Don't Worry About the Vase Podcast Secretary of War Tweets That Anthropic is Now a Supply Chain Risk
Mar 2, 2026
A deep breakdown of the Pentagon's push to label an AI company a supply-chain risk and the legal mess that tweet caused. Discussion of the DoD seeking bulk analysis of commercial user data and the company refusal. Contrasts between one firm's quick contract acceptance and another's firm red lines on surveillance and lethal autonomous weapons. Analysis of market, legal, and verification consequences.
AI Snips
Chapters
Transcript
Episode notes
OpenAI Relied On Trust And A Safety Stack
- OpenAI's agreement with the Department of War relied primarily on trust and a safety stack rather than contract language to enforce red lines.
- Zvi argues OpenAI is trusting the DoW to interpret "all lawful use" and relying on technical controls and the ability to pull the plug to prevent abuse.
Hegseth's Tweet Escalated And Threatened Anthropic
- Pete Hegseth tweeted a supply chain risk designation 13 minutes after Anthropic said it needed more time, escalating a negotiated off-ramp.
- Zvi calls the tweet legally questionable and a potential corporate-murder attempt with massive market and supply-chain damage.
Supply Chain Risk Has Narrow Legal Scope Not Blanket Power
- A supply chain risk designation legally targets procurement use, not blanket commercial bans, and requires procedures not followed here.
- Zvi explains such a sweeping designation would be arbitrary, illegal, and could not lawfully force cloud providers to sever ties.