The Trump administration on Friday directed federal agencies to stop using Anthropic’s artificial intelligence and announced additional penalties, intensifying a public dispute over AI safety and government access to advanced models. The action followed a standoff after Anthropic CEO Dario Amodei declined to give the military unrestricted use of the company’s chatbot by a set deadline.
President Donald Trump, Defense Secretary Pete Hegseth and other officials criticized Anthropic, accusing the company of jeopardizing national security. Amodei argued Anthropic could not remove safeguards designed to prevent misuse. Trump said on social media the government would not do business with the company again.
Hegseth described Anthropic as a “supply chain risk,” a label typically used for foreign adversaries that could damage the company’s government contracts and partnerships. Anthropic called such a designation unprecedented and legally questionable and warned it could set a harmful precedent for U.S. firms negotiating terms with the government.
Anthropic had sought limited, written assurances that its Claude model would not be used for mass surveillance of Americans or in fully autonomous weapons. The Pentagon responded that it would use the technology lawfully and not for those prohibited purposes but maintained it needed unrestricted access to the models for defense uses. The clash highlights broader tensions over AI’s role in national security, especially where lethal force, sensitive data, or surveillance are involved.
The administration ordered most agencies to cease using Anthropic products immediately, while giving the Pentagon up to six months to phase out the technology where it is already integrated. Officials also warned of possible civil and criminal consequences if the company failed to cooperate during the transition. Administration spokespeople reiterated that the Defense Department must have full access to AI models for lawful national defense purposes.
The public denunciations and the supply-chain designation prompted concerns among some lawmakers and national security officials about whether the measures were based on technical analysis or political considerations. Sen. Mark Warner of Virginia said the rhetoric and actions raised serious questions about the basis for the decisions.
The dispute unsettled parts of Silicon Valley. Many AI researchers, venture capitalists and employees at firms including OpenAI and Google voiced support for Amodei’s refusal to remove safeguards, calling his conditions reasonable. The dispute could advantage rival chatbots such as Elon Musk’s Grok; Musk expressed public support for the administration, while OpenAI CEO Sam Altman defended Anthropic and noted OpenAI shares similar red lines.
Retired Air Force Gen. Jack Shanahan, a former Pentagon AI lead, cautioned that singling out Anthropic could be damaging. He noted Claude is already in wide government use, including in classified contexts, described the company’s red lines as reasonable, and warned that large language models are not yet ready for many national security applications—particularly autonomous weapons.
