Research map / Large language models / AI and society
Position: AI-Powered Autonomous Weapons Risk Geopolitical Instability and Threaten AI Research
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.
A position paper arguing that ML-enabled autonomous weapons lower the human cost of war, raising geopolitical instability risks and threatening academic AI research, and proposing policy and academic recommendations.
Problem
Near-future risks that machine-learning-enabled autonomous weapons systems pose to geopolitical stability and to the openness of AI research, which have received less attention than AGI risks.
Approach
Argues that AWS substitute for human soldiers, reducing the political cost of initiating war and making low-intensity conflicts more likely even if civilian casualties fall; it links this to military-civilian AI research overlap and proliferation, and recommends human-presence requirements, autonomy standards, transparency, and university oversight of military funding.
Key results
- Argues that AWS lower the human cost of conflict initiation, so a rise in lower-intensity conflicts is likely even if civilian casualties decline.
- Cites a public opinion survey (2020-2021, 28 countries) finding about 61 percent of adults oppose fully autonomous weapons.
- Notes the US export control threshold of 4800 TOPS is over 48 times the 100 TOPS of the Jetson Orin NX used on the V-Bat platform, arguing hardware restrictions will not curb AWS proliferation.
Surprising findings
- Argues that hardware export restrictions on GPUs will not limit AWS proliferation because AWS compute needs are modest and overlap with civilian edge chips.
- Reports that a 2024 study found military-related LLMs recommended pro-escalation tactics, including escalations that provoked arms races and nuclear weapons use.
Limitations
- Position paper with no original experiments; claims rest on cited reporting and assumptions about future trajectories.
- Assumes AWS do not increase civilian casualties, which the authors acknowledge is optimistic.
Closest papers by meaning
- Military AI Needs Technically-Informed Regulation to Safeguard AI Research and its Applications (NeurIPS 2025 · similarity 0.88)
- Dynamic Risk Assessments for Offensive Cybersecurity Agents (NeurIPS 2025 · similarity 0.81)
- The WMDP Benchmark: Measuring and Reducing Malicious Use with Unlearning (ICML 2024 · similarity 0.80)
- Position: AI Safety Must Embrace an Antifragile Perspective (ICML 2025 · similarity 0.80)
- CTRL-ALT-DECEIT Sabotage Evaluations for Automated AI R&D (NeurIPS 2025 · similarity 0.79)
- GT-HarmBench: Benchmarking AI Safety Risks Through the Lens of Game Theory (NeurIPS 2026 · similarity 0.79)
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.