AI Regulation and Government Procurement

Anthropic briefed the House Homeland Security Committee behind closed doors as a Pentagon court filing on March 17 warned that the company’s large number of foreign nationals — reportedly including many from the People’s Republic of China — create "adversarial" supply‑chain risks because they could be compelled under China’s National Intelligence Law. The filing contrasts Anthropic with other labs even as the Defense Department continues to use its tools and may extend off‑boarding deadlines; Axios also notes industry recognition of Anthropic’s operational‑security measures (including disrupting an alleged Chinese cyber‑espionage campaign and banning PRC users) and that a hearing on its request for temporary relief is set for March 24.

In a March 17 declaration filed in federal court, Pentagon undersecretary Emil Michael argues that Anthropic poses a heightened national‑security risk because it employs 'a large number of foreign nationals,' including 'many from the People’s Republic of China,' to build and support its large‑language‑model products, warning those workers could be compelled to spy under China’s National Intelligence Law. The filing, part of the Defense Department’s bid to dismiss Anthropic’s lawsuit challenging its designation as a 'supply chain risk,' says the Pentagon’s worries extend beyond disputes over domestic surveillance and autonomous weapons and distinguishes Anthropic from rival labs it says provide stronger security assurances. At the same time, DOD acknowledges it is still relying on Anthropic’s tools and is prepared to extend deadlines for federal systems to off‑board them, underscoring the government’s dependence on commercial AI even as it questions specific vendors’ security. Axios notes that foreign‑born talent, and Chinese‑origin researchers in particular, make up a large share of top U.S. AI researchers, and quotes analyst Samuel Hammond calling insider threats 'genuine and tricky' while saying Anthropic is widely seen inside the industry as unusually aggressive in policing such risks and has previously disrupted a Chinese espionage campaign on its own platform. A hearing on whether to grant Anthropic temporary relief from the supply‑chain‑risk designation is scheduled for March 24, making this an early legal test of how far Washington can go in using procurement rules and national‑security designations against an AI company over workforce composition and policy fights.