The US Department of War announced on May 1 that it has entered into AI agreements with seven leading technology companies, authorizing the deployment of AI models on classified networks at Impact Level 6 (secret) and Impact Level 7 (top-secret). This move is part of the Department's strategy to become an 'AI-first' organization, aiming to leverage machine intelligence for critical defense operations such as intelligence analysis and automated decision-making.
Seven Giants In, Anthropic Out
The selected vendors include SpaceX, OpenAI, Google, NVIDIA, Reflection AI, Microsoft, and Amazon Web Services. Each will provide AI models or infrastructure for deployment within IL6 and IL7 environments. Notably, Anthropic was excluded from the agreements due to a dispute over safety restrictions, highlighting tensions between the company's cautious approach and the military's operational demands.
Multi-Vendor Strategy to Avoid Lock-In
The War Department emphasized that involving multiple providers is a deliberate strategy to prevent vendor lock-in and ensure flexibility across the AI ecosystem. NVIDIA will contribute its open-source Nemotron family, while Reflection AI will provide additional open-weight systems. Google will deploy its Gemini models, and SpaceX will support infrastructure linked to xAI's Grok models. Microsoft and AWS will continue to serve as the cloud and infrastructure backbone, ensuring scalability and reliability.
Open-Source Architecture and Rapid Prototyping
The Department's AI Acceleration Strategy prioritizes open-source architectures and rapid prototyping, favoring domestic vendors to maintain control over sensitive technologies. The focus now shifts to which models will first achieve IL6 deployment and how OpenAI's safety guardrails will be enforced as classified workflows expand. This multi-layered approach aims to build a robust, future-proof AI infrastructure for national security.