Anthropic boss summoned to Pentagon for refusing to make its AI system available to US military
Dario Amodei, head of Anthropic Photo: Ludovic MARIN / AFP / Profimedia
The American Secretary of Defense, Pete Hegseth, summoned Dario Amodei, CEO of the artificial intelligence start-up Anthropic, for Tuesday, an official told AFP on Monday, the company refusing for the moment certain uses of its AI for military and surveillance purposes, reports Agerpres.
The Pentagon chief has asked his AI vendors to lift restrictions on the use of their default models to broaden possible use cases as long as they remain within the bounds of legality, a source close to the situation told AFP, confirming a report by the website Axios.
All vendors agreed, including Anthropic, but the young California-based company wanted to prevent its AI model Claude from being used in two scenarios: mass surveillance of populations and automating a lethal attack.
Ultimatum for Dario Amodei
The Department of Defense is putting pressure on them and, according to several American media outlets, has threatened to terminate the contract that binds them or even to include Anthropic on the list of companies that present a “supply chain risk”.
The companies on this list face drastic restrictions on their contracts with the US administration. It generally applies to companies whose products US authorities fear could be used for activities likely to threaten US national security.
Currently, only foreign companies are on the list, including Chinese equipment maker Huawei and Russian antivirus software specialist Kaspersky.
“Anthropic is committed to supporting US national security through artificial intelligence,” a spokesperson told AFP, noting “productive and good faith discussions” with the defense secretary.
According to Axios, citing a Defense Department official, Pete Hegseth plans to give Dario Amodei an ultimatum. A rupture would have serious consequences for Anthropic, which would lose the $200 million contract signed in June, as well as potentially others, but also for the Department of Defense, as Claude is the only AI model used for classified defense operations.
Founded in 2021 by former OpenAI employees, Anthropic has always supported an ethical approach to AI. In early 2026, the start-up published a document called its charter, which details a series of instructions given to AI Claude to control its production. These instructions are aimed, in particular, at “preventing actions that pose an undue risk”.
