Anthropic caught between military use and defense industry exit

Anthropic is currently in an awkward position. Its models remain actively used in the ongoing conflict between the U.S. and Iran, even as the company faces government directives to wind down its operations with the Department of Defense. Anthropic is in an active dispute with the Department of Defense. President Trump directed civilian agencies to discontinue using Anthropic products, but the company was given a timeframe of six months to wind down its operations with the Department of Defense. However, the next day, the U.S. launched a surprise attack on Tehran. This initiated a continued conflict before the wind-down directive could be fully executed, leaving Anthropic models in use for targeting decisions in an active war zone.

New details have emerged regarding how these systems are being deployed. On Wednesday, The Washington Post reported that Anthropic’s systems are being used in conjunction with Palantir’s Maven system, which is a targeting system used by the U.S. military. According to the report, as military officials planned strikes, the integrated systems “suggested hundreds of targets, issued precise location coordinates, and prioritized those targets according to importance.” The Washington Post characterized this function as “real-time targeting and target prioritization.” Despite this active deployment, there are currently no legal barriers preventing this usage, as no official steps have been taken to restrict the company’s systems.

At the same time, defense-tech clients are fleeing from Anthropic. Reuters reported this week that Lockheed Martin and other defense contractors have begun swapping out Anthropic models. A managing partner at J2 Ventures stated 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.” This represents a significant retreat by defense industry clients who are actively replacing the models.

The transition comes amid potential regulatory actions. Secretary of Defense Pete Hegseth has pledged to designate Anthropic as a supply-chain risk—a designation that could restrict the use of a company’s products in defense contexts. While no official steps have been taken to designate Anthropic as a supply-chain risk, such a designation would likely result in a heated legal case. In the meantime, the lack of official designation means no legal barriers currently prevent the military’s ongoing use of the systems.

Why it matters

The situation highlights the friction between AI labs and the defense sector, where the dual-use nature of LLMs creates significant liability and operational risks for companies caught between government policy and military application.