A memo that orders rip-and-replace of a vendor’s stack inside nuclear, missile defense, and cyber systems is not a contract dispute; it is a contamination order. CBS News reported on March 6, 2026, that Defense Department CIO Kirsten Davies signed a memorandum giving military commanders 180 days to remove Anthropic AI from Department of War systems and networks, one day after the Pentagon formally designated Anthropic a supply chain risk.
The language is supply-chain, not service-level
CBS News quoted the memo alleging Anthropic’s AI presents an unacceptable supply chain risk for use in all Department of War systems and networks. Davies warned adversaries can exploit vulnerabilities and that exploitation could pose potential catastrophic risks to the warfighter, with exemptions only through her office and only with a mitigation plan. CBS News also reported the memo demands any company on Pentagon work stop using Anthropic products on Defense contract work within the same 180-day window. That reads like quarantine language applied to an American AI firm for the first time.
Reuters, on March 5, 2026, carried the supply chain risk designation and official confirmation. CNN later reported Anthropic arguing the designation is narrower than first implied, but the memo’s operational effect is still removal from key national security systems. Microsoft told CNBC it would keep Claude available to customers outside the Department of War scope, underscoring how the military is severing a vendor while enterprise channels stay open elsewhere.
From vendor to liability in one designation
CBS News tied the escalation to Anthropic’s two red lines against mass surveillance on Americans and fully autonomous weapons, which CEO Dario Amodei said reflect American values, while the Pentagon wanted Claude for all lawful purposes. CBS News also noted Claude is deployed on classified networks and has been used in the Iran conflict context, with retired Navy Admiral Mark Montgomery describing AI-accelerated targeting workflows. Once policy and product clash at that level, the memo treats the product as infrastructure that must be purged, not tweaked.
What This Actually Means
Treating Anthropic like a contaminant signals the military now classifies certain AI stacks as untrusted until proven otherwise across the whole contract base. Enterprise buyers will watch how rip-and-replace language migrates from defense memos to procurement checklists.
What is the Pentagon’s Anthropic memo and who signed it?
The internal memo dated March 6, 2026, was signed by Chief Information Officer Kirsten Davies and orders removal of Anthropic products from Department of War systems within 180 days, including nuclear, ballistic missile defense, and cyber warfare contexts. It follows the March 2026 supply chain risk designation and Anthropic’s lawsuits alleging retaliation.