Court keeps Anthropic label active
A U.S. federal court refused Anthropic’s bid to pause a government ‘supply chain risk’ designation tied to access demands for its Claude model, leaving the label in place for now. The ruling underscores how national‑security concerns can limit an AI vendor’s ability to supply public‑sector customers without meeting government access expectations. (euronews.com) (nytimes.com)
A federal appeals court in Washington left Anthropic under a Pentagon “supply chain risk” label on April 8, so defense contractors still cannot use Claude on Department of Defense work while the case continues. The judges said Anthropic had not cleared the high bar for an emergency pause, even though they signaled they would move the appeal quickly. (cnbc.com) That label is not a general ban on Anthropic everywhere in government. It is a Pentagon procurement warning that forces contractors to certify they are not using Anthropic’s models on Defense Department contracts, which is the part of the market tied to military work. (politico.com) The fight started after Anthropic refused to give the government what reports described as broad or “unfettered” access to Claude for national-security use. The Pentagon then argued that relying on a model it could not fully access created a risk in a wartime supply chain, which is the legal hook behind the designation. (euronews.com) Anthropic says the government is using procurement law like a punishment tool. In court, the company argued the designation was unlawful retaliation and said the label was already hurting its business with public-sector customers. (nytimes.com) What makes this unusually messy is that Anthropic is fighting on two legal tracks at once. A federal judge in San Francisco issued a preliminary injunction on March 26 that blocked a broader federal move against Anthropic, while the Washington case covers the separate Pentagon designation that stayed in place this week. (cbsnews.com) So the company has one court saying the government likely went too far in one part of the dispute, and another court refusing to freeze a different part of the same campaign. For contractors, that means the practical rule today is simple: Claude may still be usable in non-Defense work, but not in Pentagon contract work touched by this designation. (reuters.com) This is why the case reaches beyond one company. Cloud software vendors usually sell to government the way a food supplier sells to a school district, but a “supply chain risk” finding treats the software itself like a suspect ingredient that prime contractors must keep out of a specific kitchen. (computerworld.com) The Pentagon’s move also appears to be a first against a major American artificial intelligence company. That makes the case a live test of whether the government can pressure an artificial intelligence vendor to provide deeper access by threatening its route into defense procurement. (politico.com) Anthropic says it still expects to win on the merits, and the appeals court did not decide the full case on April 8. But until that fuller ruling arrives, the government keeps the leverage that matters most in procurement fights: the power to tell contractors which tools are too risky to bring into the building. (bloomberg.com)
