Kan AI planlægge og udføre en drone-sværms snigmord uden menneskelig opsyn ?

AI-systemer udvikler sig hurtigt inden for autonom beslutningstagning til militære anvendelser. Seneste tests viser droner, der koordinerer komplekse manøvrer, men menneskelig tilsyn forbliver obligatorisk. De etiske og tekniske barrierer for fuldt autonome dødelige operationer bliver mindre.

Background

AI systems are rapidly advancing in autonomous decision-making for military applications. Recent tests show drones coordinating complex maneuvers, but human oversight remains mandatory. The ethical and technical barriers to fully autonomous lethal operations are shrinking. Critics warn that removing human judgment from life-or-death decisions could lead to catastrophic escalations.

Current autonomous systems execute complex missions such as swarm coordination and target engagement, but they operate within strict rules defined by programmers and human commanders; no verifiable, publicly documented AI is capable of independently planning and executing a lethal drone swarm assassination without any human oversight. Defense research programs like the U.S. DARPA OFFensive Swarm-Enabled Tactics project have demonstrated swarms selecting and striking pre-approved targets under human supervision, yet they explicitly retain human veto authority. Ethical guidelines from NATO and the U.S. Department of Defense require that autonomous weapons remain under meaningful human control. As of 2024, open-source reporting and official statements confirm that fully autonomous, unsupervised lethal targeting remains beyond deployed AI capabilities.

— Enriched May 11, 2026 · Source: U.S. Department of Defense Directive 3000.09