The dispute between Anthropic and the U.S. Department of Defense has created an unusual situation in which the company’s AI models remain in use for military operations even as the firm begins distancing itself from defense contracts. Policy directives issued under President Donald Trump instructed civilian agencies to discontinue use of Anthropic systems, while allowing the company a six-month window to wind down existing Pentagon agreements.
During that transition period, Anthropic models have reportedly continued to support military planning systems used alongside Palantir’s Maven platform, where the AI tools have been used to suggest targets, provide location coordinates, and assist with prioritization during ongoing U.S. operations against Iran. At the same time, several defense contractors, including Lockheed Martin, have begun replacing Anthropic’s models with alternative AI systems amid uncertainty about future restrictions.
The situation escalated further after Anthropic rejected a new Pentagon proposal that would have granted broad access to its AI technology for “any lawful use.” The company had sought explicit guarantees that its systems would not be used for domestic mass surveillance or autonomous weapons.
Following the breakdown in negotiations, the Pentagon reached a separate agreement with OpenAI, whose CEO Sam Altman said the contract includes safeguards addressing those concerns. Anthropic CEO Dario Amodei criticized the arrangement internally, arguing that the company had declined the Pentagon deal due to concerns about preventing misuse of advanced AI systems.
