A new lab, GenWar, is set to open in 2026 at Johns Hopkins Applied Physics Laboratory. Its mission: revolutionize defense wargaming using generative AI. More details on the GenWar lab.
This initiative aims to harness large language models, allowing human players to rapidly test strategies and interact with AI agents simulating advisors or even enemy leaders.
Traditional wargames, like the 19th-century Prussian “kriegsspiel,” are labor-intensive. GenWar lab seeks to offer faster, more in-depth analyses and multiple iterations. Learn about kriegsspiel.
The lab will enable direct interaction with sophisticated computer models and even allow wargames played entirely by AI actors on both sides, accelerating human learning.
Tools like GenWar TTX create digital environments, while GenWar Sim, built on AFSIM, facilitates human interaction with physics-based models through plain speech.
Developers acknowledge AI players might not make optimal decisions but aim for realistic enough behavior to explore strategies effectively.
Experts emphasize AI will supplement, not replace, traditional methods. It provides accelerated human learning rather than definitive answers.
Benjamin Jensen of the Center for International and Strategic cautions against reducing strategic analysis solely to LLM outputs, stressing proper documentation and evaluation.
The key challenge remains benchmarking foundation models against strategy and statecraft to avoid pitfalls in integrating AI into national security policy.
The embrace of generative AI, however, comes with its own set of critical considerations, as evidenced by recent reports concerning potential biases and vulnerabilities embedded within advanced models like China’s DeepSeek-R1. Such findings highlight the necessity of meticulous scrutiny.
Research indicates these models can generate code with significant security flaws when prompted with politically sensitive topics, raising serious questions about the integrity of AI-driven software development in critical defense applications. This calls for robust oversight.
Beyond code generation, the proliferation of AI technology also brings risks of deceptive applications, with instances of fake AI and messaging apps on third-party stores concealing spyware and perpetrating ad fraud. Such malicious uses underscore the broader challenges.