Research map / Large language models / AI and society
Military AI Needs Technically-Informed Regulation to Safeguard AI Research and its Applications
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Position paper arguing that AI-powered lethal autonomous weapon systems (AI-LAWS) need distinct, technically grounded regulation, proposing a behavior-based two-criterion definition and five policy directions.
Problem
Existing LAWS governance frameworks and definitions do not distinguish modern AI-powered weapon systems, whose risks (brittleness, opacity, drift, escalation) are not addressed by high-level policy, and AI researchers are largely absent from military AI policy discussions.
Approach
Proposes a behavior-based rubric defining AI-LAWS as systems that (1) rely on AI/ML essential to function with AI-specific risks and (2) are involved in semi- or fully autonomous targeting or force application, with or without human oversight, and recommends technically informed policies (nuclear AI ban, international validation standards, ban on AI command of soldiers, clarifying legal status of
Key results
- Proposes a two-criterion behavior-based rubric for identifying AI-LAWS requiring enhanced oversight, contrasting it with narrower definitions that exclude most deployed systems.
- Tabulates 30+ real-world AI-enabled military systems across air, land, sea, and command domains with status (In Dev, Demo, Deployed).
- Recommends five policy directions: banning AI control over nuclear deployment, international AI-LAWS validation standards, banning AI command of human soldiers, clarifying legal status of civilian AI infrastructure, and institutional civil-military boundaries.
Surprising findings
- Argues that narrow definitions of LAWS (e.g., requiring full autonomy or superhuman learning) exclude most deployed and near-term systems, thereby preventing oversight of systems already in use.
Limitations
- Rubric is qualitative and not empirically validated against real systems.
- Many cited deployments rely on public reporting that the paper itself notes is incomplete or unverified.
Experimental setup
Baselines: Meaningful Human Control framework Appropriate Human Judgment framework DoD Directive 3000.09 UK 2024 military AI guidance
Builds on: Meaningful Human Control Appropriate Human Judgment DoD Directive 3000.09
Closest papers by meaning
- Position: AI-Powered Autonomous Weapons Risk Geopolitical Instability and Threaten AI Research (ICML 2024 · similarity 0.88)
- Regulating AI: Applying Insights from Behavioural Economics and Psychology to the Application of Article 5 of the EU AI Act (AAAI 2024 · similarity 0.84)
- Position: Intent-aligned AI Systems Must Optimize for Agency Preservation (ICML 2024 · similarity 0.83)
- Towards Computational Foreseeability (AAAI 2025 · similarity 0.83)
- GT-HarmBench: Benchmarking AI Safety Risks Through the Lens of Game Theory (NeurIPS 2026 · similarity 0.83)
- AI Testing Should Account for Sophisticated Strategic Behaviour (NeurIPS 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.