Anthropic's AI systems are currently being used by the U.S. military for real-time targeting and prioritization in the conflict with Iran, despite political pressure to discontinue such use. This contradiction stems from overlapping government directives, including a civilian ban and a six-month wind-down period for defense contracts.
Concurrently, many defense contractors and subcontractors are proactively replacing Anthropic's models with competitors due to the uncertain regulatory environment. The situation remains in flux, pending a potential official designation of Anthropic as a supply-chain risk by the Secretary of Defense.
The main topics covered are the contradictory U.S. government restrictions on Anthropic, the active military use of its AI for targeting, and the broader decoupling of the company from the defense industry.
The aftermath of Anthropic’s dispute with the Department of Defense has left the company in an awkward place — both actively in use as part of the ongoing conflict between the U.S. and Iran, and decoupling from many of its clients in the defense industry.
Part of the confusion is the overlapping and contradictory restrictions made by the U.S. government. President Trump has directed civilian agencies to discontinue use of Anthropic products, but the company was given six months to wind down its operations with the Department of Defense. The next day, the U.S. and Israel launched a surprise attack on Tehran, launching a continued conflict before Trump’s directive could be fully executed.
The result is that, as the U.S. continues its aerial attack on Iran, Anthropic models are being used for many targeting decisions. And while Secretary of Defense Pete Hegseth has pledged to designate the company as a supply-chain risk, no official steps have been taken to that end, so there are no legal barriers to the use of the system.
An article in The Washington Post on Wednesday unearthed new details on how Anthropic’s systems are being used in conjunction with Palantir’s Maven system. As Pentagon officials planned the strikes, the systems “suggested hundreds of targets, issued precise location coordinates, and prioritized those targets according to importance,” the Post reports. The article characterized the system’s function as “real-time targeting and target prioritization.”
At the same time, many companies involved in the defense industry have already replaced Anthropic models with competitors. Lockheed Martin and other defense contractors began swapping out the company’s models this week, according to a Reuters report. Many subcontractors are caught in a similar bind: a managing partner at J2 Ventures told CNBC that 10 of his portfolio companies “have backed off of their use of Claude for defense use cases and are in active processes to replace the service with another one.”
The biggest open question is whether Hegseth will make good on the supply-chain risk designation, which would likely result in a heated legal case. But in the meantime, one of the leading AI labs is quickly being partitioned out of military tech — even as it’s used in an active war zone.
Meridiano
