After previously cutting off access to its AI systems for firms linked to the Chinese Communist Party (CCP), Anthropic has reiterated its commitment to supporting U.S. national security while drawing firm boundaries around certain military applications of artificial intelligence.
Anthropic said it had voluntarily foregone several hundred million dollars in potential revenue by terminating access to its Claude models for companies associated with the CCP, including some designated by the U.S. Department of War as Chinese Military Companies. The company also reported shutting down CCP-sponsored cyberattacks that attempted to misuse its systems and has publicly advocated for stricter semiconductor export controls to maintain a democratic advantage in AI development.
Positioning itself as a strong supporter of U.S. defence objectives, Anthropic noted that it was the first frontier AI firm to deploy models within classified U.S. government networks, at National Laboratories, and to deliver custom AI systems for national security customers. Claude is currently used across defence and intelligence agencies for applications such as intelligence analysis, operational planning, cyber operations, and modelling and simulation.
However, Anthropic said it will not agree to two specific use cases in its contracts with the Department of War: mass domestic surveillance and fully autonomous weapons. While supporting lawful foreign intelligence and partially autonomous defense systems, the company argued that AI-enabled mass domestic surveillance risks undermining democratic freedoms, and that current frontier models are not sufficiently reliable to power fully autonomous lethal systems without human oversight.
The company stated that U.S. defence authorities have requested broader “any lawful use” provisions and removal of these safeguards, warning of possible consequences if Anthropic does not comply. Despite this, Anthropic maintained it cannot in good conscience abandon its red lines, while expressing readiness to continue supporting U.S. national security under its proposed guardrails.
