Krux
February 16, 2026
Pentagon Demands AI Giants Unlock Models for Classified Military Use
Published: February 16, 2026 at 3:09 AM
Updated: February 16, 2026 at 3:09 AM
100-word summary
The Pentagon is pushing OpenAI, Anthropic, Google, and xAI to deploy their frontier AI models across classified networks with fewer restrictions. OpenAI just brought ChatGPT to GenAI.mil for 3 million military personnel on unclassified systems, with data isolated from public models. But the military wants more—access at all classification levels for broader operations. Anthropic's resisting, refusing blanket authorization for weapons development or mass surveillance, while reports claim Claude was already used in a Venezuela raid via Palantir. This standoff could reshape how commercial AI enters defense networks, balancing innovation with ethical guardrails as the military races to weaponize cutting-edge models.
What happened
The Pentagon is pushing OpenAI, Anthropic, Google, and xAI to deploy their frontier AI models across classified networks with fewer restrictions. OpenAI just brought ChatGPT to GenAI.mil for 3 million military personnel on unclassified systems, with data isolated from public models. But the military wants more—access at all classification levels for broader operations. Anthropic's resisting, refusing blanket authorization for weapons development or mass surveillance, while reports claim Claude was already used in a Venezuela raid via Palantir.
Why it matters
This standoff could reshape how commercial AI enters defense networks, balancing innovation with ethical guardrails as the military races to weaponize cutting-edge models.