.
MUTUAL REINFORCEMENT
PRINCIPLE
“The three pillars are mutually reinforcing, with Capacity and Capability enabling responsible
adoption,
AI-assisted decision-making demonstrating responsible application and assurance, safeguards and responsibility
providing the frameworks and assurance necessary to sustain responsible use.”
01 / 03
PILLAR 01
Capacity and Capability
Developing capability across compute, data, talent, models and use cases to enable equitable participation
across the AI value chain.
Effective governance of AI in military requires that states have the technical, institutional and human
capacity to evaluate, procure, deploy and oversee these technologies. Pillar 1 addresses foundational
capability:
- Compute & Infrastructure: Access to computational power, secure environments and
resilient infrastructure.
- Data Governance & Sovereign Repositories: High-quality, representative and secure
training datasets.
- Human Talent & Education: Training military operators, legal advisers, technical
engineers and policy specialists.
- Knowledge Exchange: Multilateral research collaborations, technology-sharing
partnerships and South-South exchanges.
Review related outcome documents →
02 / 03
PILLAR 02
AI-Enabled Decision Making and Human
Control
Examining practical uses of AI in military contexts with strict emphasis on meaningful human control and
situational awareness.
Pillar 2 addresses operational integration, ensuring that artificial intelligence systems function
reliably as decision-support tools rather than autonomous decision-makers:
- Meaningful Human Control: Guiding system execution strictly through human intent and
operational situational awareness.
- Defined Operational outcomes: Context-appropriate applications with clear constraints
on autonomous functionalities.
- Information Integrity: Ensuring systems work on verified, timely and unbiased
operational intelligence.
- Command Oversight: Sustaining informed human judgement at every critical stage of
military execution.
Explore the Summit outcome →
03 / 03
PILLAR 03
Assurance, Safeguards and Responsibility
Establishing policy, legal, regulatory and oversight mechanisms to guarantee human accountability and
compliance with international law.
Pillar 3 focuses on accountability, verifiable safeguards and ethical oversight throughout the AI
lifecycle:
- Command Responsibility under IHL: Upholding moral agency and legal accountability
under applicable International Humanitarian Law.
- Procurement Assurance: Clear supplier information requirements, rigorous testing and
independent evaluation access.
- Incident Learning: Multilateral protocols for reporting anomalies, post-incident
reviews and risk mitigation.
- Confidence-Building Measures: Transparency protocols that enhance strategic stability
among participating states.
Register for Pillar Working Groups →
