Court keeps Anthropic barred
A U.S. court denied Anthropic’s bid to lift a 'supply‑chain risk' designation, which keeps government vendors from partnering with the company under the national‑security‑linked label. (x.com) Commentators noted the irony given Anthropic’s own vulnerability‑finding tools, but the near‑term effect is a tightened market for federal partnerships and a likely ripple for contractors weighing Anthropic access. (x.com) (x.com)
A federal appeals court in Washington left Anthropic’s “supply chain risk” label in place on April 8, which means Pentagon contractors still cannot use Claude on Defense Department work while the case moves ahead on a fast schedule. The same order set briefing through May 13 and oral argument for May 19. (media.cadc.uscourts.gov) That label is not a general ban on Anthropic everywhere in government. The current effect is narrower and more practical: contractors doing Defense Department work must certify they are not using Anthropic’s models for that work, even if they can still use them on other projects. (cnbc.com) The legal tool behind this is a 2018 federal procurement law aimed at “covered articles” such as information technology. It lets an agency cut off a supplier when officials decide there is a significant supply-chain risk and that less intrusive measures are not enough. (uscode.house.gov) Anthropic says the Pentagon used that tool after a fight over how Claude could be used in war. In its March 9 lawsuit, the company said it wanted to block two uses in particular: mass surveillance of Americans and fully autonomous weapons. (pbs.org) The Pentagon’s designation came in early March, and Anthropic answered with two separate court challenges because two different legal authorities were involved. One case went to federal district court in California, and the other went straight to the District of Columbia Circuit because that statute gives that court exclusive review. (pbs.org) (aoshearman.com) That split is why the story looks contradictory from the outside. A California judge on March 26 blocked a separate Pentagon designation and related executive directives, but the 41 United States Code section 4713 designation stayed alive, so Anthropic still needed relief from the Washington appeals court. (aoshearman.com) (politico.com) The appeals judges did not say Anthropic’s case was weak across the board. They said a temporary stay requires a high bar, and they found the immediate harm to Anthropic looked mostly financial, while the government said the issue touched “vital AI technology during an active military conflict.” (media.cadc.uscourts.gov) (cnbc.com) The unusual part is who got tagged. Multiple reports say this is the first known time the federal government has used this kind of supply-chain-risk designation against a United States company rather than the foreign-linked vendors these powers were largely built to police. (pbs.org) (politico.com) That makes the near-term business impact pretty simple. If you are a defense contractor choosing an artificial intelligence model for Pentagon work, Anthropic is now the awkward option, because using Claude can trigger compliance problems even before the courts decide who ultimately wins. (cnbc.com) (aoshearman.com) The next date that matters is May 19, 2026, when the District of Columbia Circuit is set to hear argument. Until then, Anthropic can keep selling outside Defense Department contract work, but the Pentagon lane stays partly closed. (media.cadc.uscourts.gov) (cnbc.com)
