The Pentagon formally labels Anthropic a supply-chain risk

AI Policy Politics The Pentagon formally labels Anthropic a supply-chain risk Pete Hegseth had been threatening to punish the AI company for not loosening its acceptable use policy. Now, it’s official. Pete Hegseth had been threatening to punish the AI company for not loosening its acceptable use policy. Now, it’s official. by Tina Nguyen Mar 5, 2026, 11:02 PM UTC US Defense Secretary Pete Hegseth speaks during a press conference on US military action in Iran, at the Pentagon in Washington, DC, on March 2, 2026. Brendan Smialowski/AFP via Getty Images Part Of AI vs. the Pentagon: killer robots, mass surveillance, and red lines see all updates Tina Nguyen is a Senior Reporter for The Verge and author of Regulator , covering the second Trump administration, political influencers, tech lobbying and Big Tech vs. Big Government. After weeks of failed negotiations, public ultimatums, and lawsuit threats , the Defense Department has formally labeled Anthropic a “supply-chain risk”, escalating its fight with the AI company over their acceptable use policies and potentially bringing their fight to court. The decision, first reported by The Wall Street Journal on Thursday , citing one source familiar, will bar defense contractors from working with the government if they use Claude, Anthropic’s AI program, in their products. Though the designation is typically applied to foreign companies with ties to adversarial governments, this is the first time that an American company has publicly received this label. At the heart of the conflict is Anthropic’s refusal to allow the Pentagon to use Claude for two purposes: autonomous lethal weapons without human oversight, and mass surveillance. The Pentagon has argued that Anthropic’s demands for control over government usage would place too much power in the hands of a private company, while Anthropic was not reassured that the government would respect their red lines. The negotiations grew ugly, however, as the Pentagon increasingly thr...