The US government has warned it may remove AI startup Anthropic from Pentagon supply chains and void existing agreements as relations between the company and the Defense Department deteriorate.
US Under Secretary of Defense Emil Michael publicly rebuked Anthropic CEO Dario Amodei, accusing him of dishonesty and asserting that Amodei sought to exert undue control over the US military while putting national security at risk.
The confrontation followed a meeting earlier in the week between Amodei and US Defense Secretary Pete Hegseth. According to sources familiar with the discussion, Hegseth gave Anthropic until Friday, February 27, to grant the US military full access to its Claude model. If Anthropic refused, Hegseth reportedly threatened to remove the company from government supply chains or to force it to prioritize government orders, potentially invoking the Defense Production Act — a Cold War–era authority that allows the president to direct domestic industry for national defense needs.
Anthropic has declined to provide unrestricted access to Claude for classified military applications that could include fully autonomous lethal systems or domestic mass surveillance. In a statement, Amodei said Anthropic understands that military decisions are made by the Department of Defense, not private companies, and that the company has not objected to specific operations. He added that in a narrow set of cases the company believes AI could undermine, rather than defend, democratic values.
Anthropic says it will not permit Claude to be used in situations where final targeting decisions are made without human intervention or where the model would enable mass domestic surveillance, arguing such uses require safeguards that do not yet exist. Geoffrey Gertz, a senior fellow at the Center for a New American Security, told DW that invoking the Defense Production Act against an AI firm would be unprecedented and could damage Anthropic’s standing as a leader in responsible AI.
The company has supplied the Claude model to US intelligence and defense agencies since November 2024. The Wall Street Journal has reported that the US military used Claude during a 2026 raid in Venezuela that resulted in the capture of Nicolas Maduro; neither Anthropic nor the Defense Department commented on that report, and details of Claude’s role were not disclosed.
Hegseth’s threat to exclude Anthropic from Pentagon supply chains would have financial implications. In July 2025, the Department of Defense awarded Anthropic a $200 million contract to prototype frontier AI capabilities intended to advance US national security. At the time, Anthropic described the deal as a commitment to US security while stressing responsible deployment and efforts to make its systems reliable and interpretable.
Anthropic, founded in 2021 by seven former OpenAI employees, has long promoted a safety-first posture. But the company signaled a shift on February 24, the same day as the Hegseth meeting, announcing it was relaxing a core safety policy to remain competitive with other leading AI models. Anthropic said the policy environment had moved toward prioritizing AI competitiveness and economic growth while federal safety discussions had lagged; a company spokesperson told DW the policy change was unrelated to the Pentagon talks.
The clash raises ethical and strategic questions about government access to powerful AI. If Anthropic accedes to Hegseth’s demands, or if the Defense Department compels cooperation via the Defense Production Act, critics argue it could erode the company’s safety-focused reputation and prompt concern about AI deployment in lethal operations or broad surveillance.
The dispute also fits within a broader administration approach of direct intervention in sectors it deems critical, following large-scale support for domestic chip manufacturing and rare-earth mining. Observers warn that such interventionist policies alter the relationship between corporations and government, moving away from a traditionally hands-off posture in technology development.
The standoff is likely to produce legal and political challenges as both sides balance national security imperatives, corporate autonomy, and ethical limits on AI use. The reporting has been updated to include statements from Anthropic CEO Dario Amodei and comments by US Under Secretary of Defense Emil Michael.