Anthropic CEO Dario Amodei has resumed last-minute talks with the U.S. Department of Defense to try to preserve the company’s access to Pentagon contracts as the Trump administration weighs designating it a supply chain risk. The status would force defense contractors to cut ties with any company so labeled, effectively shutting Anthropic out of military procurement networks.
Amodei has been negotiating with Emil Michael, the undersecretary of defense for research and engineering, to finalize terms for Pentagon use of Anthropic’s AI models, people familiar with the matter told the Financial Times. A new agreement could let the Defense Department continue using Anthropic technology while addressing the Pentagon’s policy concerns.
Negotiations collapsed last week after a sharp breakdown in tone: Michael reportedly called Amodei a “liar” and accused him of having a “God complex.” The two sides failed to agree on specific contractual language Anthropic said was necessary to prevent misuse of its models. According to an internal memo seen by the FT, the Pentagon signaled near the end of talks that it would accept Anthropic’s broader terms if the company removed a clause that restricted the “analysis of bulk acquired data.” Anthropic views that phrase as a protection against potential mass domestic surveillance and refused to back down; preventing use of its models in lethal autonomous weapons is also a stated red line for the company.
The dispute escalated after Defense Secretary Pete Hegseth warned that Anthropic could be designated a supply chain risk, a move that would bar the company from U.S. military procurement networks. The possible designation comes despite Anthropic’s existing defense work: in July 2025 the company won a contract worth up to $200 million from the Defense Department and became the first AI provider whose models were deployed in classified settings and by national security agencies. U.S. military personnel reportedly used Anthropic’s Claude model to support a major air strike on Iran hours after President Trump ordered federal agencies to stop using the company’s systems.
Tech industry groups have cautioned the administration against treating a U.S. AI firm as a supply chain threat. In a March 4 letter to President Trump, trade associations including the Software & Information Industry Association, TechNet, the Computer & Communications Industry Association and the Business Software Alliance warned that labeling a domestic company this way could undermine U.S. AI leadership, discourage innovation and weaken the country’s competitive position against China. Those groups represent hundreds of American tech firms, including larger players such as Nvidia, Google and Apple.
The renewed talks aim to bridge the gap between the Pentagon’s operational and security concerns and Anthropic’s safeguards on misuse. As discussions continue, the outcome will determine whether Anthropic retains access to U.S. defense contracts or faces a designation that could cut the company off from military customers.
This report is based on available sourcing and follows journalistic standards; readers are encouraged to verify details with original reporting.