Pentagon signs classified contracts with Nvidia, OpenAI and others amid Anthropic dispute

What This is

• The Pentagon signed agreements with seven major tech companies to deploy their AI tools on the Defense Department’s classified networks.
• The companies named in the deals are SpaceX, OpenAI, Google, Microsoft, Nvidia, Amazon Web Services, and Reflection.
• The agreements target the Pentagon’s most sensitive cloud environments, described as Impact Level 6 and 7 classified networks.
• The Pentagon says the goal is to improve data synthesis, situational understanding, and decision-making for military operations, including through its GenAI.mil platform.
• Defense officials describe the effort as moving toward an “AI-first” fighting force and “decision superiority.”
• Anthropic was excluded after the government designated it a supply-chain risk tied to its refusal of terms allowing unrestricted military use of its technology.
• Anthropic sued the government and a federal judge recently blocked the government action described in the dispute.
• The Pentagon portrays the multi-company approach as diversifying its military AI architecture, while public descriptions of oversight safeguards and what “lawful operational use” means remain limited.

Why This Matters

• Embedding powerful AI in classified systems can reshape how a military finds targets, prioritizes threats, and escalates force, raising high-stakes questions about human control and accountability.
• Classified deployment reduces outside scrutiny, so governance choices made now can set durable precedents for secrecy, oversight, and civil-military trust.
• Contract terms about “unrestricted use” versus “guardrails” will influence whether safety constraints survive contact with operational demands.
• Locking in a small set of vendors for sensitive AI infrastructure can affect competition, security resilience, and long-term cost to taxpayers.