Pentagon says it has stopped using Anthropic AI after phase-out

The Pentagon says it has stopped using Anthropic’s AI products, months after a phase-out was announced. An unnamed defence official told the BBC on Monday that use had ceased, but did not say when the last systems were switched off or spell out the statement’s coverage of contractor-operated systems. [1]

People familiar with the matter told the BBC that Anthropic’s Claude model was still being used as recently as the previous week for research, intelligence work and military operations against Iran. Those accounts concern use before Monday’s statement; they do not establish use after the Pentagon’s unspecified cutoff. The BBC’s sources also described intelligence analysis using Claude through Palantir’s Maven system. [1]

That Maven account sits alongside an earlier Pentagon assessment. On 10 September, research chief Emil Michael estimated that roughly 90% of the move away from Anthropic in classified military systems was complete. He said Maven and Palantir work had moved off Anthropic months earlier. His estimate did not specify which systems were counted, while the BBC’s Maven account did not date each use, leaving it unclear whether the two descriptions concern the same deployments and period. [2] [1]

Donald Trump announced a six-month phase-out in February. A subsequent Pentagon memorandum called for Anthropic products to be removed as soon as practical and within 180 days, and barred contractors from using them for department work. The instruction set a requirement; it did not record when the transition was completed. [4] [3]