The rapid advancement of military artificial intelligence (AI), especially its potential integration into nuclear systems, presents significant risks to strategic stability and established deterrence practices. Despite these concerns, no dedicated governance framework currently exists to address the specific challenges of the AI–nuclear nexus. Existing initiatives have primarily focused on ensuring human control over nuclear decision-making.

There are a number of state-led initiatives on the governance of military AI more broadly. They can be adapted to address the use of AI in nuclear weapons, but applying them will not be straightforward. There is thus a need to extend the conversation beyond the ‘human in the loop’ concept and develop targeted governance measures. Future discussions could investigate the precise level and degree of required human control and set clear red lines for both the extent and the type of AI integration in nuclear and related systems.

Table of contents

I. Introduction

II. AI governance approaches applicable to nuclear forces

III. Challenges and pathways to governance of AI in nuclear forces

IV. Conclusions