US national security faces volatility in AI arms race with Anthropic

The United States is navigating a complex and often contradictory relationship with artificial intelligence developers as it attempts to maintain a technological edge over China. A central point of tension involves the AI firm Anthropic. The Pentagon previously threatened to designate the company a ‘supply-chain risk’ following disputes over the military’s use of its Claude models. Anthropic CEO Dario Amodei has resisted certain applications, including the use of AI for fully autonomous weapons and mass domestic surveillance, citing reliability concerns and democratic values.

While the U.S. government has issued conflicting directives—at one point ordering defense contractors to purge Anthropic products by September 1, only to later rescind the order—the strategic stakes remain high. Officials have described the potential loss of advanced models like Anthropic’s Mythos as ‘unilateral disarmament’ in the context of offensive cyber operations and network defense.

The competition is increasingly viewed through the lens of a digital arms race. U.S. officials express concern that China could rapidly develop comparable, lower-cost AI capabilities that could penetrate American critical infrastructure, such as power grids and water systems. To mitigate these risks, the government has implemented oversight for ‘frontier’ models to prevent their use in creating biological weapons or compromising nuclear command systems, though critics note that ‘open’ models remain a potential loophole.