Autonomous Weapons — AI Glossary

Autonomous weapons are military systems that use artificial intelligence to identify, select, and engage targets without direct human intervention. Also called lethal autonomous weapons systems (LAWS), they range from semi-autonomous platforms requiring human approval before firing to fully autonomous systems capable of independent kill decisions. The defining characteristic is the delegation of targeting decisions — partially or entirely — from human operators to machine learning algorithms.

Why Autonomous Weapons Matter

Autonomous weapons sit at the intersection of AI capability and existential risk. As computer vision, reinforcement learning, and sensor fusion improve, the technical barrier to building weapons that operate without human oversight drops significantly. This raises urgent questions: Who is accountable when an autonomous system kills a civilian? Can algorithmic targeting comply with international humanitarian law's requirements for distinction and proportionality?

The UN Convention on Certain Conventional Weapons (CCW) has held discussions on LAWS since 2014, but no binding treaty exists. Major military powers — including the US, China, and Russia — are actively developing autonomous capabilities, from autonomous drones to AI-assisted missile defense. The AI research community remains deeply divided, with thousands of researchers signing open letters calling for bans while defense contractors accelerate development.

How Autonomous Weapons Work

Modern autonomous weapons typically combine three AI subsystems:

Perception: Computer vision and sensor fusion to detect and classify targets — distinguishing combatants from civilians, military vehicles from civilian infrastructure
Decision-making: Algorithms that evaluate engagement rules, threat levels, and mission parameters to decide whether to fire
Navigation: Autonomous path planning that allows the system to operate in GPS-denied or communications-degraded environments

The "autonomy spectrum" is critical. Most deployed systems today are human-on-the-loop — they can act independently but a human monitors and can override. Fully autonomous systems that operate without any human oversight remain largely experimental, though loitering munitions and autonomous drone swarms are narrowing that gap rapidly.

Google DeepMind: A leading AI lab whose research in reinforcement learning and robotics has dual-use implications for autonomous systems
Fine-Tuning: The process of adapting pre-trained models to specific tasks, including military target recognition
Agentic Coding: AI systems that act autonomously toward goals — the same agent architecture pattern underlying autonomous weapons decision-making

Want more AI insights? Subscribe to LoreAI for daily briefings.