Generative AI Adoption in the US Military

Smarter Spending Wins

From discussions with the GenAI and defense community, we heard that allocation of DOD funds matters more than total budget. Internal development and fine-tuning, plus assessing organizational impact (e.g., determining that a $5m tool that supports one team, while a tool equal in cost could be scaled across multiple branches), often yield greater value than vendor acquisition alone.

Efficiency vs. Resilience

Weighing efficiency gains against technological dependency is critical. Once GenAI evaluation concludes, we recommend immediate Red-Teaming to assess impacts on processes. Each scenario should be tested through three lenses: full automation (total dependence on GenAI), collaboration (partial dependence), and abstention (no dependence).

Strategic AI Partnerships

In 2024–2025, major partnerships emerged between defense, tech, and AI leaders—examples include OpenAI with Anduril, Palantir with Shield AI, and Meta with Scale AI. These alliances share key traits:

1. Deep experience in aggregating and synthesizing data for DOD platforms;

2. Access to large-scale datasets and funding to train advanced GenAI systems;

3. Concentrated engineering talent and proven RDT&E cycles for developing foundation models.

Data Rights Dilemma

Data rights remain a core challenge in licensing agreements between AI vendors and the DOD. The DOD holds vast but fragmented datasets, making contracts a balance between security and pricing tied to model training. Vendors must also project how pilot-phase pricing scales to thousands of users. We recommend the DOD invest in durable GenAI contracting practices—backed by funding, labor, and time—to ensure long-term viability for both sides.