The Pentagon and Anthropic are locked in a high-stakes dispute after Anthropic’s CEO refused a Defense Department demand to weaken safety safeguards on the company’s AI model, Claude, or face being cut off from military work. The conflict threatens hundreds of millions of dollars in contracts and access to advanced tools for the armed forces.
Why they’re clashing
Anthropic CEO Dario Amodei has repeatedly said the company will not permit its models to be used for domestic mass surveillance or to power fully autonomous weapons that select and kill targets without human approval. Amodei describes those uses as “bright red lines” and “entirely illegitimate.”
The Pentagon says it does not plan to use Anthropic’s tools for surveillance or autonomous weapons, but maintains that contractors should not unilaterally restrict how government technology is used. Defense officials have insisted vendors must allow the U.S. government to use their tools “for all lawful purposes,” and have placed responsibility for legality on the end user.
Amodei rejected the department’s latest contract language, saying the changes would not adequately prevent Claude’s use in the two prohibited areas and included legal clauses that could negate safeguards. He said those uses were never part of existing contracts and should not be added retrospectively.
Escalation and a deadline
Tensions rose after a reported meeting between Defense Secretary Pete Hegseth and Amodei, in which Hegseth reportedly threatened to cancel Anthropic’s roughly $200 million contract if the company did not comply. A Pentagon spokesman, Sean Parnell, posted on X that Anthropic had until 5:01 p.m. ET on Friday to accept the department’s terms or face termination of the partnership and potential designation as a supply chain risk; those posts also used a rebranded label for the department.
Anthropic replied that the department’s overnight contract revision made “virtually no progress” on protecting against mass surveillance or fully autonomous weapons and said it remains willing to keep negotiating and to ensure operational continuity for the Defense Department and warfighters.
What a “supply chain risk” designation could do
Labeling Anthropic a supply chain risk would be an unusual step, historically used against foreign-adversary technologies. Experts say the practical effects are unclear: it could bar Pentagon contractors from using Anthropic’s tools in Defense work, or it could more broadly prevent contractors from using those tools, which would be particularly damaging to Anthropic.
The Pentagon has also threatened to invoke the Defense Production Act to compel Anthropic to remove guardrails — an extraordinary and highly unusual use of a law typically reserved for rare emergencies to prioritize or direct commercial production.
Contradiction and next steps
Observers note a tension in the government’s posture: Anthropic is portrayed both as too risky to be trusted and so essential that it might be forced to stay integrated into defense systems. The immediate contract in question is worth up to about $200 million, a small portion of Anthropic’s reported roughly $14 billion in revenue, and Anthropic was previously cleared for classified government use.
If the Pentagon cancels the contract, the dispute could end quickly. But if the department attempts to force the company to strip guardrails or issues a sweeping supply-chain designation, legal challenges are likely. Analysts predict Anthropic would push back in court if the department escalates.
The disagreement underscores broader tensions about how commercial AI safety policies intersect with national defense needs and foreshadows continued debate over who sets the limits on military use of advanced AI.