ChatGPT for the Military? Large Language Models in the Military Domain and the Role of Article 36 Weapons Reviews

Abstract

The advent of ChatGPT and large language models (LLMs) has heralded a new era in generative artificial intelligence (AI). While commercial applications of LLMs abound, the military is now exploring their potential across various areas, including battlefield operations. This chapter investigates whether, and under what parameters, LLMs used in the military domain fall under the purview of Article 36 of Additional Protocol I to the Geneva Conventions, enshrining the obligation for all States Parties to conduct a weapons review. It examines whether LLMs qualify as weapons, means, or methods of warfare by analysing relevant state practice in the interpretation and implementation of Article 36. Although LLMs do not directly inflict harm on combatants or military objects, their potential role in supporting targeting decisions may subject them to the obligation of a weapons review, as demonstrated by parallels in the cyber domain. The chapter concludes with the observation that, although in their infancy, LLMs herald a general trend toward increased military reliance on non-weaponised AI, in turn, raising critical questions regarding compliance with International Humanitarian Law.

Ähnliche Arbeiten