U.S. Department of Defense Chief Technology Officer Emil Michael said on Thursday that Anthropic's Claude artificial intelligence models could contaminate the Department of Defense supply chain because "different policy preferences" are embedded in the models. "We cannot allow a company to contaminate the supply chain by embedding different policy preferences in its model through its charter, core, and policy preferences, resulting in our combatants receiving ineffective weapons, ineffective body armor, and ineffective protective equipment."
"That's really where supply chain risk identification comes in," Michael said on CNBC's "Squawk Box."
Anthropic is the first U.S. company to be publicly listed as a supply chain risk — an unusual step that has historically been reserved for foreign adversaries.
The designation will require defense contractors and suppliers to pledge not to use Claude on Pentagon-related projects.
The startup sued the Trump administration on Monday, calling the government's actions "unprecedented and unlawful."
Anthropic said in its filing that the company is suffering "irreparable harm" and that contracts worth hundreds of millions of dollars are in jeopardy.
"This move is not intended to be punitive," Michael stressed Thursday.
He added that Anthropic has a "large commercial business" and that revenue from the U.S. government represents only a "tiny fraction."
Michael also dismissed Anthropic's claims that the government proactively contacted companies to ask that their products not be used, calling it a "rumor."
“As long as they don’t enter our supply chain, the Department of War will not contact companies to tell them what to do,” he said.
Founded in 2021 by a group of researchers and executives who left OpenAI, Anthropic is best known for its Claude family of models and has made early progress working with large enterprise customers, including the U.S. Department of Defense.