<style type="text/css"> .no-show { display: none; } .disable-fade-in{ opacity: 1 !important; transform: none !important; visibility: visible !important; } </style>
Studies in Intelligence Vol. 69, No. 4 (Extracts, December 2025)

Snow Globe Multi-Player AI System: Lessons from Human-AI Teaming in War Games

Andrea Brennan, Rachel Grunspan, Daniel Hogan, Jessica D. Smith, Elizabeth VanderVeen

The authors: Andrea Brennan is senior vice president and deputy director of In-Q-Tel Labs. Rachel Grunspan is a retired CIA officer and the former director of Digital Futures in CIA’s Directorate of Digital Innovation. Daniel Hogan is a senior data scientist at In-Q-Tel Labs. Jessica D. Smith is a digital intelligence strategist in DDI Futures. Elizabeth VanderVeen is an AI strategist in DDI Futures.

Introduction

As the US Intelligence Community continues to adapt to emerging threats and rapid technological change, human-AI collaboration is becoming a critical enabler of mission success. To meet this need, CIA’s Directorate of Digital Innovation (DDI) Futures team and In-Q-Tel (IQT) have collaborated on a research project to explore how humans and conversational AI can work together more effectively and, ideally, achieve outcomes neither could accomplish alone.

This project leverages Snow Globe, a multi-player AI system built by IQT’s Applied Research Team, that uses large language models (LLMs) to play open-ended war games. Through a series of jointly designed games in which human participants play alongside (or against) simulated personas, the team has demonstrated how AI war games can serve as a testbed for human-AI teaming in intelligence work.

Download PDF to read complete article (7 pages).