Anthropic loses roughly $200M DoD contract after refusing Pentagon data-access demands
- Anthropic lost access to a Pentagon AI contract after the Defense Department moved in late February and early March 2026 to label it a supply-chain risk. - The key figure was a two-year Defense Department prototype agreement with a $200 million ceiling, awarded through the Chief Digital and Artificial Intelligence Office. - The next public marker is Anthropic’s court fight in Washington, where its appeal from the Pentagon’s March 3, 2026 order remains pending.
Anthropic’s fight with the Pentagon appears to be real, but the cleanest version of the story is narrower than the social-media posts. Anthropic had announced a two-year Department of Defense prototype agreement with a $200 million ceiling in July 2025, and by late February 2026 the company said Defense Secretary Pete Hegseth had directed the department to designate Anthropic a “supply chain risk.” March 3, 2026 is the key date in the court record. A CourtListener docket for Anthropic’s D.C. Circuit appeal says the case seeks review of a Defense Department order dated March 3, 2026, and Bloomberg reported the designation led to a ban on government use of Anthropic’s AI technology. ### So did Anthropic “lose a $200 million contract”? July 14, 2025 is where the $200 million number comes from. (anthropic.com) Anthropic said the Pentagon’s Chief Digital and Artificial Intelligence Office awarded it a two-year “prototype other transaction agreement” with a $200 million ceiling to develop frontier AI capabilities for national security. A ceiling is not the same thing as guaranteed revenue. (courtlistener.com) Anthropic’s own later statement said it had reached an impasse with the Pentagon over contract terms, and Dario Amodei said the company chose to forgo “several hundred million dollars in revenue” in connection with its stance on certain uses. ### What did Anthropic say it refused to allow? February 27, 2026 is the date Anthropic publicly described the dispute. (anthropic.com) The company said negotiations broke down over “two exceptions” it wanted preserved: “the mass domestic surveillance of Americans” and “fully autonomous weapons.” Dario Amodei repeated that position in a separate statement. He said Anthropic supported lawful national-security uses of AI, but said mass domestic surveillance was “incompatible with democratic values” and that today’s frontier models were not reliable enough for fully autonomous weapons. (anthropic.com) ### What does the Pentagon action appear to have done? The Pentagon’s action is described in public sources as a supply-chain-risk designation. (anthropic.com) Anthropic said Hegseth was directing the department to make that designation under 10 U.S.C. 3252, and Bloomberg reported the move led to a ban on government use of Anthropic technology. March 9, 2026 is when Anthropic took the dispute to court. CourtListener shows both a Northern District of California complaint for declaratory and injunctive relief and a D.C. (anthropic.com) Circuit petition for review filed that day. ### Did OpenAI replace Anthropic? February 28, 2026 is the date OpenAI announced its own agreement with the Department of War. OpenAI said it had reached an agreement for deploying advanced AI systems in classified environments and said the department planned a working group that OpenAI would join. (anthropic.com) OpenAI did not say in that statement that it had “stepped in” to take Anthropic’s contract. What it did say is that its agreement barred domestic surveillance of U.S. persons and that OpenAI retained red lines against mass domestic surveillance, directing autonomous weapons systems, and high-stakes automated decisions. (courtlistener.com) ### What is the most defensible takeaway right now? The most defensible account is that Anthropic’s Pentagon relationship broke down over contract terms tied to surveillance and autonomous weapons, after which the Defense Department issued a March 3, 2026 supply-chain-risk order that Anthropic is challenging in court. (openai.com) The $200 million figure is verified as the ceiling on Anthropic’s 2025 prototype agreement, but public sources reviewed here do not show a new Pentagon document explicitly saying “we terminated exactly $200 million” or that OpenAI formally assumed that specific award. May 19, 2026 is the latest public milestone in the appeal. Bloomberg reported that a federal appeals panel in Washington appeared skeptical of Anthropic’s challenge at oral argument, and CourtListener lists the appeal as still active. (bloomberg.com) (anthropic.com)
