Amodei is the final opponent of the Trump administration's desire for unfettered access to artificial intelligence

Anthropic, the AI company behind the Claude chatbot, is in a public standoff with the US Department of Defense after refusing to remove safety restrictions on its AI, even under pressure from the Pentagon.

The Pentagon wants unrestricted access to Claude for all lawful uses, which could include things like autonomous weapons and mass domestic surveillance. Anthropic’s CEO Dario Amodei says the company will not agree to these changes because they could cross ethical lines and make the technology unsafe or harmful.

Anthropic has a contract with the Pentagon worth up to $200 million, and the company was previously the first AI firm cleared for use in classified military systems. But now the US defence department has given Anthropic a deadline to comply or risk losing the contract.

The Pentagon has warned that if Anthropic does not drop the safeguards, it could be labelled a “supply chain risk,” which might block the company from future government work and even lead to the use of emergency powers under the Defense Production Act to force compliance.

Amodei says his company supports AI use in national security but draws firm lines against allowing its models to be used for fully autonomous weapons or widespread surveillance of US citizens. Anthropic has also said it is open to continued talks and helping transition services if needed.

Other major AI firms, including Google, OpenAI and xAI, have agreed to the Pentagon’s terms, leaving Anthropic as the only large AI developer resisting the request.