AI / ML / Autonomous Systems — 2026-03-18
Pentagon Plans Secure Environments for AI Companies to Train Models on Classified Data
The Pentagon is developing plans for generative AI companies to train military-specific model versions on classified data within accredited secure centers. AI models including Claude are already deployed in classified settings for target analysis in Iran. Classified training would bring private-sector firms into unprecedented proximity with sensitive national security data.
Analysis
Pentagon simultaneously ordered Anthropic removed from military systems (prior digest) while deploying Claude in classified Iran targeting. The contradiction suggests different Pentagon offices operating on different policy tracks. Ukraine's battlefield AI data sharing (this digest) represents a parallel track of military AI training on real combat data.
Pentagon simultaneously ordered Anthropic removed from military systems (prior digest) while deploying Claude in classified Iran targeting. The contradiction suggests different Pentagon offices operating on different policy tracks. Ukraine's battlefield AI data sharing (this digest) represents a parallel track of military AI training on real combat data.