The Pentagon has formally notified Anthropic that the company and its products are being treated as a supply chain risk, a designation that could force government contractors to stop using Anthropic’s chatbot, Claude. The department said the notice to Anthropic leadership is effective immediately.
The move follows a week of public criticism from President Donald Trump and Defense Secretary Pete Hegseth. They argued Anthropic posed a national security concern after CEO Dario Amodei resisted requests to remove safeguards that limit applications tied to mass surveillance of Americans and fully autonomous weapons. President Trump ordered the military to phase out Claude within six months; Claude is already integrated into a number of military and national security systems.
Anthropic said it will challenge the designation in court, calling the action legally unsound. Amodei said the company sought narrow exceptions that would affect only certain high-level uses, not operational decision-making, and that talks with the Pentagon had been “productive” about continued use or a “smooth transition.” He emphasized the need to avoid depriving warfighters of tools during major combat operations.
The Pentagon defended the designation as necessary to ensure the military can “use technology for all lawful purposes” and to prevent vendors from limiting lawful uses in ways that could put warfighters at risk. How broadly the department will apply the designation remains unclear. Anthropic said a Pentagon notification suggests the designation applies only when Claude is used “as a direct part of” military contracts. Microsoft said, pending legal review, it can continue to work with Anthropic on non-defense projects.
Some defense contractors have already responded by cutting ties. Lockheed Martin told the government it will follow presidential and Defense Department direction and seek other large language model providers, saying it expects minimal impact because it does not rely on a single LLM vendor.
The designation has drawn criticism for using a rule originally aimed at threats from foreign adversaries. Federal supply-chain risk definitions focus on actors who might sabotage, introduce unwanted functions, or subvert systems to disrupt, degrade, or spy on them. Critics, including Sen. Kirsten Gillibrand, called the action “a dangerous misuse of a tool meant to address adversary-controlled technology.” Neil Chilson, a former FTC chief technologist, described the move as “massive overreach” that could harm the U.S. AI sector and reduce military access to leading technologies.
A coalition of former defense and national security officials — including former CIA director Michael Hayden and retired senior service leaders — warned lawmakers in a letter that using this authority against a domestic company is a “profound departure” from its intent. They cautioned that penalizing a U.S. firm for refusing to remove safeguards against domestic surveillance and fully autonomous weapons could have far-reaching consequences.
At the same time, Anthropic has seen a surge in consumer interest. The company reported more than a million people signing up for Claude each day last week, and the app rose ahead of OpenAI’s ChatGPT and Google’s Gemini to become the top AI app in Apple’s store in more than 20 countries.
The dispute has intensified tensions between Anthropic and OpenAI. After the Pentagon’s initial actions, OpenAI reached a deal to replace Anthropic’s services in classified military settings. OpenAI CEO Sam Altman later expressed regret, saying he regretted rushing a deal that “looked opportunistic and sloppy.” Amodei apologized for an internal message that criticized OpenAI and suggested Anthropic was being punished for not offering political praise.
The legal challenge and policy debate that follow will determine how the designation is applied and whether it sets a new precedent for regulating domestic AI providers under supply chain risk authorities. For now, contractors and the companies involved are navigating a fast-moving dispute with implications for defense procurement, industry competition, and the boundaries between national security policy and commercial AI governance.