AI #158: The Department of War

The AI industry faced its most severe political crisis as the Department of War (DoW) moved to potentially destroy Anthropic, declaring it a national security supply chain risk. This escalation followed OpenAI's decision to sign a contract with the DoW, creating a stark divide between AI labs on military collaboration. The situation deteriorated after Secretary of War Hegseth went rogue, overriding former President Trump's more measured executive order that merely banned Anthropic from direct federal use with a six-month transition period. Internal Anthropic communications, including a damaging Slack message from CEO Dario Amodei written during tense negotiations, further poisoned the atmosphere, revealing deep mistrust between the AI safety-focused company and military leadership.

The core conflict revolves around autonomous weapons systems and surveillance applications, with Anthropic seeking contractual language to maintain its safety 'stack' and refuse certain requests, while the DoW insists on ultimate authority over deployment. Ironically, both sides reportedly agree on maintaining human oversight in kill chains (referencing policy '3000.09') and claim disinterest in mass surveillance. Zvi argues the contractual language is largely symbolic, as the DoW could proceed with alternative technology regardless, making the push to 'murder' Anthropic as a company an alarming overreach that threatens national security and economic stability. The crisis highlights the fragile relationship between transformative AI developers and government power, setting a dangerous precedent for state intervention in the tech sector.