Pentagon Formalizes $1 Billion Maven Contract Built on AI It Just Banned

The Defense Department crowned Palantir's Maven its core military AI platform on March 20 — fifteen days after blacklisting Anthropic's Claude, the model that powers it. The US is now formally committed to waging war with AI it has officially prohibited.

Probability sourced from prediction markets. This reflects the collective wisdom of traders betting real money on this outcome.

On March 20, Deputy Defense Secretary Steve Feinberg formalized Palantir's Maven Smart System as a Pentagon "program of record" — elevating it to the permanent backbone of AI-enabled military operations across all US armed forces. The designation unlocks stable, recurring defense funding and solidifies Palantir's contracts, potentially worth over $1 billion, from short-term bridges into long-term commitments.

The problem: Maven runs on Anthropic's Claude.

Fifteen days earlier, on March 5, the Pentagon designated Anthropic a supply-chain risk — the first American AI company to receive the label — after it refused to allow "all lawful uses" of Claude, including autonomous weapons targeting. Defense Secretary Pete Hegseth ordered contractors to sever commercial ties with Anthropic immediately. Anthropic has since filed two lawsuits, sought an emergency appeals court stay, and attracted an amicus brief from nearly 150 retired federal judges. And yet the military reportedly used Claude-powered Maven during airstrikes on Iran, striking roughly 1,000 targets in under 24 hours.

This is the collision US defense policy never planned for: a 6-month phase-out order crashing into a 12-18 month minimum replacement cycle for certified military AI systems. Palantir must now rebuild core Maven workflows — stripped of Claude — while performing under a formalized program-of-record obligation. Lockheed Martin has already pledged compliance. Palantir has not disclosed its transition plan.

The bigger signal is blunt. The Pentagon has decided that AI safety restrictions — autonomy guardrails, refusal clauses, surveillance limits — are operationally incompatible with US wartime doctrine. Anthropic's blacklisting is not a procurement dispute; it is a policy declaration. What replaces Claude inside Maven will define the architecture of American AI warfare for years. And every AI company now knows the price of admission.