Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

AI as Military Commander: An Analysis of the Legal Implications of LLM Operations in Modern Warfare

Tolevska, Rebecka LU (2026) LAGF03 20261
Department of Law
Faculty of Law
Abstract (Swedish)
Tillgänglig information om den militära användningen av stora språkmodeller är fortfarande begränsad, vilket speglar den bredare bristen på transparens kring integrationen av artificiell intelligens (AI) inom militären. Även om informationen är begränsad tyder den på att stora språkmodeller i allt högre grad integreras i krigföring. Uppsatsen syftar till att identifiera juridiska utmaningar kring användningen av stora språkmodeller, hur de förhåller sig till de grundläggande principerna i internationell humanitär rätt (IHL) samt om lagen är tillräcklig för att uppfylla syftet med regleringen.

Vapen bör tolkas i vid bemärkelse, inklusive sättet på vilket de används. Granskningen som anges i artikel 36 i tilläggsprotokoll I till... (More)
Tillgänglig information om den militära användningen av stora språkmodeller är fortfarande begränsad, vilket speglar den bredare bristen på transparens kring integrationen av artificiell intelligens (AI) inom militären. Även om informationen är begränsad tyder den på att stora språkmodeller i allt högre grad integreras i krigföring. Uppsatsen syftar till att identifiera juridiska utmaningar kring användningen av stora språkmodeller, hur de förhåller sig till de grundläggande principerna i internationell humanitär rätt (IHL) samt om lagen är tillräcklig för att uppfylla syftet med regleringen.

Vapen bör tolkas i vid bemärkelse, inklusive sättet på vilket de används. Granskningen som anges i artikel 36 i tilläggsprotokoll I till Genèvekonventionerna från 1977 (AP I) blir alltmer relevant för att hantera utvecklingen av nya vapen när ett specifikt traktat saknas, såsom i fallet med språkmodeller och annan AI. Fokus i artikel 36 i AP I ligger på vapnets avsedda användning och det förutsebara operativa sammanhanget snarare än hypotetiskt missbruk. Den rättsliga granskningen har en framåtblickande skyddande effekt, men denna förloras när vapnet redan är införlivat i statens arsenal.

Genom en kritisk analys har uppsatsen argumenterat för både efterlevnad och risker för överträdelser av IHL:s grundläggande principer. Det kan variera beroende på vilken nivå av stöd språkmodellerna tillhandahåller, situationens allvar och vilka konsekvenser som kan uppstå. Den militära befälhavaren har fortfarande kontroll, men kan acceptera vissa beslut eftersom rekommendationer från språkmodeller kan påverka beslutsfattandet och därigenom riskera att moral och andra mänskliga aspekter åsidosätts. Med stor vikt på humanitetsprincipen visar uppsatsen att IHL bygger på idén om god tro hos människor och mänskligt samvete. Maskiner kan inte garantera denna kompetens eller uppfylla de krav som anges av principerna i IHL. Samtidigt förändras krigföringens natur, och en strikt tillämpning av IHL kan förbise samtida utmaningar. Resultaten visar på behovet av kontinuerlig utvärdering av AI-vapen med tanke på de snabba framstegen inom militära teknik. (Less)
Abstract
While publicly available information regarding the military use of Large Language Models (LLMs) remains limited, reflecting the broader lack of transparency surrounding the integration of Artificial Intelligence (AI), emerging developments suggest that LLMs are increasingly being incorporated into warfare. This essay aims to examine the legal challenges associated with the use of LLMs by determining whether they comply with the fundamental principles under International Humanitarian Law (IHL) and whether the law is
sufficient to fulfil the purpose of the regulation.

“Means and methods” should be understood as weapons in the widest sense, including the manner in which they are used. The review set out in Article 36 of Additional... (More)
While publicly available information regarding the military use of Large Language Models (LLMs) remains limited, reflecting the broader lack of transparency surrounding the integration of Artificial Intelligence (AI), emerging developments suggest that LLMs are increasingly being incorporated into warfare. This essay aims to examine the legal challenges associated with the use of LLMs by determining whether they comply with the fundamental principles under International Humanitarian Law (IHL) and whether the law is
sufficient to fulfil the purpose of the regulation.

“Means and methods” should be understood as weapons in the widest sense, including the manner in which they are used. The review set out in Article 36 of Additional Protocol I to the 1977 Geneva Conventions (AP I) is becoming increasingly relevant for managing the development of new weapons, as in the case with LLMs. The focus in Article 36 of AP I is on the weapon’s intended use and its foreseeable operational context, rather than on hypothetical misuse. This reveals a regulatory gap concerning new weapons under IHL.

Through a critical analysis, the essay has argued for both potential compliance with and risks of violations of the principles of IHL. These risks may vary depending on the level of support provided by LLMs, the criticality of the situation, and the possible consequences. The military commander retains control but may accept certain decisions because recommendations generated by LLMs influence the decision-making process, thereby risking the neglect of moral judgement. With emphasis on the principle of humanity, the essay concludes that IHL is rooted in the idea of human good faith, and moral conscience. Machines cannot guarantee this competence nor meet the requirements set out by IHL. However, the nature of warfare is evolving and a strict application of IHL may overlook contemporary challenges. The findings show the need for continuous assessment on AI-enabled weapons, considering the rapid technological advancements in the military domain as systems are updated with new software and retrained on new data. (Less)
Please use this url to cite or link to this publication:
author
Tolevska, Rebecka LU
supervisor
organization
course
LAGF03 20261
year
type
M2 - Bachelor Degree
subject
keywords
IHL, Warfare, LLM, folkrätt:public international law
language
English
id
9228315
date added to LUP
2026-08-25 15:47:58
date last changed
2026-08-25 15:47:58
@misc{9228315,
  abstract     = {{While publicly available information regarding the military use of Large Language Models (LLMs) remains limited, reflecting the broader lack of transparency surrounding the integration of Artificial Intelligence (AI), emerging developments suggest that LLMs are increasingly being incorporated into warfare. This essay aims to examine the legal challenges associated with the use of LLMs by determining whether they comply with the fundamental principles under International Humanitarian Law (IHL) and whether the law is
sufficient to fulfil the purpose of the regulation. 

“Means and methods” should be understood as weapons in the widest sense, including the manner in which they are used. The review set out in Article 36 of Additional Protocol I to the 1977 Geneva Conventions (AP I) is becoming increasingly relevant for managing the development of new weapons, as in the case with LLMs. The focus in Article 36 of AP I is on the weapon’s intended use and its foreseeable operational context, rather than on hypothetical misuse. This reveals a regulatory gap concerning new weapons under IHL. 

Through a critical analysis, the essay has argued for both potential compliance with and risks of violations of the principles of IHL. These risks may vary depending on the level of support provided by LLMs, the criticality of the situation, and the possible consequences. The military commander retains control but may accept certain decisions because recommendations generated by LLMs influence the decision-making process, thereby risking the neglect of moral judgement. With emphasis on the principle of humanity, the essay concludes that IHL is rooted in the idea of human good faith, and moral conscience. Machines cannot guarantee this competence nor meet the requirements set out by IHL. However, the nature of warfare is evolving and a strict application of IHL may overlook contemporary challenges. The findings show the need for continuous assessment on AI-enabled weapons, considering the rapid technological advancements in the military domain as systems are updated with new software and retrained on new data.}},
  author       = {{Tolevska, Rebecka}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{AI as Military Commander: An Analysis of the Legal Implications of LLM Operations in Modern Warfare}},
  year         = {{2026}},
}