#國防AI

Pentagon's AI Divorce: Anthropic Deemed 'Untrustworthy' as DoD Pivots to OpenAI and AWS for Classified Model Training

The Pentagon has fractured its relationship with Anthropic, with the DOJ labeling the firm 'untrustworthy' due to restrictive AI safety guardrails. In response, the DoD is moving to train models on classified data through a new OpenAI-AWS partnership, signaling a shift toward 'sovereign' defense AI tailored for lethal military operations.

Jason·Mar 18, 2026·3 min read