Large Language Models and International Security: A Primer

IOANA PUSCAS

Source Link

Introduction

Large language models (LLMs) represent one of the most prominent types of contemporary artificial intelligence (AI) systems. They are best known for their ability to generate content or summarize text when embedded in chatbots, but the range of applications of this technology is far broader – including emerging and potential use cases with impacts for international security.

LLMs are increasingly of interest to intelligence and military organizations, including for analysis, planning and other operational tasks. LLMs are also relevant to international security insofar as malicious actors could exploit capabilities afforded by LLMs for a range of nefarious purposes, such as to enhance disinformation campaigns, to conduct attacks in the cyber domain or to seek assistance with the production of weapons, including biological weapons.

This primer aims to provide an overview of LLMs and their relevance to international security: first, by introducing and explaining the basics of the technology, including how it works and where key vulnerabilities lie, and second, by illustrating the impact of LLMs on international security through select examples of uses and applications.