White House Orders Agencies to Cut Ties With Anthropic

The standoff's roots trace back to a January 2026 AI strategy memo from Defense Secretary Pete Hegseth, mandating all Pentagon AI contracts include "any lawful use" language. This directly conflicted with Anthropic's existing agreement, which specifically prohibited using its Claude model for fully autonomous weapons or domestic mass surveillance. The "supply chain risk" designation is an unprecedented move against a U.S. company, a label typically reserved for foreign adversaries. This action legally bars any contractor, supplier, or partner doing business with the military from conducting commercial activity with Anthropic, with a six-month transition period for current systems. Anthropic has stated it will challenge the designation in court. While the Pentagon claims it has no interest in using AI for mass surveillance or autonomous weapons without human oversight, it refused to codify these exceptions. The dispute reportedly intensified after a discussion about a hypothetical nuclear missile attack, where a defense official characterized Anthropic CEO Dario Amodei's response to using Claude for missile defense as non-committal, a claim Anthropic denies. This conflict highlights a growing divide in the AI industry. While Anthropic, founded by former OpenAI employees over safety concerns, is taking a hard line, competitors like xAI and OpenAI have agreed to the Pentagon's "all lawful purposes" standard for unclassified work. Shortly after the White House directive, OpenAI announced a new deal to supply its AI to classified military networks, though CEO Sam Altman stated the agreement includes the same prohibitions Anthropic fought for. The administration's invocation of the Defense Production Act, a Korean War-era law, to force a private company to alter its software's ethical safeguards is a novel and legally contentious application of the statute. This move is seen by some as contradictory to the "supply chain risk" label, as one designates the company a threat while the other deems its product essential to national security. The Pentagon's aggressive stance aligns with its new "AI-first" mandate, which seeks to overhaul military operations with AI at their core and requires that the latest commercial models be available to military users within 30 days of public release. This rapid adoption policy creates inherent friction with AI companies focused on deliberate, safety-oriented scaling. The petition signed by hundreds of Google and OpenAI employees signals significant unease within the broader AI engineering community. It represents a strategic, cross-company effort to establish industry-wide ethical red lines, building on the precedent of past employee activism like the 2018 Project Maven protests at Google. Defense tech is a rapidly growing sector for startups, with the Pentagon budgeting $13.4 billion for AI-driven platforms. Companies like Anduril, Shield AI, and Palantir have secured significant defense contracts by focusing on military-specific applications. The fallout from the Anthropic dispute could reshape the procurement landscape, potentially favoring companies willing to forgo ethical constraints for market access.