NEW YORK, March 6, 2026, 07:45 EST
- The Pentagon has moved to block its contractors from using Anthropic technology in any U.S. military projects, citing “supply-chain risk.”
- Sources say Palantir’s Maven Smart Systems relies on Anthropic’s Claude code.
- Palantir shares barely budged early Friday, as the order ramped up compliance pressure.
The Pentagon has officially tagged AI startup Anthropic as a “supply-chain risk,” barring U.S. defense contractors from deploying Anthropic’s technology, according to people with knowledge of the decision. That designation is adding new pressure to Palantir Technologies’ military software business. 1
Palantir finds itself in a tricky spot. According to people familiar with the matter, its Maven Smart Systems—used for military intelligence tasks and weapons targeting—leans heavily on prompts and workflows developed with Anthropic’s Claude. Those Maven contracts, spread across the Defense Department and other U.S. national security entities, could top $1 billion in value, Reuters previously reported. 2
The squeeze isn’t only about the technicals. Defense Secretary Pete Hegseth put contractors on notice: “effective immediately” they’re barred from “conduct[ing] any commercial activity” with Anthropic. Lawyers, though, are already debating if the administration’s authority really stretches this far past direct government use. 3
A major tech industry association warned Hegseth that the plan could unsettle suppliers and make it harder to secure “best-in-class” technology, pushing the Pentagon to stick with routine procurement for resolving issues. “Emergency authorities… are typically reserved for entities… designated as foreign adversaries,” ITI CEO Jason Oxman said in a letter reviewed by Reuters. 4
Anthropic has acknowledged the designation and says it plans to challenge it in court. CEO Dario Amodei described the restrictions as having “a narrow scope,” saying they only affect Claude’s use in Pentagon contracts—not other projects from customers who also work with the military.
Pentagon Chief Technology Officer Emil Michael posted late Thursday on X, making it clear there’s no active negotiation happening with Anthropic. Reuters reports that the Trump administration has renamed the Defense Department to the “Department of War.”
Microsoft’s legal team found that Anthropic products “can remain available” to users outside the Department of War via Microsoft 365, GitHub, and its AI Foundry platform, Reuters reported. Amazon, which backs and uses Anthropic, didn’t reply right away when Reuters reached out for comment.
Reuters reached out, but Palantir, the Pentagon, and Anthropic all declined to comment on Palantir’s Maven use of Claude. At a defense tech event in Washington, Palantir CEO Alex Karp cautioned that Silicon Valley firms that “screw the military” risk pushing the U.S. toward “the nationalization of our technology,” as he said in remarks posted on X.
Lockheed Martin—among the largest U.S. defense contractors—says it will comply. “We expect minimal impacts,” the company said, noting to Reuters that none of its projects rely on a single AI vendor.
Palantir slipped 0.3% to $152.67 Friday, putting its market cap near $433 billion.
Palantir faces a tough execution challenge here—switching out a model that’s already running in a live environment isn’t just a matter of picking a new back-office provider. According to people familiar with the situation, the company would need to swap Claude for a different model and rework sections of Maven. Timing is uncertain, but the Pentagon has made it clear it wants adjustments immediately.
Palantir’s focus has long been selling data platforms to both government agencies and private-sector clients, but a closer alignment with U.S. defense tech upgrades has sharply boosted its market value. Still, the latest episode shows that even top-tier contractors aren’t immune to risks tied to the policy disputes and supply chain jolts of their AI suppliers.
