IBM is launching a large language model purpose-built for defense and national security work that was developed with data from Janes and built on IBM’s Granite foundation models. Company and Janes officials provided DefenseScoop an exclusive preview ahead of the model’s public rollout, saying the system is engineered for deployment in air-gapped, classified, and edge environments and can be connected to secure customer networks via an application programming interface.
Janes is the primary data source for the model and supplies structured, human-vetted information collected from manufacturers, public government statements, and on-the-ground reporting at events like air shows. That dataset is delivered to customers through secure feeds on a schedule, and the model is designed to query live Janes data rather than memorize every fact. Janes and IBM officials emphasized that the approach reduces reliance on inconsistent internet sources and helps the model produce more reliable outputs about equipment, terminology, standards, and mission context.
Officials described early target use cases as operational planning, intelligence support, and strategy work within the defense industrial base. Integrators could embed the model into broader systems, including efforts tied to CJADC2 and the Maven smart system. The companies plan a subscription-based pricing model that supports continuous updates and integration work. IBM and Janes stressed the model is a decision-support tool meant to augment human analysts, not replace them, and they expect initial customer implementations in the coming months. The preview also noted the broader context that generative Artificial Intelligence tools can produce convincing but sometimes inaccurate outputs, and that the Department of Defense is actively investing in advanced algorithms and large language model capabilities.
