Updated Feb. 26, 2026 — The U.S. Defense Department and AI firm Anthropic are locked in a high-stakes standoff that could end a $200 million partnership, with the Pentagon giving the company until Friday at 5:01 p.m. to permit the military to use its Claude model for ‘all lawful purposes.’ Anthropic has resisted, seeking explicit contractual limits to prevent Claude from being used for mass surveillance of Americans or to operate weapons autonomously.
Pentagon Chief Technology Officer Emil Michael told CBS News the department has offered concessions, including written acknowledgments of federal laws that restrict surveillance of Americans, existing Pentagon policies on autonomous weapons, and an invitation for Anthropic to join its AI ethics board. Michael said those limits already exist in law and policy and emphasized the military does not use AI to power fully autonomous weapons, arguing the U.S. must prepare for adversaries such as China and cannot promise in writing to forgo defensive capabilities.
Anthropic pushed back, saying the Pentagon’s latest contract language ‘made virtually no progress’ toward blocking Claude’s use for domestic surveillance or for fully autonomous weapons. CEO Dario Amodei argued the draft paired protective phrases with legal wording that could be ignored and said the department’s threats to cut off contracts ‘do not change our position: we cannot in good conscience accede to their request.’
The dispute escalated on social media when Michael posted on X, calling Amodei a ‘liar’ and accusing him of having a ‘God-complex’ and of trying to exert undue influence over U.S. military policy. If the parties do not reach an agreement, the Pentagon says it will end the partnership, label Anthropic a supply chain risk and is considering invoking the Defense Production Act to compel compliance. Michael has not confirmed DPA use but warned no vendor will be asked to remove software that the department depends on until a replacement is in place; he also said officials are exploring other AI partners.
Anthropic’s technology is currently the only commercial model deployed on the Pentagon’s classified networks via a tie-up with Palantir, which makes the outcome particularly consequential for defense AI operations.
The clash underscores a broader split between technology companies pressing for safety and transparency and defense leaders worried that strict limits will undermine military effectiveness and U.S. competitiveness. Amodei has argued that ‘frontier’ AI systems are not reliable enough for fully autonomous weapons and warned that advanced models could aggregate dispersed data into comprehensive surveillance dossiers. By contrast, some defense officials, including Defense Secretary Pete Hegseth, have warned that excessive restrictions could hobble the military, saying forces will not use AI models that compromise their ability to fight.
Michael framed the dispute in ideological terms, suggesting some actors fear AI’s power and that a single company should not be able to set the rules for how the U.S. military and government operate. He reiterated that the Pentagon intends to use AI lawfully and to treat it like other critical technologies, while continuing to seek partnerships that meet its operational needs and legal obligations.