Pentagon Stops Using Anthropic’s Claude AI Following Prior Supply‑Chain Blacklist

This week, the U.S. Department of Defense said it will cease using AI products from Anthropic, the firm that created the Claude language model, reversing a policy adopted only months earlier when the agency added the company to its supply‑chain risk list.

Earlier this year, defense officials classified Anthropic as a “supply chain risk,” warning that its technology might leak classified information or be susceptible to foreign manipulation. That label grouped Anthropic with other commercial suppliers whose offerings now must undergo extra review before being incorporated into military platforms.

Despite that, multiple insiders reported that the Claude model had been adopted broadly across several Pentagon departments in the months after the blacklist. The system was said to be integrated into diverse tools—ranging from intelligence analysis to logistics planning—thanks to its high accuracy on natural‑language tasks and rapid deployment capability.

Removing the tools highlights the clash between the department’s desire for state‑of‑the‑art AI and its obligation to protect the supply chain’s integrity. Commercial AI can speed up mission‑critical processes, yet it also sparks concerns over data provenance, model transparency and possible backdoors, particularly when the vendors are newer, privately‑owned firms.

Going forward, the Pentagon announced plans for an exhaustive review of its AI acquisition policies, focusing on tighter vetting and increased dependence on approved, domestically‑sourced technologies. Legislators and oversight agencies are likely to watch the effort closely as the department strives to marry swift innovation with the need to safeguard national security.