Don't Worry About the Vase Podcast AI #158: The Department of War
Mar 5, 2026
A fast tour of the Anthropic–Department of War crisis, showing policy fights, legal maneuvers, and industry fallout. Discussions of deepfakes, jailbreak toolkits, and debates over defense production and supply‑chain risk. Coverage of regulatory bills, chip export limits, and the political messaging swirling around AI. Ends with reflections on aligning smarter‑than‑human systems and a lighter, humorous wrap‑up.
AI Snips
Chapters
Transcript
Episode notes
Anthropic Crisis Nearly Triggered Corporate Murder
- The Anthropic vs Department of War crisis nearly escalated into corporate destruction that could harm national security and the economy.
- Zvi describes a rapid sequence: Trump’s de-escalation truth, Hegseth's rogue SCR tweet, OpenAI signing with DOW, and a leaked Dario Slack that reignited the fire.
The Dispute Was Narrow And Solvable
- The core disputes are narrow and resolvable: autonomous weapons, surveillance, and the military's need for operational control.
- Zvi argues everyone broadly wants the same outcomes and that contract language plus wind-down terms could have solved most issues.
Guarantee Wind Down And Transition For Military Use
- Offer bounded operational guarantees: keep deployed models available for a fixed wind-down period and provide transition assistance for ongoing operations.
- Zvi proposes explicit X-month/Y-month commitments to remove DOD's rug-pull fear and enable integration.