Pentagon: Anthropic's foreign workforce poses security risks

The Pentagon is highlighting new national security concerns about Anthropic's use of foreign workers, including from China, according to a court filing. Why it matters: The Defense Department is raising red flags about a key element of the AI industry — its reliance on global talent — as it moves to dismiss Anthropic's lawsuit.

What they're saying: "Anthropic employs a large number of foreign nationals to build and support its LLM products, including many from the Peoples Republic of China (PRC)," a March 17 declaration from Pentagon undersecretary Emil Michael states.

The use of those workers, Michael wrote, "increases the degree of adversarial risk should those employees comply with the PRC's National Intelligence Law." The risks with other major U.S. AI companies that use foreign workers are reduced by "the technical and security assurances of the other labs' leadership, along with their consistently responsible and trustworthy behavior" when working with the Pentagon, the filing states. "Anthropic's case, however, is different."

Between the lines: The Pentagon argues its concerns with Anthropic go beyond disagreements over domestic mass surveillance and autonomous weapons, and extend to broader national security risks.

However, the Pentagon is still relying on Anthropic and is willing to extend the deadline to offboard the company's tools as necessary. "As a matter of Department of War policy, we do not comment on ongoing litigation," a Pentagon spokesperson said.

Context: The Pentagon earlier this month designated Anthropic a supply chain risk. Anthropic is asking courts to undo the supply chain risk designation, block its enforcement and require federal agencies to withdraw directives to drop the company. Zoom in: Foreign-born workers make up a significant share of the top AI and tech talent in the U.S., according to several recent reports.

Chinese-origin researchers constituted roughly 38-40% of top AI talent at U.S. institutions as of 2023, according to a talent tracker.

The big picture: Anthropic was an early adopter of operational security techniques like research compartmentalization and audit trails, in part because they were the first AI lab to partner with the Pentagon.

The company last year disrupted a first-of-its-kind AI-orchestrated cyber Chinese espionage campaign on their platform and banned the PRC from their services.

Zoom out: "Insider threats are a genuine and tricky concern," the Foundation for American Innovation's Samuel Hammond said.

"Ironically, within the industry Anthropic is widely considered to be the most serious and proactive about policing insider threats from foreign nationals and otherwise."

What's next: A hearing on whether to grant Anthropic temporary relief is set for March 24. Go deeper... Scoop: Anthropic meets with House Homeland Security behind closed doors